 -----------
| Changelog |
 -----------

***** Tagged with galax-1-1 for GODI O'Caml release 3.10 *****

March 24, 2008 - Mary

  o Typing_call 

    - Fixed insidious bug in Typing_call in which static type checking
      for binary overloaded function call did not match on all
      possible pairs of types, thus excluding valid output types,
      which lead to type-based rewriting incorrectly converting the
      overloaded/polymorphic call to a monomorphic call.

March 21, 2008 - Jerome

  o Typing:
     - Fixed wrong signatures for op:is-same-node, op:node-*.

***** RE-Tagged with galax-1-0 for GODI O'Caml release 3.10 *****

March 12, 2008 - Jerome

  o Installation:
     - Now installing the extensions examples directory.
     - Added missing META.in.

  o Serialization:
     - Fixed a few issues with serialization (showing up while doing
       the tutorial).

  o Documentation:
     - Added table of contents back.
     - Pass on installation chapter. Added section on how to run the
       W3C XQuery test suite.
     - Removed the 'accessing and storing XML' part relating to Jungle
       and datamodel, empty or obsolete.
     - Added back chapter on APIs.
     - Shuffled the chapters in the reference part (alignment is
       renamed conformance and comes first, 'general' last).
     - Removed tutorial part on serialization as well-formed,
       buggy. Added text for more useful 'standard' mode.
     - Pass on tutorial chapter added a short tutorial section on
       printing compilation phases.

  o Testing:
     - Fixed the default config file to not point to the
       Galax-specific test suite. (for distribution).

February 21, 2008 - Jerome

  o Release:
     - Changed the status to 'Stable'

  o Documentation:
     - Pass on Chapter 1. 

February 21, 2008 - Jerome

  o Debian:
     - Folded in Stefano's patches.

  o Documentation:
     - Re-enabled documentation generation.
     - Fixed some formatting issues.

***** Tagged with galax-1-0 for GODI O'Caml release 3.10 *****

February 12, 2008 - Mary 

  o Ran XQuery Test Suite

  o Fixed the APIs 

January 29, 2008 - Mary [DXQ]

  o Galaxd

    BUG ALERT:
    Added default configuration: Conf.set_materialize_tables Conf.Always;
    Changes Conf.materialize_tables_kind from Analysis (i.e., always
    do analysis) to Always, due to quadratic memory usage bug somewhere
    in Alg_path_structutil.path_sequences_with_disjoint_roots

  o Typing_expr
    
    Bugs in typing Eval and Insert.

  o Webgui

    Added new graph-drawing for <node> and (sub)<tree>
 
  o Top_util

    load_graph raises error if named graph does not exist.

  o Procmod_phases
  
    Permitted code selection phase to be optional

January 14, 2008 - Mary [DXQ]
  o Playing with Willow

December 5, 2007 - Mary

  o Java API
    - Changes to get Java API to link
    
    TODO: align examples/galapi/java_api


November 26, 2007 - Mary 

  o C API

    Makefile.galax: Had to add "-lrt" and CONF_C_LIBPCRE to C-API link
    opitons.  

November 16, 2007 - Mary [DXQ]

  o Changed syntax of box-closure expression to:
      for server P implement Q box { E }
    
    Assumption is that closure will be evaluated at some server P
    that implements interface Q, but we do not know the ultimate
    destination. 

    NB: Now that we ship arbitrary code in closure, we have even more
    motivation to distinguish between functions that are (1) location
    dependent and/or (2) updating.  We ue 'updating' modifier to
    describe both, which conflates meaning.  This will eventually bite
    us when we want to optimize an expression for a particular server.

  o Updated Willow example
    - Changes friend for a domain based on latency
    
November 7, 2007 - Mary [DXQ]

  o Changed glx:gettime() to use Unix.gettimeofday() 
  o Fixed bug in normalization & code selection for 
    'for server P box { E }' to have same semantics as the remote
    execute expressions.

November 6, 2007 - Mary [DXQ]

  o Changed config files to use uniform graph format for neighbors and
    latencies.

    See examples/dxq/narada/latencies.xml & examples/dxq/willow/topology.xml

  o Webgui 
    Added "-t graph.xml" option to load initial neighbor topology from file. 

October 25, 2007 - Jerome

  o DXQ:
     - Fixed precedence conflicts in the parser for boxstuff
       expressions.

  o Anonymization:
     - Factored out occurrences of 'Galax' in strings into
       a single constant, to facilitate anonymization. (SIGMOD stuff).

***** Tagged with galax-0-9-9 for Debian Release to Stefano Zacchiroli <zack@bononia.it> *****

October 24, 2007 - Mary 

  o Documentation: 
    
    Woefully incomplete vis a vis Updates, XQueryP, APIs, etc. 

  o Code_selection:

    Fixed insidious bug in code selection for references to imported
    variable or imported function: Incorrectly returned
    code-selection-context of imported module instead of importing
    module.

  o configure script

    Disabled C and Java APIs until we figure out what we want to do
    with them.

  o OCaml API examples

    Aligned with new processing model 

October 23, 2007 - Jerome

  o Unicode:
     - Added support to change the location of Unicode character maps
       through an environment variables.

October 16, 2007 - Mary [DXQ]

  o Typing_expr:

    Added default typing rules for update expressions so that
    sub-expressions are also typed. 

  o Conf.mlp:

    XQueryP is containted in DXQ: Fixed is_xqueryp() to account for
    inclusion. 

  o Narada: 

    Initialized global variables to be non-empty, to avoid this bug:

    [TODO : There is a nasty stack overflow bug in Subtyping_gluskov.includes:
     Internal Error: Subtyping_gluskov.includes 
       small = element route of type xs:untyped, text?, element route of type xs:untyped (min 0) (max 0)
       big = attribute *  of type, (...rest of item()*...)
    ]

***** Tagged with galax-0-7-10 *****

October 15, 2007 - Mary 
  o Release:
    - Set config/Makefile.galax RELEASE number to 0.7.10

    Should this be 1.0...finally?!

October 15, 2007 - Mary [DXQ]

  o Added two new constructs to DXQ: boxing an expression in a closure
    and evaluating a boxed closure.  This expression:

      for server S box Expr 

    constructs a closure for Expr in the context of server
    implementation S.  The closure is an XML tree.  This expression:
    
      eval box Expr              

    unboxes the closure in Expr and evaluates the closure.  A
    rough definition of at-server-do (similarly, for-server-return) in terms
    of box and eval:

        at server S do E 
      === 
        let $v := for server S box E return
        at server S do { eval box $v }
 
    Semantics of these constructs still a bit fuzzy.  

  o Implementation required changes all the way from parser to code
    selection and plan serialization.

October 10, 2007 - Jerome

  o Join detection:
     - Disabled rewriting of "some" into "let" to facilitate join
       detection in a few specific cases (affects q10 in XMark).
     - Changed how XMark query 10 is written (manual inlining of two
       variables) to avoid issues with independance checking
       interacting with optimization.

September 20, 2007 - Jerome

  o Location hints:
     - Fixed schema_import.mlp to use the new location hint calls.

September 20, 2007 - Jerome

** Please run configure again due to changes to detect Camomile's version **

  o Galax Compilation:
     - Fixes to ./configure to support camomile's version.
     - Upgraded to Ocaml 3.10
       (changes in compilation due to new camlp4)
     - Added support for Camomile 0.7.* on top of Camomile 0.6.
       (changes due to new library names)
     - Now forcing use of Netsys, coming with latest ocamlnet.

  o Join optimization:
     - Another fix to the compiler, to properly clean-up selec
       clause for join detection.

Sept 25, 2007 - Mary [DXQ]

  o Grammar + Xquery_ast/Xquery_core_ast

    - Added optional "implements URI [at Expr]" clause to module declaration:

      "module" "namespace" NCNAME "=" URI ["implements" URI ["at" Expr]];

    - GLXTS : Galax Test Suite

      Added "Interfaces" test group for testing interfaces & modules
 
  o Processing_context
    - Location hints are completely managed in this module
    - Now distinguish between interface, module and schema location
      hints. 

Sept 19, 2007 - Mary [DXQ]

  o Grammar + Xquery_ast/Xquery_core_ast

    - Added global server declaration to prolog:
        "declare server NCNAME implements NCNAME at Expr;"
      aligns with :
        "let server NCNAME implement NCNAME at Expr return Expr"
    
      It doesn't match other declarations that define namespaces:
      e.g., "module namespace X = URI";

  o Namespace_names + many normalization modules

    Added support for interface and server prefixes:
    
    - Extended Namespace_names.prefix to distinguish between interface
      and server prefixes. 
      (Pre-requisite to implementing function shipping correctly.)

    - Fixed scoping of declare/let-server prefixes and added static
      checks that server-functions are actually defined in
      corresponding interface.

    Open questions/TODO
    ~~~~~~~~~~~~~~~~~~~
    x Should globally declared server in a module be imported into
      importing module?  Right now, it's not, but it probably should
      be. 

    x Norm_ident_top/expr are incomplete vis a vis declare-server and
      let-server.

  o DXQ Examples 
    Use new declare-server syntax

  o General clean-up of Web GUI 

    Webgui:
    - Flashplayer interface is now primary GUI
    - Excised all SVG code and event reporting from HTML interface
    - HTML interface only useful for quick access to servers'
      programs, debug logs, and sending queries
    - Still not very clean code, but there's a lot less to grapple
      with

September 15, 2007 - Jerome

  o Streaming:
     - Disabled stream inside tuple fields because of some serious
       problem with the streaming analysis...

September 13, 2007 - Jerome

  o Compilation:
    - Fixed bug in the compilation of where clauses, affecting some
      queries with conjunctive/disjunctive predicates.

  o Processing model:
    - Added a debug option for the compiler.

  o Namespaces:
    - Fixed expected results for Usecase NS, and regression tests.

September 12, 2007 - Mary

  o General clean-up of Web GUI 

    Webgui:
    - Flashplayer interface is now primary GUI
    - Excised all SVG code and event reporting from HTML interface
    - HTML interface only useful for quick access to servers'
      programs, debug logs, and sending queries
    - Still not very clean code, but there's a lot less to grapple
      with

September 10, 2007 - Mary

  o General clean-up of DXQ server code 

    Galax_server_util:
    - Contains server utility functions shared by Galax_server,
      Galaxd, Webgui: server logging, signatures for server functions,
      etc.

    Galax_server:
    - Consolidated GADS of replicated code: 
      queries from peers, queries from GUIs, error handling
    - Defines HTTP/TCP and UDP server functions
    - Builtin DXQ functions except those depending on
      Thread/Mutex/Condition modules
    - All calls to Browser/GUI isolated here 

    Galaxd:
    - Thread server definition and definitions of thread primitives
      exposed as builtin DXQ functions.

    - Mainline for server start-up: reads directories, parses files,
      etc., but does not handle query evaluation or call GUI.  All
      query evaluation and GUI I/O in Galax_server.

September 7, 2007 - Mary

  (Warning: NS,SGML,XMP usecases currently failing regression)

  o General clean-up of HTTP code

    Http : 

      All HTTP functions (client-side, server-side, and rpc-like) are
      now in http.ml.  This module is in a "hybrid" state: the HTTP
      get function is implemented using Http_client; all others are
      implemented using home-grown HTTP request code.

      Client-side functions only return HTTP message payloads (no
      status or headers) and an raise exception if any status other
      than 200/"OK" is received.

    Galax_http :

       Is replaced by Galax_url, which only handles URL decoding and
       user-installed schemes.

September 4, 2007 - Nicola

  o added a new function, glx:http-get-request() that (temporarily)
    implements HTTP get calls with full HTTP/1.1 features (using Netclient), 
    until we fix and unify the HTTP support.

August 30, 2007 - Jerome

  o Namespaces:
    - Added a couple of optimizations to the namespace resolution
      process.
    - Now avoiding serialization of namespace bindings that are
      already in scope.

August 29, 2007 - Nicola

  o Workflows:
    - Added support for HTML forms.

August 27, 2007 - Jerome

  o XQueryP:
    - Removed obsolete while normalization code.

August 27, 2007 - Jerome

  o XQueryP:
    - Fixed normalization of already normalized expressions with while
      loops.

August 27, 2007 - Jerome

   ** NOTE: Complete re-implementation of while loops in XQueryP. It
      has only been tested with simple kinds of while loops and
      Narada. Please report any problem. **

  o XQueryP:
    - Full re-implementation of while loops. Now preserved as such all
      the way down to the algebra. This fixes problems with variable
      scoping, and (seems to) improve performances.
    - Added proper checks in the parser for some of the XQueryP
      expressions (while, letvar, set).
    - Fixed the core pretty-printer for: while,letvar,set,imperseq
      expressions.

August 23, 2007 - Jerome

  o Join support:
    - Re-enabled support for conjunctive normal form of predicates
      during compilation of where clauses, needed to support proper
      hash/sort-join detection. [Likely disabled by mistake]

August 23, 2007 - Jerome

  o Rewritings:
    - Fixed bug in inlining inside the fn:error() function.

  o Updates:
    - (Somewhat) improved performances for insert operations. Most
      significant for "insert into" or "insert as first into".

    ** Note: insert into now behaves as insert as first!! Please use
    explicit location to override that behavior.

August 21, 2007 - Jerome

  o Testing:
    - Updated the reference for the test results.

August 21, 2007 - Jerome

  Feature Galax
  Minimal Conformance	          14555 / 69 / 14637  (99.4%)
  Optional Features
    Schema Import Feature	  0   	/ 0   / 174
    Schema Validation Feature	  0   	/ 0   / 25
    Static Typing Feature	  46  	/ 0   / 46
    Full Axis Feature	 	  130 	/ 0   / 130
    Module Feature	 	  32 	/ 0   / 32
    Trivial XML Embedding Feature 0 	/ 0   / 4

  o Oops:
    - Guarded printing of table materialization size by proper debug
      flag.

  o Namespaces:
    - Fixed bug in the fn:in-scope-prefixes() function.

  o XQueryP:
    - Added support for multiple declarations in blocks.
      In the style of: declare $x := 1, $y := 2;

  o Testing:
    - Re-generated expected plans for the Galax test suite, aligning
      with the current compiler.

August 21, 2007 - Jerome

  o Processing model:
    - Fixed issue with compilation of main modules in the context of
      the standard module. (semi-hackish)

August 17, 2007 - Jerome

  o Testing:
    - Cleaned up some dead-wood code in test_core.

  o Updates:
    - Propagated several bug fixes in the update usecases.

  o DXQ:
    - Fixed namespace issue related to I/O of query plans in XML.

August 17, 2007 - Jerome

  o Processing model:
    - Added call to compile a statement into a logical plan.

August 17, 2007 - Jerome

  Feature Galax
  Minimal Conformance	          14535 / 89 / 14637  (99.3%)
  Optional Features
    Schema Import Feature	  0   	/ 0   / 174
    Schema Validation Feature	  0   	/ 0   / 25
    Static Typing Feature	  46  	/ 0   / 46
    Full Axis Feature	 	  130 	/ 0   / 130
    Module Feature	 	  0 	/ 0   / 32
    Trivial XML Embedding Feature 0 	/ 0   / 4


*** WARNING ***

  This commit includes grammar changes to the update feature. This
  implements the new W3C grammar for updates. Users code is affected
  and should be modified accordingly.

  Here is a summary of the changes:

  do delete Expr     	      ==>   delete node Expr
                     	      or    delete nodes Expr

  do insert Expr..   	      ==>   insert node Expr..
                     	      or    insert nodes Expr..

  do rename Expr..   	      ==>   rename node Expr..

  do replace Expr..  	      ==>   replace node Expr...

  do replace value of Expr..  ==>   replace value of node Expr...

  transform copy $x := Expr.. ==>   copy $x := Expr..

*** WARNING ***

  o Testing:
    - Fixed text comparison in the test harness to be equivalent to
      fragment comparison.
    - Precompiling the queries producing the test results, should lead
      to major improvement in running time for the test harness.

  o Updates:
    - Implemented the new update grammar. (See changes above).
    - Reflected new update grammar in usecases and dxq code in
      ./examples directory.

  o Pretty-printer:
    - Fixed pretty-printing of interface files, now printing
      function signatures properly. [For Nicola]
    - Fixed pretty-printing of updates, following new grammar as well
      as fixing bugs in printing of core updates.

  o Namespaces support:
    *** Still more work needed -- Should be functional though ***
    - Pretty major re-implementation of namespace propagation in
      constructed elements. Notably properly implements in-scope namespaces.
    - Fixed lurking bug in namespaces for the produced test results.
    - Fixed the NS usecase to produce and check against the correct
      results (finally...).

  o Processing model:
    - Added data structures for compilation to algebraic plans,
      without doing code-selection. Still more work needed.

August 15, 2007 - Michael

  o Galapi:
    - Added validate_document Caml API function

  o Ast_printer:
    - Print_type_core's output should now be parseable as an XQuery type

  o Ast_walker:
    - Added fold_over_cxtype

  o Config:
    - Combiner the printer and walker in Makefile.galax so allow for 
      Print_type_core's use of fold_over_cxtype

  o Datamodel:
    - Fixed comment

August 9, 2007 - Jerome

  o XQueryX:
    - Added support for module interfaces in xqueryx2xquery.

  o Interfaces:
    - Fixed pretty-printing of module interfaces.

  o Galax Compilation:
    - Added missing dependencies for some target executables to the Galax
      library.

August 9, 2007 - Mary

  o Galaxd: 

    Fixed bug that snuck in when added support for async calls to self
    to use TCP

August 8, 2007 - Mary

  o Planio : Added planio serialization for let-server-implements and
    at-server-do expressions so that they can be shipped as remote
    expressions.

    Static scoping of function calls is still not quite right.  

August 7, 2007 - Mary

  o Galaxd 
    - Fixed bug that did not handle asynchronous calls to the local
      server (aka "self") correctly.
   
  NB: In the process of consolidating and cleaning up code in Galaxd,
  Galax_server, Http, and Webgui.

August 3, 2007 - Jerome

  o XQueryX:
    - Added support for library modules in xqueryx2xquery.
    - Added support for batch processing of modules in xqueryx2xquery.
          Usage: xqueryx2xquery -batch q1.xqx ... qn.xqx
          Produces: a set of files: q1.xq ... qn.xq
    - Replaced XQuery parser by XML parser in xqueryx2xquery.

August 3, 2007 - Nicola
  
  o Added a new built-in function, glx:doc-of-string() for
    parsing a string into an XML document
  o Modified glx:http-request() to return only the content
    of the message, without headers.  

August 2, 2007 - Nicola
  
  o Added a new built-in function: glx:http-request() for
    sending a generic HTTP request 
  o Enabled zerod to send the content of the HTTP message to
    galaxd.

August 1, 2007 - Jerome

  o Normalization:
    - Resolved inconsistency when dealing with schema import between
      Norm_ident_top and Norm_top.

  o Rewriting:
    - Properly hooked-up rewriting support for the XQuery Prolog.

August 1, 2007 - Mary

   o Code_execute (See July 13, 2007)

     - Sigh.  Re-enabled optimization of Execute expressions to
       local server, which simply evaluate the remote expression
       locally.  Evaluating closure in place and exporting/importing
       closure should be absolutely equivalent, but they are not
       behaving as such:

       Willow example throws dynamic type errors when optimization is
       enabled; Narada doesn't run when optimization is
       disabled. Gotta figure this one out.

August 1, 2007 - Jerome

  o Toplevel:
    - Added flag to turn warnings on/off ('-print-warnings').

August 1, 2007 - Jerome

  o Rewriting:
    - Started cleaning up rewriting support for the XQuery Prolog.

  o XQueryX:
    - Made direct embedding of XQueryX elements in queries an option
      (off by default). This restores the conformant behavior of the
      processor in the presence of those elements.

      You can turn the behavior back by using '-embed-xqueryx on'.

August 1, 2007 - Jerome

  o XQueryP: Fixed bug in syntax of while expression.

  ******************************************************
  WARNING: while expr return ... is no longer supported.
  You have to write: while (expr) return expr
  ******************************************************

  o XQuery: Added missing modules! Ugh too...

Aug 1, 2007 - Mary
  o schema/schema_import.mlp : incorrectly updated the generated file
    schema_import.ml, then put it back in CVS. Ugh.

    removed schema_import.ml from CVS, again.

August 1, 2007 - Jerome

  o XQueryX: First-cut support for XQueryX with trivial embedding
    in Galax. It supports the following features:

     (1) Compiling XQueryX to XQuery using a command line tool called
         xqueryx2xquery.

     (2) Running XQueryX with galax-run with the flag '-syntax
         xqueryx'.

     (3) Embedding XQueryX expressions directly within XQuery.

  (1) and (2) work on whole modules, while (3) only works with
  expressions since it has to be integrated with the rest of the
  grammar.

  o Processing model:
     - Fixed ugly bug in the way parse handlers were built. Now
       properly taking configuration into account.

August 1, 2007 - Mary

  o XQuery Test Suite Modules Group

    - Normalization : Numerous changes to pass XQTS Modules group

    - Location hints: schemata, modules and interfaces may have
      multiple associated location hints.
  
      Their meaning for modules and interfaces , i.e., alternative
      locations or locations of multiple fragments, is
      implementation-defined.  However, the XQTS *requires* that they
      mean locations of multiple fragments, which are then
      concatenated during normalization.

    - Processing_context: Added flag indicating whether to merge
      module fragments from multiple locations.  false, by default. 

    - regress/testconfig-tmpl.xml : added Modules group

  o DXQ examples : 
    - Align with new module requirements

July 31, 2007 - Jerome

  o Parsing:
     - Fixed another lurking bug in the lexing of sequence types,
       transitioning to the wrong state... (Was tripped by e.g.,
       instance of or typeswitch inside a constructor).

       [Replaced old hack by better working new hack...sigh

        Queries to test:
        <a>{1 instance of item()}</a>
        <a>{1 instance of item()*}</a>
        <a>{1 instance of item()(::)*}</a>
        every $a as item()* in (1, 2), $b as item()* in $a satisfies $b]

  o Prolog:
     - Fixed another lurking bug in cyclic variable dependency
       checking (not clearing up the previous variable table in
       parallel branches of the dependency graph).

  o XQueryX:
     - Started to prepare support for Trivial Embedding. Added hook in
       normalization.

July 27, 2007 - Nicola

  o Zero Daemon: Moved the logic for dispatching messages to DXQ 
    functions from the daemon into a special function called
    dispatch() that is to be constructed in each module. The 
    namespace prefix of dispatch() is passed as a command-line argument 
    to zerod.

July 24, 2007 - Nicola

  o Zero Daemon: Implemented the full protocol between the proxy
    and the DXQ workflow server, including management of newly
    enabled activities, except for the case of "pick" and
    "while" activities.

July 21, 2007 - Nicola

  o Workflows: 
    First version of an HTTP server providing an interface for 
    ProjectZero workflows and acting as proxy towards the DXQ
    server running the workflow script.
    The current version does not support yet receive activities
    that are not enabled from the beginning. 

July 20, 2007 - Jerome

  o Prolog:
     - Fixed bug in cyclic variable dependency checking (not handling
       multiple occurrences of the same dependence).


July 19, 2007 - Mary & Trevor [DXQ]
   o Willow: added Willow:replace_attributes() function to work-around
     replace-node bug in Galax 
   
July 13, 2007 - Mary & Trevor [DXQ]
   o Willow : kind of, sort of working

  o DXQ:
     - Added support for conditional variables.

July 18, 2007 - Jerome

  Feature Galax
  Minimal Conformance	          14528 / 96 / 14637  (99.3%)
  Optional Features
    Schema Import Feature	  0   	/ 0   / 174
    Schema Validation Feature	  0   	/ 0   / 25
    Static Typing Feature	  46  	/ 0   / 46
    Full Axis Feature	 	  130 	/ 0   / 130
    Module Feature	 	  0 	/ 0   / 32
    Trivial XML Embedding Feature 0 	/ 0   / 4

  o WSDL:
     - Removing wsdl_load.ml at cleanup-time, since it's being
       generated.

  o Prolog:
     - Added check for cyclic definitions of global variable in
       modules.

  o Testing:
     - Added group testing for the XQuery Appendices.

July 18, 2007 - Nicola

   o WSDL support: changed AST creation of attributes to always use
     EText constructors


July 13, 2007 - Mary

   o Xquery_algebra_ast.mli + all files that depends on Algebra AST
     - Changes names of AST nodes for imported variables & functions. 
       Conflated interfaces with imported 

   o Changed Pervasives _back_ to an XQuery module from an interface,
     because it really is a module whose functions are built-in.
     Sorry for churn on this.
 
     - Removed stdlib/pervasive.xqi 
     - Added glx:string_of_item
   
   o Planio:
     - Fixed bug in parsing of ElementKind and AttributeKind

   o Code_execute:

     - Temporarily disabled optimization of Execute expressions to
       local server, which simply evaluate the remote expression
       locally, because there is an insidious bug when its called.

   o Parse_xquery:
     - Changed parsing of interfaces so that variables and functions
       are correctly tagged as coming from interfaces.  Previously,
       did this with an ugly hack in normalization. 

   o Galaxd & Galax_server:
     - Added exception-handler argument to tcp_server and udp_server
       so that internal errors and async errors can be reported to the
       Gui.

July 12, 2007 - Nicola

  o Normalization of XQueryP functions: Functions with side effects
    need to be annotated with the "updating" flag. The compiler was
    checking this annotation only for XQuery! programs, but not for
    XQueryP. I corrected it.

July 10, 2007 - Jerome

  o Pretty-printer:
     - Fixed pretty-printer for blocks inside function declarations.
     - Fixed pretty-printer for fixed attributes in elements! [See
       Nicola's thread]

July 6, 2007 - Jerome

  o DXQ:
     - Fixed bug in variable access analysis not taking the variable
       in the [from server ... at server ...] expressions into
       account.

  o Optimization:
     - Turned variable inlining on by default.

July 5, 2007 - Mary [DXQ]

  o Factored dxq "utility" functions in glx: namespace into
    examples/dxq/dxq.xqi with new namespace dxq:

    Need to update DNS and Narada to use this module.  They won't work
    until updated.

  o Changed Galaxd to implement dxq: functions

  o Added some bit-manipulation functions 

  o Worked on Willow definition

July 5, 2007 - Nicola

  o WSDL: fixed serialization of WSDL elements and attributes
    (there were extra quotes and curly braces)

July 5, 2007 - Jerome

  o Normalization:
     - Fixed normalization of function calls whose expected type is
       xs:anyURI (wrongly casting to xs:string).

  o Optimization:
     - Added clean-up optimization rule for inverse MapFrom/MapTo:
         Map{#x}(Map{[x:ID]}(Op)) == Op

June 30, 2007 - Mary

  o Modules: Fixed bugs in normalization.  

    - References to functions and variables in an imported module were
      normalized as references to funcs/vars in an imported interface, 
      which wraps them in DXQ let-server expressions under the
      assumption that they are remote references.

      Need to distinguish between these references in Core but
      subsequent phases can/do treat them equivalently.

    - Extraction of a module's interface neglected to include
      namespace declarations.

June 28, 2007 - Nicola

  o WSDL: Updated the WSDL parser to make it work again with the 
    latest changes in Galax.

  o Removed wsdl_load.ml from the repository, as it is supposed to
    be generated from wsdl_load.mlp

June 7, 2007 - Mary [DXQ]

  o WebGUI: Simulated network delays were not being injected into UDP
    messages.  Now they are.

  o Galax_server: returns XML results in HTTP message with xhtml+xml
     header. 

  o Galaxd: latencies/topology file are now loaded from simulation
    directory. 

  o Narada example: added simple multicast and unicast messages.

June 6, 2007 - Mary [DXQ]
  o Lots of last minutes changes

  o WebGui : Yet another logging bug

June 5, 2007 - Mary [DXQ]

  o WebGui : Added labels to neighbor edges; tweeked DNS example

May 30, 2007 - Mary [DXQ]

  o WebGUI 
    - Fixed bug that caused clock skew between webgui and gui.swf.

May 29, 2007 - Mary [DXQ]

  o DNS example : coded up scenarios for live demo

  o WebGUI logging
  
    At start-up, webgui creates directory $TMPDIR/webgui-<processid>
    and writes all logs into this directory.

      webgui.log contains complete playback log to be run by
      tracegui.swf.

      Content of each message is written to a file with name:
      <clock-tick>-<source>-<target>.msg 

May 29, 2007 - Mary [DXQ]

  o Implemented Flashplayer/WebGUI messages:

    - Display program
    - Broadcast/unicast query
    - Added "-msg true" option to enable reporting of message content
      to Flashplayer, which are displayed when player is paused and
      "Details" button is clicked.

      [By default, messages are written to files in directory
       /tmp/webgui-<processid>/ and read on demand when "Details"
       button is clicked in Flashplayer (still working on this feature).]

    - Fixed start-up and timing bugs: nodes now appear on start-up and
      messages appear "on time".

May 25, 2007 - Mary [DXQ]

  o Temporary bug fix : In galaxd, allow_streamed_tuple_fields is
    false, because calls to the built-in glx:gui-* functions are
    failing when streaming is enabled.  Don't know why.

  o Variety of changes to WebGui to handle messages from flashplayer

May 23, 2007 - Jerome

Grammar changes summary:

  declare function f { { declare $x ... } };
    ==> declare function f { declare $x ... };

  let server ... return { Expr }
    ==> let server ... return Expr

  from server T return { Expr }
    ==> from server T return Expr

  at server T do { Expr }
    ==> at server T do Expr

  o XQueryP/DXQ:
    - Fixed bug in normalization of blocks, not registering variables
      declarations properly.
    - Fixed parser for function declarations. Now using directly a
      block instead of requiring extra curly braces.
    - Fixed issue with parsing of DXQ expressions. Removed unnecessary
      curly braces.
    - Fixed issue with lexing state when parsing DXQ expressions
      inside constructors.
    - Fixed parsing precedence conflict for 'at server' expression.

  o Optimization cleanup:
    - Removed the '-pruning' option, unused. Pruning code is still
      called from the staircase join implementation.
    - Turned '-code-selection-by-physical-type' on by default, removed
      the corresponding option from the command line. Removed the
      corresponding old versions of code selection.
    - Turned '-allow-streamed-tuple-fields' on by default.

  o Code selection:
    - Fixed code selection for tuple field assignment. Was doing it at
      run-time instead of at compile time!!!
    - Fixed type assertion support in tuple construction over lists of
      nodes.

May 22, 2007 - Mary [DXQ]

  o Updated Narada example with new syntax. 

May 21, 2007 - Mary [DXQ]

  o Implemented syntactic sugar for references to remote variables &
    functions, e.g., remote function call is the same as a synchronous
    call to function at the remote server.

      T:foo() => from server T return { T:foo() }

    - Reports an error when remote function/variable is not in the
      scope of an implementation namespace.

    - Updated DNS example w/ new syntactic sugar. 

  o TODO: 

    - Need real syntax for interface declarations.  "external" is
      overloaded: Used both for genuinely external functions and
      imported functions.

    - Import module/interface: exclude symbols that are not in
      module/interface's namespace. 

    - Physical types for external and imported variables: Right now,
      they are completely materialized XML values.

    - Need help from Jerome on weird parsing problems with braces. 

   o APIS:
     - examples/caml_api/test.ml
     - C or Java APIs

May 18, 2007 - Mary [DXQ]

   o Implemented new DXQ syntax (note curly braces!)

     let server NCName_1 implement NCName_2 at Expr_1 return { Expr_2 } 
     from server NCName return { Expr }
     at server NCName do { Expr }

   o Migrated DXQ programs in examples/dxq/test and examples/dxq/dns
     to use new syntax. 

     - Working on Narada example

May 16, 2007 - Mary

  MODULES and INTERFACES 
  *** PLEASE READ THE WHOLE MESSAGE ***

  o Deep changes to support separate module interfaces and module
    implementations.

    - Compilation/evaluation now occur within context of a "compiled
      program".  (In new module code_selection/Compiled_program_units)

      A compiled program consists of imported module interfaces,
      imported library modules, and an optional main module.

    - All program units (statements,prologs,modules) are compiled in
      context of a compiled_program unit.

    - New pre-processing phase (before normalization), computes
      transitive closure of imported library modules.  Used to
      populate compiled_program unit.

    - Recursive module imports (as per standard) are prohibited. 

   o Changes to ASTs: 

     We distinguish between "external" functions/vars (which are
     defined in external programming environment) from "imported"
     functions/vars (which are defined in imported XQuery modules)

     New kinds of declarations: 
     Xquery_ast : EFunctionInterface and EVarInterface 
     Xquery_core_ast : CEFunctionInterface and CEVarInterface 
     Xquery_algebra_ast : AOEFunctionInterface and AOEVarDeclInterface

   o Importing interfaces and modules

     Module interfaces are imported during normalization.  Variable
     and function declarations in the Core rep of the imported
     interface are merged into the Core rep of the importing module.

     If a module does not have an interface, an interface is created
     for it during the pre-processing stage. 

   o New interface stdlib/pervasive.xqi replaces stdlib/pervasive.xq
 
     - Removed stdlib/pervasive.xq 
     - NB: new file suffix ".xqi" for module interfaces. 

   o Code selection

     References to imported variables and functions are resolved
     during code selection of the importing module.  

     The code_selection_context for a module points to the
     code_selection_contexts of the modules that it imports to resolve
     references to imported variables/functions.

   o What works:

     - Usecase, XQuery testsuite, and Galax regressions all pass

   o What doesn't work:

     - examples/caml_api/test.ml
     - C or Java APIs
     - Any DXQ program 

May 8, 2007 - Mary

  o TODO: Change pervasive.xq to an interface and don't give it
    special treatment in procmod_compiler.  

May 2, 2007 - Mary

  Changes to prepare for separate compilation of modules:

  o Processing_context:
    - processing_context now contains STATIC context for 
      PROGRAM unit, i.e., configuration options and values.

    - module_processing_context contains STATIC context for
      MODULE unit & is contained in norm_context.
      name_generators moved here from processing_context.

  o Algebra_context: 
    - algebra_context contains DYNAMIC context for PROGRAM unit. 
      alive_documents moved here from processing_context. 

April 25, 2007 - Jerome

  o Parsing:
    - Replaced "declare" "server" token by "let" "server" token.

April 23, 2007 - Mary [DXQ]
  o Narada example : Got rid of workaround for syntax bug

April 23, 2007 - Jerome

  o Parsing:
    - Added missing "implements" token for DXQ.

April 23, 2007 - Jerome

  o Parsing:
    - Fixed bug in XQueryP parser. Lexing states for opening blocks
      not transitioned properly.
    - Added DXQ tokens in the lexer.

April 20, 2007 - Mary [DXQ]

  o Narada example : workaround for syntax bug

April 10, 2007 - Jerome

  o Parsing:
    - Fixed lexing transition states for updates and XQueryP
      tokens. Bug only showing up when popping the state for the curly
      braces in direct constructors. The new parser has been checked
      against the ULTF usecases and Nicola's bug example. [Bug
      reported by Nicola]

April 9, 2007 - Mary [DXQ]

  o Webgui fix
  o DNS : implementing a variety of resolution algorithms

April 9, 2007 - Mary [DXQ]

  o Changed format of Webgui output to XML

March 21, 2007 - Mary

  o Galax_server: 
    - Removed log of query-response size that was growing without
      bound!
    - Drip, drip, drip: other small memory leaks abound...

March 18, 2007 - Mary

  o Evalution_expr

    - Fixed two long-standing problems in evaluation of tail-recursive
      functions, resulting from compilation of XQuery while loop:

      * Slow growing memory leak in O'Caml heap due to recursive
        evaluator in algebra_execute_aux not being tail recursive.

      * Unbounded growth of O'Caml stack due to Galax evaluator not
        "unrolling" tail recursive calls to the correct Galax stack
        frame.

    ***NOTO BENE***
    Seemingly innocent changes to the core recursive evaluator in
    algebra_execute_aux can cause the O'Caml compiler to yield a
    *non-tail-recursive* function. 

    The code is well documented, but DO NOT CHANGE this code without
    consulting Mary or Jerome first.

  o Added glx:livewords() 
    - Function executes an O'Caml major
      collection then returns number of live words in heap.  Useful for
      debugging, so I left it in.

March 9, 2007 - Mary

  o AST fold operators
 
    - Changes Ast_walker_fold.compute_cexpr to Ast_walker_fold.fold_over_cexpr
    - Added Optimization_walker_fold.fold_over_algop
    - TODO:

      Define Optimization_judge.{has_side_effect_internal,has_non_trivial_snap_subexpr, 
      subexpr_has_side_effect,contains_update} all in terms of Optimization_walker.fold_over_algop.
 
  o Datamodel TODO:

    - Noted recursive functions in Cursor.ml that are not *tail
      recursive*.  Must be fixed.

March 9, 2007 - Mary (DXQ)

  o Code_execute

    Added "label" to remote query (name of first user-defined function
    called in plan, which corresponds to "message" name) to report to
    Webgui.

  o Webgui

    Added "-log [false/true]" option to log every message in
    Javascript Object Notation (JSON) for flash-movie generation. 
   
March 8, 2007 - Mary (DXQ)
  o More changes to Galax_server and Webgui. 
    
    Added -drop option for Galax_server and/or Webgui to drop
    non-status messages (i.e., asynchronous inter-server messages), 
    but to record/show status messages sent from application. 

March 2, 2007 - Mary (DXQ)
  o Playing with graph drawing

March 1, 2007 - Mary (DXQ)

  o Galaxd & GUI 
    
    - XQuery GUI report functions: changed names to glx:gui-*
    - Added new GuiNode message "g" to print a graph
    - Added glx:gui-status, which takes graph of 
      (source, target, label) edges to print.  XML type:

      element graph { 
        element edge { 
          attribute s of simple type xs:string,
          attribute t of simple type xs:string,
          attribute l of simple type xs:string
         }* }

  o Webgui

    Added "-drop" option to drop all messages and node-reports except
    "!" (user request), "x" (error), and "g" (graph status).

  o Top_util.InternalQuery

    Three modules apply internal queries : Test_core, Galaxd, Webgui.

    Moved common code for evaluating internal queries (e.g., setting
    up context, boxing/unboxing XQuery values, etc.) into
    Top_util.InternalQuery.

March 1, 2007 - Mary (DXQ)

  o Galaxd: 
   
    Added "-t" option to galaxd to load file containing pair-wise
    latencies between servers.  See
    examples/dxq/narada/latencies.xml. 
  
    Useful for approximating network delays.  Right now, just
    latencies.  Could be generalized to network topology.

February 28, 2007 - Mary (DXQ)

  o Clean-up of Galax Server code
    
    - New module Galax_server_util:

      Factorized functionality shared by Galax_server, Galaxd, and
      Webgui into Galax_server_util.  Contains: 
        * Server UDP defaults 
        * Server-name management (maps from virtual to actual names)
        * Shared GUI utilities

    - Moved file and directory management utilities into Gmisc.

    - Moved DXQ options into Top_options.

    - Processing_context contains "short" hostname and port of server.

February 21, 2007 - Jerome

  o Options:
     - Moved the shebang options to some more logical places in the
       options list.
     - Removed option for streamed variable, now always allowing
       potential streaming for variables.

February 21, 2007 - Nicola

  o Rewriting:
     - Added rewriting rule for removing CECopy from CEInsert when
       source expr. is an elem. constr. (It was previously done
       at normalization)

February 12, 2007 - Jerome

  o Testing:
     - Fixed small bug in Makefile.

February 12, 2007 - Jerome

  o Code clean up:
     - Removed obsolete modules: Factorize_update, Factorize_util.

  o Galax Test Suite:
     - Added a first version of the Galax test suite in the CVS
       repository. This currently contains a few tests for: plan
       stability of the use cases, and join detection tests.

  o Testing:
     - Finalized support for comparing query plans in the test
       harness.
     - Added an option to automatically generate expected results in
       case the file for that expected result is missing.

       Usage: [galax-test -generate-results] to generate results.

  o Toplevel:
     - Added three plan conversion utilities for convenience.
        [xquery2plan] generates a plan from an XQuery. **
        [xquery2xmlplan] generates an XML plan from an XQuery.
        [xmlplan2plan] generates a plan from an XML plan.

      **you can turn optimization on by writing -optimization on.

  o Updates/Parsing:
     - Removed 'snap delete' etc. variants for the update
       operations. Fixing shift/reduce conflicts. (e.g., WITH could
       parse as either a do replace or a snap replace).

       The changes are:
         MapFromItem[$x]{P1}(P2) --> Map{P1}(P2) where occurrence of $x
         in P1 is turned into ID.
         MapToItem{P1}(P2) --> Map{P1}(P2)
         INPUT --> ID
         snap { P } --> Snap(P), same for Delete,Replace,Insert,Rename.

  o Pretty printer:
     - Fixed the algebraic pretty-printer to align with the current
       formalization of the algebra [Ghelli,Onose,Rose,Simeon].

  o Compiler:
     - Consolidated name creation (variables, tuple fields, etc.)
       throughout the compiler. This is now handled by a proper module
       [Namespace_generate]. This is important notably for the
       regression tests to ensure stability of variable names accross
       separate sets of compilation.

February 11, 2007 - Nicola

  o Rewriting:
     - Added tests for side effects in rewriting rules 
       that may change the semantics of the query when
       triggered on expressions containing side effects.

***** Tagged with galax-0-7-2 *****

February 5, 2007 - Mary
  o Release:
    - Set config/Makefile.galax RELEASE number to 0.7.2

February 1, 2007 - Jerome

  o Compilation:
     - Fixed issues after the merge [not compiling!!].

February 1, 2007 - Jerome

  o Code cleanup:
     - Cleaned up all the source file headers. Added module
       descriptions when missing, as well as CVS Id.
     - Removed obsolete modules: Optimization_rules_treepattern_old,
       Factorize_side_effect.

  o AST Walker:
     - Added support for a generic fold operation on the AST (useful
       to compute a boolean property on the AST for instance).

  o Normalization:
     - Small fix to the normalization of comparisons to re-enable join
       detection, not using let bindings for the comparator
       anymore. [hack]

  o Rewriting:
     - Added a judgment to check for side-effects, removed
       corresponding obsolete judgment in Factorization.
     - Moved the snap removal rule from optimization to rewriting,
       cleaning up the plans as early as possible.

February 1, 2007 - Mary

  o ***Warning: You need to re-run ./configure before compiling***

    Changes to configure script, config/Makefile.galax
    to align with O'Caml 3.09.3.  

    Added conditional linking of Netsys library. 
    In preparation for godi-galax package release.

January 30, 2007 - Mary [DXQ]

  o Execute operator

    - Removed hack in which default port was always "0".
      Port is now computed dynamically from host-port argument, or if
      port is missing, is determined by Galax_server.

  o Narada example

    - Fixed insidious bug in which sequence numbers were being
      compared *lexicographically* because XML values are treated as
      untyped strings after export then load.  Shows what happens when
      you don't have a semantics for export/load of typed values.

    - Added N:init() function so that nodes can be restarted with
      initial state without having to restart Galaxd processes.

    - Added individual "loop" functions so that each message-loop can be
      debugged in isolation of all others; argument is delay between
      each message. 

January 30, 2007 - Jerome

  Current testing results on XQTS 1.0.2:

  Feature Galax
  Minimal Conformance	          14514 / 110 / 14637  (99.1%)
  Optional Features
    Schema Import Feature	  0   	/ 0   / 174
    Schema Validation Feature	  0   	/ 0   / 25
    Static Typing Feature	  46  	/ 0   / 46
    Full Axis Feature	 	  130 	/ 0   / 130
    Module Feature	 	  0 	/ 0   / 32
    Trivial XML Embedding Feature 0 	/ 0   / 4

  o Normalization:
     - Now properly detecting mal-formed option declarations
       (unresolved prefixes notably).

  o Optimizer:
     - Fixed bug in selection pushdown rule, bogus when pushing down
       over a concatenation (expecting a single tuple). Bug triggered
       when variable inlining is on.

  o Namespace declarations:
     - Fixed bug: now handling undeclaration of non-declared prefixes
       properly.

  o Parser:
     - Fixed bug in lexing of typeswitch case branches.
     - Fixed wrong transition for DECLAREVALUEINDEX and DECLARENAMEINDEX.

  o Value index support:
     - Fixed code selection to first perform code selection for global
       variables before indices, allowing indices to access global
       variables.
     - Turned glx:keyref to a standard built-in Galax function,
       providing direct support for static typing, and cleaning up the
       corresponding code.

January 28, 2007 - Philippe

  o Fixed SBDO rewrites

January 26, 2007 - Philippe

  o Some more changes to sbdo optimization.
    Sbdo optimization is in flux, moving currently moving to enable
    the optimization over more than just tree patterns. This means
    propagating the state of the automaton from one tree pattern to 
    the other.

January 26, 2007 - Philippe
 
  o Fixed treejoin rewrite rule to clean up some more projects
  o Fixed label setting on Attribute SAX events, so that
    streaming aggregation functions return the correct result
  o Fixed bug in SBDO rewriting (nested tree patterns should pass on their state)

January 25, 2007 - Trevor

  o Added a broadcast action to the gui

  o Added filtering by regexp and nodes

January 24, 2007 - Mary  [DXQ]
  o Narada example: added routing tables
 
January 24, 2007 - Mary 

  o Normalization

    Changed normalization of expressions so that newly created
    functions (e.g., for while-loops) are added to global context
    correctly.  Got rid of a lot of unused, hackish code.

  o Planio 

    Added Dep="No|One|Two|Many", Ind="No|One|Two|Many" attributes to
    operator elements to permit easier parsing. 

January 23, 2007 - Mary 

  o Changes to support tail-recursive function calls.

    Code_user_defined_fn:
    Code selection for user-defined function calls now returns code
    for (default) non-tail-recursive calls and entry and exit code 
    for tail-recursive calls. 

    Evaluation_expr:
    Interpreter changed to detect tail-recursive calls and evaluate
    them within the same stack frame. 

    Status:
    Only works for XQueryP "while" loops, which are compiled into
    self-tail-recursive calls.  Need analysis to detect arbitrary
    tail-recursive calls. 
    
  o Webgui - better error messages

January 18, 2007 - Trevor

  o Tweaked the narada cluster example for a smaller network.

January 18, 2007 - Trevor

  o Added options to gui that will let us run multiple instances on
    a machine without conflict:

      -port <n>  Use <n> as the default port for servers
      -guiport <n>  Use port <n> on localhost for the gui

January 18, 2007 - Mary 

  o PlanIO:
    - Paired down XML rep of plans to something reasonable.  
      TODO: Omit function signatures from serialization of Call ops

  o Tail recursion:

    - Modified Call operators in Core/Algebra ASTs to include tag for 
      tail-recursive function calls.  

      XQueryP while loops, compiled into tail-recursive calls, are
      tagged.

    - Still working on changes to interpreter to evaluate
      tail-recursive XQuery function calls in O'Caml while loops
      (which are tail recursive!)

  o Webgui [DXQ]

    - Moved XHTML pretty printing into GUI.  

January 18, 2007 - Trevor

  o New file to start up multixterm for cluster example

January 16, 2007 - Trevor

  o GUI prints error events in red

January 12, 2007 - Trevor

  o GUI can now filter out events based on type

January 11, 2007 - Mary 

  o Optimization_rewrite:

    Modified two "Inlining" rules: singleton_product_to_concat and 
    singleton_map_inline to not inline independent operators that are
    _updating_, i.e., have mutable state. 

  o DXQ Narada example : got rid of all the P2 rules, which were
    not helpful. 

  o Pervasive: Added "updating" modifier to functions with mutable
    state.

January 11, 2007 - Trevor

  o Report errors on async exec to GUI.

January 11, 2007 - Mary 

  o Added glx:random_int which exposes O'Caml Random.int function.  We
    need to have a consistent way of exposing O'Caml standard library
    functions.

  o Started clean-up/simplification of PlanIO closures etc.

  o More work on DXQ Narada example.

January 9, 2007 - Trevor

  o Exposed GUI report functionality to XQuery in galaxd.

January 9, 2007 - Trevor

  o Cluster testbed for Narada example.

January 9, 2007 - Trevor

  o Use the same port for TCP and UDP in communications
    between the servers and gui.

  o Now use UDP for asynchronous plan execution between galax servers,
    with plans > 16384 bytes falling back to TCP.

January 8, 2007 - Mary 

  - Added examples/dxq/chord directory and pseudo-code for Chord
    protocol. 

January 5, 2007 - Mary & Kristi

  - Continued work on Narada definition.  TODO: latency.

January 5, 2007 - Trevor

  o Removed remaining fork() calls from GUI, so that it is completely
    threaded.  Added -v flag for verbose operation.  Fixed a file
    descriptor race bug.

  o Narada example now spawns a new thread on go() instead of looping
    infinitely without returning.  make clean; make run needed to see
    this change.

December 23, 2006 - Jerome

  o Testing:
     - One more step toward support for comparison of Query
       plans. Still not fully functions.

  o Compilation:
     - Removed separate compilation of single let expressions into
       AOELet, now uniformly compiling FLWORs to the tuple part of the
       algebra.
     - Changed AOELet to AOELetvar, which are now handled separately.

  o Optimization/Rewriting:
     - Fixed function inlining to handle return types properly.
     - Started moving function inlining from optimization to
       rewriting.
     - Removed obsolete rewriting_ddo_* modules.
     - Removed "Double Null Index Removal" optimization rule, both
       buggy and wrong!
     - Fixed several bugs in "Map item on Singeton tuple, inline"
       optimization rule, ignoring the presence of type assertion in
       the tuple constructor, and removing computation that may have
       side effects of raising errors.
     - Turned off "Combine Groups" optimization rule, wrong! (sigh...)

  o Code:
     - Fixed bug in NullMap operator not returning a tuple in case of
       the empty sequence.

December 15, 2006 - Nicola

  - fixed bug in alg_path_analysis: an incorrect # of subops. of
    AOERename was assumed
  - changed normalization of do-insert not to wrap the source
    in a copy statement if it is an element constructor
  
December 13, 2006 - Jerome

  Current testing results on XQTS 1.0.2:

  Feature Galax
  Minimal Conformance	          14498 / 114 / 14637  (99.1%)
  Optional Features
    Schema Import Feature	  0   	/ 0   / 174
    Schema Validation Feature	  0   	/ 0   / 25
    Static Typing Feature	  46  	/ 0   / 46
    Full Axis Feature	 	  130 	/ 0   / 130
    Module Feature	 	  0 	/ 0   / 32
    Trivial XML Embedding Feature 0 	/ 0   / 4

  o Compilation:
     - Added support for separate -galax-regression option to indicate
       where the Galax Test Suite is installed.
     - Fixed circular dependency in optimization walker/judge modules,
       resulting from the latest reorg.

  o Testing:
     - Fixed support for multiple test units, generating distinct
       files instead of overriding them. Fixed printing for multiple
       test units.
     - Added support for comparison of query plans.

  o Updates:
     - Fixed 'do rename' to properly handle xs:string and xs:untyped
       values for the name expression.

  o F&O:
     - Fixed fn:dateTime to support empty sequence parameters.

  o Pragmas:
     - Now properly resolving Pragmas QNames.
     - Fixed serious parsing and lexing issues with pragmas (wrong
       lexical state created in some cases).
     - Fixed bug, now raising an error when there is no expression.
     - Added check that the pragma QName and content is separated by
       at least a whitespace character.

  o Parsing:

    Long over-due cleanup of the lexers. This fixes two major
    problems: (1) we now have keyword tables which makes for a more
    easily extensible parser, and smaller lexing tables.  (2) we don't
    rely on long keywords anymore which means we properly implement
    comments within keywords. [ Note that it's not completely true, we
    still have a few long keywords hanging around, notably those with
    QNames inside... This still need some work, but should get us
    going. ]

     - Removing numerous unused lexing rules in some of the states,
       notably in the operator's state.
     - Moved part of the keywords used in prolog declarations from the
       operator's state to the declaration state.
     - Consolidated duplicated string literals and comments lexing
       rules to use the same lexer.
     - Added support for simple kind of look ahead in lexing handlers.
     - Added keyword tables for the operator, default, and namespace
       declaration states.
     - Removed (almost) all long keywords from the default,
       operators, and namespace declaration state.
     - Fixed bug, 'as' not recognized as a qname in the default state.

December 8, 2006 - Trevor 

  o Bug fix for galaxd thread exit
  o GUI is now threaded too

December 8, 2006 - Trevor 

  o GUI changed to display asynchronous queries and jobs differently
    than synchronous queries/jobs.
  o Galaxd submits jobs to GUI using UDP

December 6, 2006 - Mary  & Trevor 

  o Norm_top: 
    - Library modules were not normalized properly for XQueryP : 
      Generated functions were not included.

December 6, 2006 - Mary & Trevor

  o DXQ & Optimization 

    ISSUE: glx:my_address() returns DXQ server state and therefore
    cannot be pushed into a remote exec.  Same is true for all Mutex
    functions.  Would labeling them as "updating", i.e., has side
    effects be good enough?  

    We should probably have DXQ functions in their own namespace,
    labeled as "updating", so the optimizer doesn't mess with them. 

December 5, 2006 - Trevor

  o Galaxd [DXQ]

    Added external Mutex XQuery module. Interface in
    examples/dxq/narada/mutex.xq.  Implementations in Galaxd.

November 29, 2006 - Jerome

  o Joins:

     - Added pattern matching case to deal with join detection when
       the query is not type-checked (i.e., with starting fn:boolean()
       call in the where clause). This should enable join detectin in
       the presence of side-effects -- long-standing bug for Nicola.

November 21, 2006 - Jerome

  o Code cleanup:
     - Removed obsolete module: Xquery_physical_algebra_ast_util

  o Dynamic typing:
     - Fixed bug in handling of type assertions in tuple constructors.
     - Fixed bug in type assertion code for attributes.
     - Fixed bug in type assertion for processing instructions with a
       name.

  o Testing:
     - Fixed bug in test harness when handling input-URIs.
     - Removed some left-over debug printings for modules.

November 13, 2006 - Nicola

  o Created directory for the DXQ webdemo in 
    examples/dxq/webdemo


November 12, 2006 - Mary and Trevor [DXQ]

  o Fixed bugs in async exec.   Works but SLOWLY!

    - I'm generally unhappy with the ill-defined interfaces between
      Galaxd, Galax_server, and Code_execute.  This needs to be
      cleaned up with a focus on error handling.

November 10, 2006 - Mary and Trevor [DXQ]

  o Added an asynchronous variant of exec. Uninspired syntax: 
      exec { Expr1 } [ Expr2 ]
    
    In DXQ server, creates thread to eval Expr2 and immediately
    returns (). 

    - Added EExecute, CEExecute, etc. to ASTs. 
    - Added AOEASyncExecute

November 7, 2006 - Mary [DXQ]

  o Added a lift-execute-over-join rewrite when both branches of Join
    are evaluated at same location. 
  o Now handle AOEExecute correctly in Compile_annotate and Cs_code_typing_top

  TODO: See long list in https://yang.princeton.edu/wiki/Things%20to%20fix
  HOW TO RUN DEMO: https://yang.princeton.edu/wiki/How%20to%20run%20demo

October 29, 2006 - Nicola

  o Normalization: corrected normalization of transform expressions
    to register their variables in the normalization context.

October 25, 2006 - Philippe

  o Fixed handling of Project operators in streaming annotation

October 25, 2006 - Philippe

  o Added missing entry for AOECharRef in planio
  o Added missing moniker for AOEPromoteAnyString in planio
  o Added a hack rule in TPNF normalization to avoid the compiler
    to go in let-only FLWOR block compilation, since this
    interferes with join detection (why is that anyway?) 
  o Added two 'robustness' rewrite rules s.t. combinations
    of selects, project and tree patterns also get 
    picked up as joins and that projects get pushed through
    outermaps in order to pick up LO-Joins

October 24, 2006 - Philippe

  o Fixed a TreePattern rewrite dealing with variables wrongly
  o Moved algebraic let-inlining rule to the cleaning rules
    (note, once inlining is done properly at the core level,
     this rule is no longer required)
  o Disabled buggy type based core rewrite 'fs_untyped_to_atomic_type_rewrite'

October 24, 2006 - Philippe

  o Work-around for the need for re-typing TPNF' expressions

October 23, 2006 - Philippe

  o Adding removal of redundant wrapping ddo's to TPNF normalization.
    All tree patterns in XMark are now picked up.

October 22, 2006 - Philippe

  o Some fixes to TPNF and ddo-optimization modules

October 22, 2006 - Philippe

  o Made printing of sbdo-mode in tree patterns default

October 22, 2006 - Philippe

  o Aligned the TPNF rewrite rules with the ICDE paper. We now need
    only five rules, plus three Galax-specific rules to deal with
    projects and MapConcats. This clean-up is a consequence of 
    introducing TPNF, and thus, for TTPs and streaming TreeJoins
    to be picked up, factorization needs to be enabled!
  o Big advantage: Tree Patterns no longer depend on the
    -inline-variables option to be introduced and the interference
    with join introduction is gone
  o The old rewrite rules are kept (for now) in
     Optimization_rules_treepattern_old
  o TODO: need to make TPNF normalization more robust, especially
    with respect to the other XPath axes.
  o TODO: More testing

October 21, 2006 - Nicola

  o Logical optimization: modified introduction of Left Outer Join
    according to the latest specifications of the algebra

October 20, 2006 - Philippe
  Added FLWOR decomposition to TPNF for robustness

October 20, 2006 - Philippe
  The reorganization of the optimizer is not expected to cause any
  trouble. All the XQTS tests pass as before. Let me know if you
  run in trouble anyway.
  o Created new module Optimization_walker and factored out AST walker
    functionality from Optimization_rewqrite to it
  o Factored out the Tree Pattern, Treejoin and Cleanup rewrites into 
    respectively Optimization_rules_treepattern,
    Optimization_rules_treejoin and Optimization_rules_cleanup
  o TODO: Lots! 
      - Try to factorize out most of the rules in separate modules, so
        that the optimizer becomes manageable again, i.e., we should
        add separate modules for the group-by and join rewrites
      - Either merge Optimzation_jugde and Optimization_util or push
        the functions in both to the appropriate module
  
  o Fixed Streaming bug: attribute axis failed to test for label on
    parent, giving weird results for some MemBeR benchmarks

October 17, 2006 - Philippe
  o TPNF in place! Lots of bug fixes to TPNF normalization rules
  o All rules have undergone some testing now and seem to do fine

October 16, 2006 - Philippe
  o Removed earlier additions in ast_walker, decided to go with own
    TPNF substitution algorithm, which is easier for the TPNF
    fragment
  o Aligned TPNF normalization with paper. More testing is needed
    to ensure conformance with TR.

October 16, 2006 - Jerome

  o Serialization:
     - Fixed corner case issue with serialization of floating point
       numbers.

October 16, 2006 - Philippe

  o Adjusted select compilation to pick up conjunctions that return
    CEEmpty in addition to those returning fn:false()
  o Added a rename_variable function to ast_walker_rewrite
  o Modified substitution in ast_walker_rewrite to deal with variable
    substitution for axis calls
  o Removed faulty access_onesub call for (among others) AOEProject
    operators in compile_annotate that compromised streaming plans

October 13, 2006 - Mary 

  o Put glx:deep-distinct() back.

October 10, 2006 - Nicola

  o Physical optimization: 
    - corrected pipeline analysis not to do double materialization    
    - implemented analysis for joins, but cannot be tested until 
      pattern matching in code_util_join is modified to accept updates

October 10, 2006 - Nicola
   
  o Physical optimization and updates: implemented analysis of pipelining
    for all operators on tables, with the exception of Joins.

***** Tagged with galax-0-6-10 *****

October 10, 2006 - Mary

  o Release:
    - Set config/Makefile.galax RELEASE number to 0.6.10.

October 5, 2006 - Nicola
   
  o Logical Optimizations: 
    - corrected/aligned the push mapconcat through group rule
      to push the MapConcat as regular MapConcat, and not as
      OMapConcat
    - added a rule for converting MapConcat {OMap ..} into OMapConcat
    - updated compilation to detect AOEParse when the argument of 
      fn:doc is fs:promote-to-anystring

October 5, 2006 - Mary [DXQ]
  o DXQ regressions started in examples/dxq/test/
    I guess they should go under the regress directory

  o Changed "exec {}{}" to take hostname:port pair, so that galax-run
    can call a server directly.

  o Fixed several bugs in unboxing of DXQ results. 

  o Hacked namespaces to include prefix-uri pairs of imported
    modules. 

    Time to fix this and implement imported modules correctly.

October 5, 2006 - Nicola
   
  o optimization and updates: enabled detection of left outer joins
    for queries with side effects
      
October 4, 2006 - Mary
  o planio/ cleanup

    Re-organized directory to align with other directories.
    Contains:
      Planio_common: shared QNames, symbols
      Planio_util: utility functions
      Planio_physical_value: Boxing & unboxing of physical values
      Planio_top: Boxing & unboxing of closures/modules/algebraic plans

    TODO: Cleanup serialization of tables

October 3, 2006 - Mary

  o Extricated threads from all modules except Galaxd.

    Galax_server is now a functor, which takes the "kind" of server
    module (Thread or Forked) as an argument.

    Extended processing context to include the server function to
    invoke a remote query, because it depends on the kind of
    Galax_server.  The "toplevel" program (galaxd) determines what
    kind of server to use. 

    The server's query function must be available to the Execute
    operator, so I extended the Processing_context to include it.  
  
October 2, 2006 - Jerome

A few more Galax bugs fixed, which were lurking in the tests listed in
knownbugs.xml. This does not change the results for XQTS 1.0.

  o Testing:
     - Cleaned up and updated the known-bugs files, reported
       corresponding bugs to Galax or XQTS bugzilla as appropriate.

  o Galax Compilation:
     - Directing ocaml compilers to not raise overridden methods
       warnings.
     - Cleaned up unused variables throughout the code.

  o Type promotion:
     - Finished implementing support for type promotion of xs:anyURI
       to xs:string.

  o F&O:
     - Properly implementing fn:resolve-uri, notably checking for
       absolute URI when necessary.
     - Fixed fn:local-name-from-QName to return a value of type
       xs:NCName instead of xs:string.

September 29, 2006 - Jerome

  Results as submitted to W3C:

  Feature Galax
  Minimal Conformance	           14410 / 195 / 14618 
  Optional Features	
    Schema Import Feature	   0   	 / 0   / 174 
    Schema Validation Feature	   0   	 / 0   / 25 
    Static Typing Feature	   46  	 / 0   / 46 
    Full Axis Feature	 	   130 	 / 0   / 130 
    Module Feature	 	   0   	 / 0   / 32 
    Trivial XML Embedding Feature  0   	 / 0   / 4 

  o Normalization:
     - Finally added proper support to check variable scoping. About
       time... This is used notably to check for the scope of
       variables inside the prolog.

  o User-defined functions:
     - Re-enabled type matching on user-defined functions, mistakingly
       disabled.
     - Checking that functions are not in no namespace.

September 29, 2006 - Mary

  o Added fn:doc* test groups 
    - Changed Test_core to load URIs so we can run fn:doc* test groups 
  o Added fn:id* test groups 
    - Partially implemented fn:id and fn:idref
      This still needs LOTS of work but I put it in anyway. 
 
  [ Other notes: Static typing (weak or strong) of the validate {}
    expression is broken. 

    ]

  Minimal Conformance	 14317 / 194 / 14618 
  Optional Features	
    Schema Import Feature	  0 / 0 / 174 
    Schema Validation Feature	  0 / 0 / 25 
    Static Typing Feature	 46 / 0 / 46 
    Full Axis Feature	        130 / 0 / 130 
    Module Feature	          0 / 0 / 32 
    Trivial XML Embedding Feature 0 / 0 / 4 

September 28, 2006 - Mary

  Minimal Conformance	          14289 / 153 / 14618 
  Optional Features	
    Schema Import Feature	  0   / 0 / 174 
    Schema Validation Feature	  0   / 0 / 25 
    Static Typing Feature	  46  / 0 / 46 
    Full Axis Feature		  130 / 0 / 130 
    Module Feature		  0   / 0 / 32 
    Trivial XML Embedding Feature 0   / 0 / 4 

  o Fixed fn:index-of and fn:deep-equal to correctly use the op:eq
    semantics.  fn:distinct-values still uses its own variant, which is
    troublesome.

  o Temporary fix to handle subtyping of derived types in
    Subtyping_build_regexp (Line 75), but it's time to bite the bullet
    and have compiled, derived types.

September 28, 2006 - Nicola
  
  o Toplevel: added a flag -infer-independence for testing independence of
    operators in the presence of updates
  o Path analysis:
    - added a modified field to the tuples of paths returned by the path analysis
    - extended the physical annotation to contain not only the concatenated
      paths of the sources, but also the returned, accessed and modified paths
      for each operand
  o Rewriting
    - when -infer-independence is set, doing inlining of let clauses if the 
      return clause is a call to fn:doc
  o Optimizer:
    - updated the push-mapconcat-through-groupby rule to work when operands
      have updates, using path analysis in the optimization_judge

September 27, 2006 - Jerome

  Minimal Conformance	           14210 / 169 / 14618 
  Optional Features	
    Schema Import Feature	   0     / 0   / 174 
    Schema Validation Feature	   0     / 0   / 25 
    Static Typing Feature	   46    / 0   / 46 
    Full Axis Feature	 	   130   / 0   / 130 
    Module Feature	 	   0     / 0   / 32 
    Trivial XML Embedding Feature  0     / 0   / 4 

  o F&O:
     - Fix to the fn:abs fix!
     - Fixed fn:ceiling, fn:floor, fn:round, fn:rount-half-to-even to
       support optional parameters.

September 27, 2006 - Mary

   o More date,time,duration fixes
   o Static context checks 
   o Handling whitespace in QNames, hexBinary, etc.
   o testconfig-tmpl.xml: 
     Minimal Conformance	 14182 / 197 / 14618 
     Optional Features	
         Schema Import Feature	 0 / 0 / 174 
	 Schema Validation Feature 0 / 0 / 25 
	 Static Typing Feature	 46 / 0 / 46 
	 Full Axis Feature	 130 / 0 / 130 
	 Module Feature		 0 / 0 / 32 
	 Trivial XML Embedding Feature	 0 / 0 / 4 

September 27, 2006 - Jerome

  o Prolog:
     - Properly checking that the encoding declaration corresponds to
       a valid encoding.

  o F&O:
     - Fixed fn:abs() to support optional parameters.

September 27, 2006 - Jerome

  o Fixed bug in typeswitch optimization rule, implemented backward
    (was removing branch when type is not subtype which is wrong as
    there might be some intersection).

September 27, 2006 - Jerome

  o Current status of regress/testconfig-tmpl.xml: 

  Minimal Conformance	          14187 / 191 / 14618 
  Optional Features	
    Schema Import Feature	  0   	/ 0   / 174 
    Schema Validation Feature	  0   	/ 0   / 25 
    Static Typing Feature	  46  	/ 0   / 46 
    Full Axis Feature	 	  130 	/ 0   / 130 
    Module Feature	 	  0   	/ 0   / 32 
    Trivial XML Embedding Feature 0   	/ 0   / 4 

 o F&O:
    - Fixed comparison operations for xs:AnyURI. Just broken...
    - Fixed default-collation() function to return a string value,
      consistently with its type.
    - Numerous fixes to the compilation of fn:sum().
        -- Added proper support for binary and unary variants.
        -- Added proper typing rules.
        -- Fixed several deep bugs in rewritings related to aggregate
           functions.
    - Fixed handling of NaN properly in fn:min, fn:max.
    - Added proper min/max support for xs:string and xs:anyURI.

 o Static typing:
    - Fixed bug in type promotion detection when the input is has the
      empty sequence type.

 o Type promotion:
    - Fixed several bugs in type promotion of xs:anyURI into
      xs:string, from typing to rewritings.
    - Fixed strange bug in type promotion for overloaded functions avg
      and abs (hoping it's ok!?). See uncommented parts in the
      function overloading tables norm_overloaded.ml

September 27, 2006 - Mary 

  o Current status of regress/testconfig-tmpl.xml: 

  Minimal Conformance	 14151 / 227 / 14618 
  Optional Features	
    Schema Import Feature	 0 / 0 / 174 
    Schema Validation Feature	 0 / 0 / 25 
    Static Typing Feature	 46 / 0 / 46 
    Full Axis Feature		 130 / 0 / 130 
    Module Feature		 0 / 0 / 32 
    Trivial XML Embedding Feature	 0 / 0 / 4 

September 26, 2006 - Mary 

  o Current status of regress/testconfig-tmpl.xml: 

  Minimal Conformance	 13878 / 149 / 14618 
  Optional Features	
    Schema Import Feature	 0 / 0 / 174 
    Schema Validation Feature	 0 / 0 / 25 
    Static Typing Feature	 46 / 0 / 46 
    Full Axis Feature		 130 / 0 / 130 
    Module Feature		 0 / 0 / 32 
    Trivial XML Embedding Feature 0 / 0 / 4 

September 26, 2006 - Mary 

 o More conformance fixes

   - Overloaded functions.  Default collations. URI comparison.
     Lexical space for binary types, etc., etc., etc.

 (Subtyping_build_regexp (Line 75) : type intersection for derived
  atomic types needs to be fixed.  See test K-SeqExprTreat-5
  Pass 13837/14027)

September 23, 2006 - Nicola 

 o Algebra: removed the unnecessary AOEMaterialize operator

 o Optimizer: corrected the is_singleton judgement for selections

 o Code selection: 
   - added a "materialization" option at the top level
   - this option, if set to "always", forces materialization  
     of tables output by the tuple operators

September 21, 2006 - Mary 
 o Makefile fixes

**********************************
*** You must re-run configure. ***
**********************************

September 21, 2006 - Mary 

 o Removed "-with-galaxd" option from configure script.  Switch no
   longer necessary.  Makefile now compiles & links all top-level
   targets appropriately: galaxd links with galax-threads.cm*a and all
   other targets with galax.cm*a

   Removed toplevel/top_server.ml from CVS b/c it's generated.

September 21, 2006 - Philippe

 o Quick compile error fix (unclosed comments in code_fn)

September 20, 2006 - Mary

 o More conformance fixes
   - Boundary & error conditions for durations, timezones, etc.
    13698/13876 (total 14618)

September 20, 2006 - Mary

 o Conformance fixes
   - Fixed lexical space of hexBinary (pairs of hex digits)
   - Fixed semantics of Castable on ()
   - Made unknown character encoding a warning (per XQuery spec) 
   - Added "declare option" to prolog.  Parses but is semantic NOOP.

September 19, 2006 - Mary

 o Re-enabled typeswitch rewrite rule, once again

 o Datatypes et al

   - Numerous changes to pass Cast/Castable conformance tests
   (Total: 13486/13876)

September 19, 2006 - Philippe

 o Completed the TPNF normalization phase in factorization
   Use -factorization on to test it
   TODO: add an extra rule to combine doc() calls for 
         conjunctive path expressions.

September 18, 2006 - Jerome

 o Type assertions:
    - Fixed some pretty-printing issues.
    - Fixed variable inlining in the optimizer to check for type
      assertions.

 o Optimization:
    - Fixed bug in join detection, not picking up when using string
      equality.

 o Testing:
    - Fixed bug in propagation of strong-typing setup in the
      test-harness.

September 15, 2006 - Jerome


 o Optimization:
    - Added some checks so that constructors (element,attribute, etc.)
      are considered as side effect operators. This is to avoid
      rewriting MapConcats into cartesian products in the case the
      dependant branch constructs something. Turning those into a
      product means it is evaluated only once, hence may not construct
      as many nodes as expected. Consider for instance, the following
      query:

      (for $v1 in (1, 2, 3, 4)
       let $v2 := <b/>
       return ($v2))/.

      eventhough $v2 does not depend on $v1, we must keep this as a
      MapConcat otherwise there will be only one b element and the
      path expressions '/.' will collapse them into one node, instead
      of the 4 expected nodes!!

      Big issue here: is that going to be too restrictive. Is that a
      limited form of side effect on which better can be done?!

  o F&O:
    - Fixed promotion of duration types in for fn:distinct-values().

September 15, 2006 - Philippe

 o Normalization: XPath annotations are now replaced by scrambling 
   annotations. 
 o Overhaul of factorization phase for TPNF

September 14, 2006 - Nicola

 o  Optimization & side-effects: corrected the management of the 
    "updating" flag for builtin functions. Factorize_side_effect
    probably became obsolete.

September 14, 2006 - Nicola

 o Optimization -- aligned (partially) the following rewriting rules 
   to the xbangalgebra draft:
   - natural product rewrite (except the independence considerations)
   - join rewrite
   - mild group creation

September 13, 2006 - Jerome

 o Website:
    - Preliminary XQuery! demo available.

 o Testing:
    - Fixed test harness to deal with static typing tests properly.
    - Changed the test harness printing to indicate group names.
    - Removed the InternalVariablesWithout test group which includes a
      recursive funtion definition which makes the test harness go
      into an loop.
    - Added support for external context items in the test harness.
    - Resynchronized the expected tests file with the current results
      for v1.0.

**********************************
*** You must re-run configure. ***
**********************************

September 13, 2006 - Mary

 o Added "-with-galaxd" option to configure script (default without)

   -with-galaxd (re)compiles all modules with threads enabled,
    link with threads library, and creates galaxd.

   !!NB!!: All top-level executables will be thread-enabled and
   therefore *RUN SLOW*.  Only use -with-galaxd if you really need it.

   
 o Added toplevel/{top,thread,unix}_server.  Makefile selects
   appropriate server based on config options. 

September 11, 2006 - Mary

 o Added processing-model options to galax-test so we can run
   StaticTyping test group (still need to change Makefile by hand)

 o Aligned config/Makefile.galax-inst with config/Makefile.galax

September 6, 2006 - Mary & Chris R

 o Added -lthreadsnat -lpthread to C-API link options to support
   threaded Galax. 

Sepember 5, 2006 - Nicola
 o Fixed the looping rewriting introduced in the combine selections rule

August 30, 2006 - Nicola

 o Corrected two bugs related to side effects

    - functions with snapped effects should still be considered
      updating. 
    - fixed the updating flag on the compilation of while loops.

August 18, 2006 - Trevor

 o Added the web interface for galaxd.

August 18, 2006 - Nicola

 o Added a modified dxq example with a DNS resolver having 
   a simple cache.

August 18, 2006 - Trevor

 o Galaxd cleanup.

August 18, 2006 - Trevor (Distributed XQuery)

 o Bug fix for threaded galax server.  (Needed to
   explicitly flush output on Thread.exit().)

August 18, 2006 - Nicola

  o Added a glx:gettime() function that returns the current time, 
    in seconds. It differs from current-dateTime(), which returns
    the time when the query context was built.
 
August 18, 2006 - Trevor (Distributed XQuery)

 o The galax server is now threaded.

**********************************
*** You must re-run configure. ***
**********************************

August 17, 2006 - Jerome

  o Updates:
    - Fixed pretty printer for updates.

August 17, 2006 - Jerome

  o Testing:
    - Added cleaner diff for expected vs. actual test results using
      XQuery.

***** Tagged with galax-0-6-8 *****

August 16, 2006 - Jerome

  o Release:
    - Set the version number to 0.6.8.
    - Added 0.6.8 entries in the documentation and Web sites.
    - Some minor fixes the documentation, notably updating the
      contributor's list.
    - Added some text for support for ULTF, XQueryP in the release
      notes.
    - Fixed the ns_usecase expected result to match the one generated
      by Galax. Namespace handling in constructors may be fixed at
      some point and result in subsequent changes.
    - Fixed usecase Makefile to not set monitoring on.
  o ULTF/XQuery!/XQueryP Trilogy:
    - Added -language ultf option for W3C Update Facility.
    - Added minimalistic tutorial in the documentation for how to run
      those.
    - Added examples in ./examples/extensions for each of the 3
      languages.
  o F&O:
    - Fixed fn:matches function to not take optional second and third
      arguments anymore.

August 16, 2006 - Nicola

  o Added -language option to galaxd.

August 16, 2006 - Mary

  o Configure script
    - Sets $ocaml_home if not set by default

  o Code_selection_context: 
    - Added enter/exit_closure_context

  o Cs_code_selection_top
    Galapi/galax
    Procmod_compiler
    Procmod_phases
    Procmod_types

    Cs_code_selection_top.code_selection_statement now correctly
    returns updated code_selection_context.  Previously, it was
    dumped. 

    Corresponding changes in modules listed above to propagate and
    return updated code_selection_context.
   
  o Cs_code_typing_top:
    type annotation of top-level expression stored as a global
    annotation so available to caller.

  o Cs_util_coercion:  
    Added missing coercion functions 

  o Code_binding:
    Tuple fields contain physical_xml_values, not physical_values:
    minor changes to support this change
    
  o Code_execute:
    Added support for shipping plans that return tables/tuple-streams

  o Code_util_materialize:
    Changed names of functions to reflect type of materialization,
    e.g., materialize_cursor_to_dom_value_array
   
    Many files updated to reflect name change:
    code_group_order.ml
    code_hash_join.ml
    code_join.ml
    code_nestedloop.ml
    code_sort_join.ml
    code_tuple.ml

  o Parse_xquery.mly:
    Fixed long-standing bug: EOF missing from end of library and main
    modules. 

  o Website/demo/querycgi.ml : alignment

August 16, 2006 - Mary (DXQ)

  o DNS examples: 
    Changed name of module interface to dns.xqi

  o Optimization_dxq:
    Working on DXQ opts

  o Planio/*:
    Added support for serializing and parsing.
    ***All these files need to be reorganized.  They are a mess***
    
  o Galax_server: fixed error-return XML

  o Galaxd: yet another debug log bug
 
August 15, 2006 - Jerome

  o Testing:
    - Added knownbugs.xml in the dependencies for testresults.xml
    - Added query file information in test results for wrongly
      succeeding tests.
  o Type matching:
    - Added missing checks for undeclared types during type matching.
  o Lexing:
    - Further fix to the lexer dealing with numeric literals followed
      by a qname without separating whitespace.

August 15, 2006 - Nicola

  o XML Updates: 
    - Changed the code generation for Insert and Replace to 
      expand document nodes into the list of their children,
      as specified by the W3C draft.
    - added two examples to test these expansions, in a new
      examples/xquerybang/tests folder

August 14, 2006 - Jerome

  o DXQ:
    - Fixed wrong pattern matching in error printing, not catching DXQ
      errors.
  o Binary types:
    - Properly implemented xs:hexBinary and xs:base64Binary types, now
      get converted to the correct internal binary value. Fixed
      corresponding comparison operators and cast into each others.

August 14, 2006 - Nicola

  o XML Updates: 
     Added a Makefile and .expect files for performing
     automatic regression tests for the 
     XQuery Update Facility Use Cases.
     The addressbook usecase is skipped, because of an issue
     related to update conflicts in the last W3C draft.
     The current output of Galax on these programs needs to 
     be checked again.

August 11, 2006 - Jerome

  o Date & Time:
    - Fixed bug in comparison of durations, not comparing the sign
      properly.
    - Added missing support for promotion of xs:yearMonthDuration and
      xs:dayTimeDuration into xs:duration.

August 11, 2006 - Kristi (Distributed XQuery)

 o Modified galax-parse.ml, top_options.ml, top_config.ml,
   top_config.mli to support -diff option (compares 2 XML files), so
   can use galax-parse -diff to diff 2 XML files

 o Added preliminary test cases for DXQ in examples/dxq/test/q along
   with 2 programs, run-tests and generate-tests run-tests is a shell
   script that calls galax-parse -diff to check if tests passed.
   generate-tests is a perl script that generates null-ary test cases

August 10, 2006 - Jerome

  o Parsing:
    - Fixed parser to properly impose whitespace around arithmetics
      operators (10div 3, etc.).
  o Date & Time:
    - Added support for addition of xs:yearMonthDuration or
      xs:dayTimeDurations as first operands, missing.
    - Fixed multiplication of double with yearMonthDuration when the
      numeric value is infinite, or nan (overflow and parameter
      mismatch respectively).

August 10, 2006 - Jerome

  o Normalization:
    - Fixed normalization of Expr[Numeric] to use subsequence only for
      integer literals, since it is wrong for other kinds of numeric
      literals.
    - Fixed predicate boolean value to not do rounding for numeric
      values.

August 9, 2006 - Jerome

  o Parser:
    - Fixed precedence for XQueryP expression.
    - Fixed {1} like expressions to be errors except when XQueryP is
      on.

  o Variable handling:
    - Fixed variable lookup in static typing to use expanded-Names
      lookups rather than triple (prexix,uri,localname) lookups.
    - Fixed name comparison in variable inlining in the optimizer.

August 8, 2006 - Trevor (Distributed XQuery)

 o One more XML printing fix for the DXQ GUI.

August 8, 2006 - Trevor (Distributed XQuery)

 o Better printing of XML for the DXQ GUI.

August 2, 2006 - Nicola

 o XQueryP: turned static typing off (by default) and
            corrected the "binding.xq" example

August 2, 2006 - Nicola

 o XQueryP:
   added ImperativeSeq in the core and algebra. It is similar
   to Seq, but it only returns the result of the last expression
   in the sequence. Syntactically, an ImperativeSeq constitutes
   the content of an XQueryP block and the delimiter between
   expressions is the semicolon.

August 1, 2006 - Nicola

 o Optimizer: fixed the variable inlining rule that did
   not respect scoping. Also, corrected annotation for
   AOESet to Many (because it redefines the variable).
 o XQueryP:
   - implemented parser and normalizer for blocks
     !! For the time being, blocks are similar 
        to sequences, concatenating their results.
        This should be changed to returning the 
        result of the last expression.
   - added "-language xqueryp" option to Galax
   - added a couple of first examples, inspired from
     the XIME-P presentation; some things are broken due
     to static typing

August 2, 2006 - Mary

 o Parse_xquery.mly: 

   Fixed long-standing bug: EOF-token was omitted from xquerymodule
   and librarymodule productions, permitting parser to ignore first
   error in a module file.

August 1, 2006 - Trevor (Distributed XQuery)

 o New primitive functions for domain name comparison, used in
   DNS example.

July 31, 2006 - Mary

  o Error : added DXQ_Error

  o Variable_manager_context, et al

    - Changed variable manager to store Physical_value.xml_value
     (Trees only) instead of Physical_value.physical_value (Tables or
     Trees), because variables never contain Tables. Avoid lots of
     unnecessary boxing & unboxing.

    - All functions for storing/accessing variables changed as well.

  o Cs_util : added function to demangle names in dxq: namespace. 

  o Code_execute: numerous bug fixes

  o Galax.compile_serialized_closure : working on adding tuples

  o Streaming_constructors : added try-with around constructors to
    propagate better error messages.

  o Galax_server: added <alg:Result> box around <glx:error/> for
    consistency. 

  o QUESTION: Why do we return item lists in Physical_load instead of
    item cursors?

July 31, 2006 - Nicola

  o Fixed treatment of letvars in the normalization of while loops
  o Fixed variable assignment to return the empty sequence.

July 31, 2006 - Nicola

  o Fixed glx:getdoc() to avoid any form of document caching.

July 31, 2006 - Nicola

  o Imperative statements:
    Corrected normalization of WHILE loops; some free variables in
    the CEInsert were missed

July 31, 2006 - Nicola

  o Imperative statements:
    finished implementing variable assignment; but code selection
    for it is probably unefficient at this moment

July 28, 2006 - Jerome

  o Alignment:
    - Correctly raises errors with mal-formed entity references in
      string literals.

July 27, 2006 - Nicola

  o Added the glx:getdoc() function that behaves similar to
    fn:doc() except that it always re-reads the document from
    the source. If the document changed in the meantime, 
    it will retrieve the new data. 

July 27, 2006 - Mary

  o Optimization_rewrite: 
    Fixed bug in re-annotation: previously reannotated "root"
    expression, but if optimization rewrites "root" as a new
    expression, the old root, not the new expression, was incorrectly
    reannotated.

  o Namespace_context.make_binding:  
    Did not handle (NSDefaultElementNamespace, NSUri "") correctly.

  o Code_selection.retrieve_annotation: 
    Added provenance of call to make debugging less painful.

  o Compilation/code-selection of user-defined, external functions
    that are not built-in.  

    Right now, the treatment of F&O built-in functions is conflated
    with user-defined external functions.  We know a function is
    user-defined & external if it is _not_ in the compile_context. 
    Must handle this case in optimization and code selection:

      Code_user_defined_fn 
      Optimization_judge
      Optimization_rewrite

    TODO: We might want to "de-conflate" these two. 

  o Galax.compile_serialized_closure:
    Returns prepared_prlog, statement pair.  Prepared_prolog contains
    all info for closures' environment vars for execution of statement.

July 27, 2006 - Nicola
  o Imperative statements: 
    - added Set as a new statement at XQuery and Core level, without
      runtime code yet
    - filled missing matches for letvar and set in the normalizer&
      compiler

July 27, 2006 - Nicola
  o Imperative statements: completed normalization of letvar's
    into AOELet 

July 27, 2006 - Mary 

  o Namespace_context:
    - Fixed bug in which duplicate namespace entries were not removed
      from delta bindings.   Resulted in serialized element with
      duplicate xmlns:<prefix> attributes, which is not well-formed.

  TODO: 
  o Serialization

    Streaming_ops.prefix_xml_stream & Namespace_context.cleanup_bindings:
    The computation of required namespace attributes in
    prefix_xml_stream seems pretty inefficient.  We end up comparing
    strings of (prefix,uri) pairs --- why not compare (int) symbols?

July 26, 2006 - Jerome

  o XML Updates:
    - Fixed precedence problems in the new XML update grammar.

July 26, 2006 - Nicola

  o Imperative statements: 
    - corrected normalization of while loops
    - added syntax for letvar's

July 26, 2006 - Mary & Kristi (DXQ)

  o Closures
    - Added boxing/unboxing for stand-alone attribute values

July 26, 2006 - Mary 

  o Debugging messages
    - Slowly changing all calls to *print* functions to use Debug
      module instead.

  o Code_selection_context 
    Add enter_closure/exit_closure to set up variable and tuple
    contexts. 

  TODO:
  o Debug module - We are using List.exists to check debug variables,
    which is really inefficient; should use bit vector.
 
July 26, 2006 - Philippe

   o Added two normalization rules + fixed bugs in other rules.
   o Changed factorization to do TPNF normalization only for now.

July 25, 2006 - Nicola

   o Implemented glx:sleep() that pauses execution for a given time

July 25, 2006 - Nicola

  o XML Updates: corrected normalization of ETransform (it was not doing
    the `copy' after my last changes)
  o Imperative statements: added parser support for while loops and
    a way to normalize them into XQuery! core
  o Norm_ident_expr now raises exceptions for the new ETransform and
    EWhile expressions

July 25, 2006 - Jerome

  o Patched SBDO stuff (makefile and call to
    Optimization_sbdo_automaton instead of Sbdo_automaton) so that
    Galax compiles again. SORRY PHILIPPE, you'll have to look this
    over again.

July 25, 2006 - Mary (DXQ)

  o Code_execute, Galax, Planio modules
    - Completed (type) boxing and unboxing of datamodel values passed in
      closure environments and returned as results from Galaxd server. 

  o Physical_load
    - Fixed very old bug in load_xml_value_from_typed_stream that
      delegated to load_document instead of load_xml_value.

July 25, 2006 - Philippe

  o Fixed compile error

July 25, 2006 - Philippe

  o Added factorization plugin module for rewriting CXQ expressions
    into TPNF normal form. Files:
    - factorization/factorize_tpnf.ml(i)
    - factorization/factorize_tpnf_rules.ml(i)
    The rules are of course only triggered when factorization is
    turned on.

July 24, 2006 - Mary & Kristi

 o Galaxd.ml updated to include prelim support for type information in
   a boxed request on the server side.
 o planio_print_xml.ml contains some udpates

July 24, 2006 - Mary

 o Debug module now prints to Conf.glx_err_formatter.
 o More DXQ changes 

July 24, 2006 - Nicola
  o XML Updates code generation:
    Implemented conflict detection for "replaces" and "renames",
    according to the W3C's July 2006 specs

July 21, 2006 - Mary & Kristi

 o Code_execute: changed to construct and emit a closure 
 o Added Galax.compile_serialized_closure
   - Unboxes an environment, extends code-selection, execution, &
     physical-type contexts with free variables and their values
   - Invokes body of closure in new context
 o TODO: Change code_execute to parse XML stream result from server
   directly instead of redirectly to a file. Yuck.
 o Changed Debug.print_*_debug() to check flags internally so that
   they do not have to be guarded by if (*_debug()) then print_*_debug()

July 21, 2006 - Mary

 o Unstable state for DXQ ONLY

July 21, 2006 - Nicola

  o Normalization & updates:
    corrected normalization of Replace Value Of; the construction
    of the text node out of the "with" clause was wrong
  o XML Updates code:
    removed the check for insert-insert conflicts

July 21, 2006 - Nicola

 o XML Updates:
   - Added support for the "transform copy ... modify" statement
     (which is normalized into other XQuery! constructs)
   - Finished implementing "replace" statements
   - disabled "snap nondeterministic"; conflict detection should
     in general be updated soon

 o XML Update Use Cases:
   - Added all the W3C usecases, with some fixes. A Makefile
     will be needed to run them automatically. 

July 20, 2006 - Jerome

 o Streaming:
   - Added test support for streaming of ','.

July 20, 2006 - Mary

 o Unstable state for DXQ ONLY

July 19, 2006 - Nicola

 o XML updates:
   - Aligned the parser with the latest W3C draft, except for
     "transform modify" which is still missing
   - Added an "updating" modifier keyword in the function declarations
     (also specified in the W3C draft) that is required in the declaration
     of functions that can return lists of pending updates
   - Propagated information provided by the "updating" flag to function
     declarations and function calls at all levels: AST, core, algebra
   - Removed the "Detach" update (at all levels), as it subsumed by "Delete"
   - Implemented the "Rename" update in the galax datamodel. Still
     working on "Replace" and "Replace_value"
   - Added a new directory usecases/xqupdate for the XML Update Use Cases
     (from W3C). For now it only has the R and the Parts usecases. The 
     rest will follow soon (but many of the queries are not correctly
     specified by W3C).

 o Configuration: set the -with-c flag for the configuration script to
   be false by default, as ./configure -help says it should be. Attn:
   there is a problem with the $RANLIB variable here, so you should
   probably still compile the C API.

July 19, 2006 - Jerome

 o Overloaded function calls:
    - Factorized type access during the generation of type promotion
      code. This significantly reduce the number of operations for
      overloaded functions (pretty pervasive).

July 19, 2006 - Mary & Kristi (DXQ)

 o Skeleton for closures in Code_execute
 o Debug module: added two new options for "typing" and "dxq"

July 18, 2006 - Jerome

 o FSA:
    - Removed duplicates in visited states list during intersection
      leading to blowup in size and huge performance issues during
      subtyping. (Issue Circa 2000).

July 18, 2006 - Kristi and Mary (Distributed XQuery)

 o Added stubs and initial code for supporting POExecuteTuple in 
   code_selection/code/code_execute.ml and 
   code_selection/code/cs_code_typing_top.ml

July 17, 2006 - Trevor (Distributed XQuery)

 o Bug fix for determining short server name in galax_server.ml
 o Added -d option to galaxd.ml to make it easier to start a cluster
   of galax servers.

July 17, 2006 - Trevor (Distributed XQuery)

 o Added dxq: name demangling.
 o Changed the example to use the dxq: name convention and
   added an interface file.  Cleanup.

July 14, 2006 - Kristi & Mary

 o Optimization_top
   - Fixed long-standing bug in which function bodies were not optimized.
     TODO : Optimize global variables definitions

 o Physical algebra
   - Now have two variants of AOEExecute: POEExecuteTree and POEExecuteTuple,
     to permit remote plans to return either tree or tuple values.

 o Galaxd
   - Added options for function & variable inlining

July 14, 2006 - Trevor (Distributed XQuery)

 o Change parser and lexer to support a new syntactic form:

        exec { "foo" } { bar() }

   executes the code bar() on host foo.

July 13, 2006 - Mary

 o Xquery_algebra_ast: 

   - Changed function body into a structure with two fields containing
     logical plan (before code selection) and physical plan (after
     code selection)
   
   - Permits in-lining & optimization of function calls in top-level
     user queries by inlining the _logical_plan_.  Previously, it
     attempted to inline physical plan, which is just wrong.

 o Updated usecases/xmark_usecase.expect to align w/ input file

July 8, 2006 - Trevor (Distributed XQuery)

 o Correct locking and reporting of debug log in galaxd.

July 6, 2006 - Mary & Kristi (Distributed XQuery)

 o Changed galaxd so that prolog is both compiled and evaluated on
   server start-up.  Led to numerous questions on what should happen
   on server start-up (see "Things-to-fix" on Wiki).

July 6, 2006 - Trevor (Distributed XQuery)

 o Eliminated useless proc_ctxt replication in galaxd
 o Correctly set dxq host and port in proc_ctxt for non-simloop case
 o Allow simulations to be run from a directory different than the directory
   containing the source code

July 6, 2006 - Mary  and Kristi (Distributed XQuery)

 o Modified ast_printer to add support for printing plans of an
   Execute{} operator after optimization.
 o Modified base/galax_http.ml to disambiguate the Http module

July 6, 2006 - Mary 

 o Fixed name conflict in base/galax_http.ml

July 5, 2006 - Trevor (Distributed XQuery)

 o debug log is now flushed properly.

July 5, 2006 - Mary and Kristi (Distributed XQuery)

 o Changed Code_execute/build_default_ceexecute_code() to detect when
   a Galaxd server issues a remote plan to itself: Now circumvents
   server and evaluates plan locally. 

 o Processing_context: added procctxt fields to identify Galaxd
   host/port. 

 o Questions:

   - In processing_context, why are monitor_context, external_nsenv
     not copied?

   - In Print_xquery_algebra, line 730, handle two cases for
     AOEExecute. See comments there. 

 June 29, 2006 - Kristi
 o Flattened the function calls in the following files to reduce recursion:
 o examples/dxq/directory-doc-remote/a.xq
 o examples/dxq/directory-doc-remote/b.xq
 o examples/dxq/directory-doc-remote/c.xq
 o examples/dxq/directory-doc-remote/d.xq
 o examples/dxq/directory-doc-remote/e.xq
June 28, 2006 - Trevor

 o -print-plan-kind added to galaxd, needed to avoid a problem with
   the standard plan printing

June 28, 2006 - Trevor

 o Debug log for prolog now prints to a file in galaxd.
 o Galaxd no longer uses a temp file for server startup.

June 26, 2006 - Mary

 o Added extra debug options to galaxd : "-print-prolog on" to print
   library module when it is loaded. 

June 21, 2006 - Kristi

 o Integrated and added support for the pre-generated logfiles (see
   note from June 20) into each .xq file:
 o examples/dxq/directory-doc-remote/a.xq
 o examples/dxq/directory-doc-remote/b.xq
 o examples/dxq/directory-doc-remote/c.xq
 o examples/dxq/directory-doc-remote/d.xq
 o examples/dxq/directory-doc-remote/e.xq
 o examples/dxq/directory-doc-remote/summary.xq

June 21, 2006 - Trevor

 o Fixed bug with -print-plan-shebang output
 o Added a debugging log to galaxd
 o Use a buffer for evaluation output rather than a temp file

June 20, 2006 - Kristi

 o toplevel/galax_server.ml: added support for logging query/response
   sizes
 o toplevel/galaxd.ml: added support for logging query/response sizes
   (fn:log query response host)

June 16, 2006 - Trevor

 o toplevel/top_options.mli: export dxq options from module
 o toplevel/galaxd.ml: added various options

June 16, 2006 - Mary

 o fn:collection()

   Added all necessary hooks for fn:collection(), which currently
   loads a document, not a document fragment.

   base/galax_io.ml: See Galax_io.pull_parser_from_input_spec for info
   on how/where document fragments should be added.

June 16, 2006 - Chris Rath
 o Fixed configure script to work with Solaris /bin/sh
 o Updated RELEASE number in Makefile.galax to 0.6.6

June 16, 2006 - Mary, Trevor, Kristi (Distributed XQuery)

 o Processing_context
   Added dxq_optimization flag

 o Top_options
   Added -dxq [on/off] option to enable DXQ optimizations.

June 16, 2006 - Chris Rath 
 o Removed == test in configure as it does not work with older
   versions of /bin/sh

***** Tagged with galax-0-6-6 *****

June 16, 2006 - Mary (Distributed XQuery)

 o Optimization rules

   Added new Optimization_dxq module for Distributed XQuery.
   Uncomment rules in Optimization_rewrite to enable. 

 o Usecases

   Changed expected outputs to align with "ugly" (non-pretty-printed)
   serialized XML output. 

 o Working on fn:collection() -- see comment in Galax_io on how to
   create a PXP entity manager for document fragments.  Also
   Pxp_ev_parser. 
  
June 15, 2006 - Jerome

 o Updates:
    - Fixed bug in physical typing of insert/replace/rename.

    - Added top-level flag to select the language support (between
      XQuery 1.0 and XQuery!). Default is XQuery 1.0. Please use the
      following option to turn XQuery! support on:

       -language xquerybang

      Notes:
        * This only currently work with optimization off!! Something
          got broken in the safety checks of the optimizer with all
          the work we are going on it...
        * This also turning static typing off since it is not
          implemented (and wouldn't be safe!).

June 13, 2006 - Chris Rath

 o Removed newer shell constructs in configure so it will work
   with machines that use an older version of /bin/sh.

June 12, 2006 - Michael

 o Streaming_conv
   Fixed bug in streamed fn:count

June 6, 2006 - Mary

 o Type simplification

   In Schema_simplification.simplify_union:

   Fixed long-standing bug in which only duplicates of atomic types
   were removed from union types, resulting in "simplified" union
   types that contained multiple occurrences of node item types, e.g.,
   element(), comment(), etc.  Now correctly removes duplicates of any
   (item) types that can be compared syntactically, that is, does not
   required full-blown type inclusion.

   If we had _compiled_type_ identifiers, we could do this more
   efficiently by sorting and removing duplicates.

   See comments in Schema_simplification.simplify_union.

June 6, 2006 - Jerome

 o Serialization:
    - Turned pretty-printing off. In preparation for an overhaul of
      serialization for purposes of improving performances.

 o Optimizer:
    - Fixed bug in handling of equality predicates that do not contain
      fs:untyped-to-any() calls. Changed the variable allocation for
      the existential quantification to a safe one.
    - Improved the detection of join predicates when
      fs:untyped-to-any() calls are removed, covering asymetric
      cases. (Impacts XMark Q11 and Q12 which for which sort joins are
      now picked up).

June 1, 2006 - Philippe

 o Optimizer
    - fixed the remove_materialize_flwor rule to use has_non_trivial_snap 
      instead of has_trivial_snap, which is the negation of the latter.
      The fact that has_trivial_snap is implemented as 
      not(has_non_trivial_snap) seems weird.
    - fixed the twig merge rule to be robust wrt. project operations that
      may/may not occur in the subexpression of the select.
    - Fixed the map_from_to rewrite. It was checking wether order matters,
      but that information is irrelevant for this rule.
    - Fixed the map_concat_pushdown, to not push MapConcats past projects
      blindly. There is a special rule for this.

    Now all of my tests regarding TTP rewrites work again.
    ADVICE: wehen running tests: verify that al rewrites you expected to be
    apllied indeed are applied. The optimizer is a mess, it is almost
    impossible to guarantee this right now.

    Good news: so far no evidence was found that buggy [Cleanup 0] is
    necessary for TTP rewrites.

May 31, 2006 - Jerome

 o Optimizer:
    - Cleaned up various modules and pieces of code. Moved things
      around to the right places. Added comments.
    - Turned [Cleanup 0] rule off, since it's wrong until we know what
      it's usage is and how/whether we want to fix it.

May 30, 2006 - Jerome & Mary

***
RANT: compile annotate is evil. we need to rewrite it.
      compile_annotate.ml
      compile/compile_annotate
      WHY IS THIS IN ./compile ANYWAY? ***

 o Optimizer:

    - Turned off group flatten rule which is incorrect, creating bugs
      in conjunction with at least TTP rewrites.
    - Improved detection of join predicates in the case the predicate
      operation is internally creating and accessing new tuple fields.
    - Improved pattern matching of join predicates, picking up
      existential quantification rewritten as a let.

May 25, 2006 - Jerome

*** Improved support for debugging. Important changes:
       - Conf.debugging is gone, you should use ./base/debug instead.
       - Please always use the Debug.print_debug printing function.
       - You can now cluster debugging messages per feature.

         There are currently two features: join and default. join
         correspond to join debugging messages, default is currently for
         everything else. You can turn those features on by using:
           ./galax-run -debug join
           ./galax-run -debug default
           ./galax-run -debug default,join
           etc.

         To add a given debugging feature 'MYFEAT' that you want to
         isolate, you should change ./base/debug.ml as follows: (1)
         add an entry in the list of debug flags (e.g., MYFEATDebug),
         (2) add a function which returns true if MYFEATDebug is on
         (e.g., val MYFEAT_debug : unit -> bool), and add "MYFEAT" as
         the recognized string for MYFEAT. ***

  o Debugging:
     - Cleaned up debugging support. Removed Conf.debugging. Now
       debugging must go through the base/Debug module.
     - Added support to categorize debugging output. Currently there
       are two categories, which can be set independantly: join
       debugging, and misc debugging for everything else.

  o Joins:
     - Fixed normalization to introduce EBV properly for the where
       clause. Consequently removed hack in the compilation phase
       which was introducing EBV back.
     - Made the join predicate analysis used to detect hash/sort joins
       more robust to variations in the core expression for the
       predicate. Those are necessary since rewritings often change
       the form of the predicate because of weak typing.

May 25, 2006 - Mary
  
  o Added implementation(s) for fs:item-sequence-to-node-sequence on
    DOM tree & XML-token cursors.

    This function is only necessary for typing purposes, but if
    rewriting is off, it must be implemented as the identity function
    on both DOM trees and XML-token cursors.   

  o galax-parse command-line options:

    -dm       => All phases through loading 
    -validate => All phases through validation

  o Documentation - some updates

May 23, 2006 - Michael

  o Compile_annotate
    Small fix to tuple field use count analysis

May 19, 2006 - Philippe

  o TTP rewriting (optimization_rewrite):
    Generalized rule TP 3 

***** Tagged with galax-0-6-5 *****

May 17, 2006 - Mary

  o Browser interface/Website
    - Updated to align w/ GODI configuration & Galax 0.6.5.
    - Fixed CGI printing bug.  All output now printed correctly. 

  o Documentation
    - Updating documentation for GODI release.

  o Logical Plan Printing 
    - Fixed old bug that replicated algebra: namespace.
  
  o Loading
    - Made interface names more mneumonic

  o Java API & Examples Makefiles
    - Added path to CAPI libraries
    - Fix to install: target 

  TODO: Document command-line options :=(

May 17, 2006 - Michael
  o Conf.ml_p_
    - Fixed initial settings of dynamic buffering for streaming

May 16, 2006 - Jerome

  *** Test suite change: we now run 0.9.0.
     WARNINGS:
       - 0.8.6 will NOT run, because of xdt: to xs: namespace changes.
       - several Galax bugs are being exposed by 0.9.0 which are not
         fixed yet. ***

  o Printing:
     - Fixed issue with flushing of formatters in pretty-printing
       stubs. (Gmisc module).

  o Alignment:
     - Removed xdt: namespace, now replaced by xs: namespace.
     - Added support for 'encoding' option in the XQuery version
       declaration.
     - Implemented fn:codepoint-equal() function.
     - Fixed bug in parsing for 'castable as' expression with an
       optional type.

  o Arithmetics:
     - Fixed long-standing bug in arithmetics of dates and time not
       dealing with the empty sequence properly, both in the type
       signatures and evaluation code.
     - Short-circuits the evaluation in the case one of the operand is
       the empty sequence.

  o Constructor functions:
     - Constructor functions are now properly normalized to cast based
       on whether the function name derives from xs:anyAtomicType,
       instead of just as being in the xs: namespace.
     - Constructor functions now properly normalized to a cast
       expression with an optional type.

  o Testing:
     - Fixed test harness to handle cases where the error comes out of
       serialization rather than evaluation.

May 16, 2006 - Michael
  o Alg_path_analysis
    - Added handling of AOEProject to path analysis

May 15, 2006 - Chris Rath
  o Added a new target, "byte" to the toplevel and subordinate Makefiles.
    - Only creates the byte-code galax library and byte-code toplevel applications.
  o Added a new target "byteworld" that works like "make world"
    - Only creates the byte-code galax library and byte-code toplevel applications.

May 14, 2006 - Michael
  o Stream_conv
    - Fixed bug in typed_of_typed_labeled_xml_stream
  o Streaming_xpath
    - Added streamed treejoin for attribute-axis
  o Dynamic_buffer
    - Set initial buffer size to zero
    - Now doubling buffer size on resize

May 12, 2006 - Chris Rath
  o Fixed missing files and typos in Makefile.galax
  o Added "regression" target to top level Makefile.
  o Removed regress/testconfig.xml from CVS; this file is now generated from testconfig-tmpl.xml
  o Updated all lower-level Makefiles to ensure they all have the standard targets

May 12, 2006 - Jerome

  o QNames:
     - Fixed bug in casting of QNames introduced when improving the
       parser.

  o Parsing:
     - Fixed bug in text-node merging during parsing, introduced while
       experimenting with parser improvements.

  o Testing:
     - Fixed the knownbugs file to align with 0.8.6 and Galax status.
    
May 11, 2006 - Jerome

  ** The test-suite should work again **

  o XML Parsing:
     - Fixed bug introduced last week about resolution of attributes
       in the default namespace, affecting testing.

May 10, 2006 - Philippe

  Until now, MapTo/MapFromItem combinations got eliminated for
  TupleTreePatterns. However, the combination of MapToItem with
  MapFromItem is actually a kind of projection. Just eliminating it is
  wrong because some fields that where invisible, now become visible
  again. This was interfering with some other rewrites, like the join
  rewrites. The MapFromItem/MapToItem combo is now replaced with a
  project operator. This and some project-related rewrites allow picking
  up both tree patterns and joins correctly. XMark queries 8-12 now get
  compiled in a combination of joins and tree patterns.

  o Added AOEProject operator
  o Added some optimization rewrites related to projects (this
    get rid of the redundant products in the XMark queries)
  
  Project is basically a no-op: it only manipulates the returned
  fields of a subexpression.

May 9, 2006 - Philippe

  Fixed wrongful bulk TTP rewrite.

May 8, 2006 - Jerome

  o New compiler:
     - Started cleaning up unused variable detected by the latest Caml
       compiler.

  o Streaming:
     - Removed unnecessary file id lookups during stream parsing.
     - Fixed expensive QName parsing calls during stream parsing.

  o Toplevel:
     - Added support for intermediate parsing phases in
       galax-parse. Command-line options allow to individually call
       the following XML processing phase operations: PXP parsing,
       Galax stream parsing, Namespace resolution, Type annotation,
       Load, Export, Erasure, prefixing the stream, and serialization.

May 7, 2006 - Michael

  o Top_config
    Added code_selection_by_physical_type to streaming_shebang
  o Compile_annotate
    Minor changes to tuple field use count analysis

April 26, 2006 - Chris Rath

  o Added toplevel command sources to the .depend target.
  o Updated the configure script to support -enable-profiling option.
  o Updated Makefile.galax to support profiling.

April 24, 2006 - Philippe

  o added print-plan-shebang for
    printing all plans (core/rewritten/logical/optim/physical) w/o 
    annots, but including logical rewrites

April 24, 2006 - Philippe

  o added "shebang" options
    -streaming-shebang on/off
    -scjoin-shebang on/off
    -twigjoin-shebang on/off

April 18, 2006 - Chris Rath

  o Fixed some bugs in the configure script.

April 17, 2006 - Chris Rath

  o Changes necessary to align with GODI


***** Tagged with pre-godi *****

April 14, 2006 - Mary 

  o Changes to align Abstract DM with Galax DM (document-constructor
  now takes (atomicAnyURI option ref) argument) 

  Does anyone know how was this compiling?  They have been out ofp
  alignment for a while...  
  Who made this change?

April 12, 2006 - Mary (Distributed XQuery)
  o Minor changes to plan printing & Galax server.

April 12, 2006 - Philippe
  Getting ready for code alignment.
  Small fix related to debug output

April 6, 2006 - Mary and Doug
  o Updated distributed directory example to work correctly

April 4 2006 - Mary (Distributed XQuery)
  o Fixed galaxd to report fatal errors to stderr on startup.

  o Changed AOEExecute operator to take arbitrary expressions for host
    and port values. 

  TODO: 
  o Change galaxd/galax_server to report errors in original files; 
    Do not copy to temp files. 

April 4 2006 - Mary (Distributed XQuery)

  o Temporary fix to GUI reporting in Galax_server

March 26, 2006 - Philippe

  o Moved SBDO optimization code to optimizer
  o Applied SBDO optimization over TupleTreePatterns

March 24, 2006 - Mary (Distributed XQuery)

  o Added pseudo-function glx:remote-plan(host, port, plan), for
    executing plan at remote Galax server. Temporary place-holder for
    new extension operator "execute at Expr {Expr}".

  o Added corresponding algebra operator AOEExecute (host, port, plan)

  o Code_selection/code/code_execute: Implementation of AOEExecute

  o Galax.mli: Added compile_serialized_logical_statement to support
    shipping of query plans for individual statements. 

  o Galax_server.ml: 
    Changed "/plan" command to take plan of an individual statement.
    Should probably permit both plans for complete modules & for
    individual statements. 
     
March 24, 2006 - Philippe
  o Snap removal fix

March 24, 2006 - Philippe
  o c1 AND c2 - conditional expressions got rewritten into
      if c1 then c2 else false
    where the 'false' was an atomic value instead of an fn:false()
    call. Since compilation was relying on fn:false() to compile
    this into a normalized select-predicate description, it
    failed at doing so. The rewriting rule now uses a boolean
    call to fix this.

March 23, 2006 - Philippe
  o Serialization and parsing for TupleTreePattern operator.

March 23, 2006 - Michael
  o Added preliminary handling of TupleTreePatterns to
    tuple field use count/path analyses

March 21, 2006 - Philippe
  o Added a fall-back rewrite rule for rewriting 
    TupleTreePatterns into TreeJoins:

      TupleTreePattern [dot1](s1/.../sk{dot2}* ) (indep) 
      ==
      MapConcat { MapFromItem{v -> [dot2 : $v]}(
        TreeJoin[sk](...(TreeJoin[s1](INPUT#dot1))))
      } (indep)

    This rule could replace all other TTP elimination
    rules, but then we would end up with a lot of maps,
    which may interact with Michael's work.
    

March 15, 2006 - Mary & Doug (Distributed XQuery)
  o Changed Galaxd to take "/plan" command, which executes logical
    plan 
  o Added glx:remote-plan(host, port, plan) function to Galax_server
    to ship plan.
  o Added examples/dxq directory 

March 15, 2006 - Mary (Distributed XQuery)
  o Changed Galaxd to compile each library module once. 

March 15, 2006 - Mary (Distributed XQuery)

  o Lifted Sim,Gui, and Server modules out of galaxd.ml and put into
    toplevel/galax_server.ml, which is a misnomer, because it contains
    both server and client code.   Need to organize this. 

  o Added client-side glx:remote-query() function to galax_server.ml. 

March 14, 2006 - Mary 
  o Compile_annotate : added some comments per meeting with Michael 

March 13, 2006 - Mary
  o Fixed bug in .depend dependencies

March 10, 2006 - Mary
  o tools/escaping/Makefile: removed unnecessary reference to galax.cma

March 10, 2006 - Mary
  o Makefile: changed .depend target to depend only on .mli files

  o In regress/test_core.ml 

    - Changed so that queries on XQTSCatalog.ml are not statically
    type checked, but actual test-suite queries are statically type
    checked.

    - Hard-coded strong typing for "StaticTyping" test group.

  o Schema_simplification.ml:

    Type simplification now takes atomic named type hierarchy into
    account when simplifying types, so that resulting type does not
    contain any types that are derived from other types and therefore
    is guaranteed to be deterministic.

March 10, 2006 - Michael
  o Revived rudimentary version of galax-project. It takes a
    single argument (the query file to be processed) and prints
    the result of path analysis together with any projected
    documents.

March 9, 2006 - Mary
  o Added file locations to warning messages.

March 8, 2006 - Mary
  o Uncommented typeswitch_rewrite rule, which was incorrectly
    disabled. 

March 8, 2006 - Michael
  o Added streaming version of fs:first-item (Code_builtin_fn).

March 7, 2006 - Michael

  o Tuple field use counts (Compile_annotate)

    Added inference rule formulation and lots of documentation;
    added dummy stream consumption in case variables/tuple fields
    are never accessed explicitly - this is necessary in case the
    cardinality of a streamed item sequence must be known (for
    iteration).

    NOTE: Dummy consumption is uniformly applied for both bindings
	  that affect values in their entirety (let-binding) and
	  those that 'chop' values into slices and bind a single
	  slice at a time (for-loop), meaning that, should a use-
	  count-zero-let-binding (which is useless except for an
	  optional type assertion) be discovered in a plan, the
	  stream cursor for the bound value will actually be
	  consumed, though it would not be necessary.

    OBSERVATION: Tuple field use count analysis should take place
		 at the physical algebra level; physical operator
		 variants of one and the same logical operator
		 behave differently at times!

  o Physical typing (Cs_code_typing_top)
    Aligned inference rule formulation and code

  o Builtin functions (Code_builtin_fn)
    Started implementing selected builtin functions over XML token
    streams - fn:count is the first

March 6, 2006 - Mary
  o Fixed bug in tools/escaping/Makefile
  o Added Xquery_type_core_ast_annotation.ml*

March 3, 2006 -  Mary

  o Weak typing is now default

    - Processing_context : changed default to weak typing

    - All 7500 of version 0.8.4 test suite tests pass except these
      two (but they'll pass soon)
      v0.8.4/Queries/XQuery/Expressions/OrderExpr/Orderexpr-19.xq
      v0.8.4/Queries/XQuery/Expressions/OrderExpr/Orderexpr-20.xq

    - All usecases pass 
  
  o Typing_call.compute_type_overloaded_function_call
    Rewriting_rules_typing.overloaded_type_rewrite:
  
    - In computation of type for overloaded function call and in
      conversion of overloaded to statically dispatched calls, fixed
      long-standing bug in which quadratic number of all possible type
      signatures were checked.  Now uses (average case) linear-time
      lookup on first type signature to match.

March 2, 2006 -  Mary

  o Numerous changes to speed up static typing & bug fixes revealed by
    weak typing:

    - Passes all of XQuery Test Suite 0.8.4 except for 12 queries that
      either Galax fails or on queries that have empty() static type.

    - Subtyping_top.includes: Now do syntactic sub-type check on all
      common top-level types (e.g., xs:anyType, numeric*?+, etc.)
      before building & running type automata.

    - Typing_util.can_be_promoted_to : replaced calls to
      Subtyping_top.is_subtype_of by faster Schema_judge.is_derived_from.

    - Xquery_type_core_ast_util.equal_cxtypes: structural equivalence
      on cxtype_descriptor, ignoring File location.

    - Schema_simplification.simplify_ty : any given type is only
      simplified once, replaced by its simplified form, and labeled as
      simplified.

    - Schema_builtin: All builtin types are labeled as simplified.

    - Schema_builtin: All type expressions over builtin types are
      predfined.  

    - Removed sub-type check to terminate recursion in descendant
      type, replaced with simple List membership test.

    - Physical_util.ml: Fixed get_ATOMICTYPE functions to take named
      type hiearchy into account, e.g., get_duration applies to
      duration, yearMonthDuration, and dayTimeDuration values;
      get_decimal to integer and decimal.

    - pervasive.xq: Fixes to signatures 

   TODO : 

    - Compiled type name identifiers.
      Xquery_type_core_ast_util: All declared named types are labeled
      with computed type identifier (pre, post) such that T1 <: T2 iff
      (T1.pre <= T2.pre) and (T2.post <= T1.post).

March 2, 2006 - Philippe
 
  o Fixed typo-bug in twig inlining

March 1, 2006 - Philippe
 
  o TTP rewrite robustness fixes

February 24, 2006 - Jerome

  o Datatypes:
     - Added limited support to check for facets on built-in derived
       integer types.

  o F&O:
     - Implemented fn:namespace-uri-for-prefix().

February 23, 2006 - Jerome

NOTES:
  *** We are now using the W3C XQuery Test Suite 0.8.6.
  *** Function inlining has been turned off by default, since this
      does not handle type assertion on the *output* type of the
      function properly.

  o Processing context:
     - Turned off function inlining by default.
     - Added 'system' flag, which must be turned on to parse
       pervasive, to account for fn: functions being reserved.

  o DateTime:
     - Fixed handling of dates and dateTimes with negative years.
     - Fixed bugs in parsing of xs:duration.

  o F&O:
     - Fixed handling of unknown timezones in fn:timezone-from-*
       functions.
     - Fixed numerous problems with fn:adjust-*-to-timezone functions.
     - Implemented fn:document-uri accessor function.
     - Implemented fn:resolve-QName function.
     - Implemented fn:in-scope-prefixes function.
     - Fixed empty sequence case for fn:QName.
     - Fixed corner case in fn:resolve-uri().

  o Sequence Type Matching:
     - Fixed bug in sequence type matching for element(*,Type) and
       attribute (*,Type) cases.

  o Casting:
     - Implemented casting from xs:double to xs:decimal.

  o Parsing:
     - Fixed lexing of character entities in namespace declarations.

  o Normalization:
     - Now raises an error for variable in the same for and at
       clauses with the same name. (e.g., for $x at $x in ...).
     - Now raises an error for two variables with the same name in the
       prolog.
     - Now properly normalizing whitespace in URI literals.
     - Now properly checking for reserved URIs in function declarations.
     - Now properly checking for duplicate parameters in function
       declarations.

  o XML Schema:
     - Fixed bug, not parsing xs:notation as a schema component.

  o Testing:
     - Switched to test suite version 0.8.6.
     - Added some hackish quoting for the text comparison to deal with
       inconsistencies in the test suite expected results...
     - Added support for external variables bound to query results.

February 23, 2006 - Philippe

  o Added a more conservative rule for pulling TTPs out
    of MapToItem operators as bulk TreeJoins
  o Added a typed rewriting rule for removing sbdo calls
    over expressions that yield subtypes of item()?
  o Fixed the SCJoin self axis to perform nodetests correctly

February 22, 2006 - Philippe

  o Fixed handling of predicates when reverting to
    bulk TreeJoin execution plan
  o Some fixes in SC-Join operators

February 20, 2006 - Michael

  o Fixed severe bug in Streaming_xpath (uncaught Stream.Failure).

  o Minor fixes to tuple field use count analysis in
    Compile_annotate.


February 18, 2006 - Philippe

  o SC Join code completed. A lot of bugs remain to be
    dealt with.

February 16, 2006 - Michael

  o Compile_annotate
    - Started refining tuple field use count analysis as to
      more closely reflect physical operator implementations.
    - It is still incomplete, but works correctly on approx.
      the subset of the algebra that should be streamable.

February 15, 2006 - Jerome

  o F&O:
     - Fix to fn:seconds-from-duration, wrong for negative seconds.
     - Implemented fn:dateTime, missing.

  o DateTime:
     - Fixed bug with 24:00:00 time.

February 15, 2006 - Jerome

  o DateTime:
     - Added support for xs:duration missing.

  o Namespaces:
     - Internal renaming of xdt: into xs:. Surface syntax still uses
       xdt: for compatibility with the current test suite.

  o Pretty-printing:
     - Fixes to pretty-printing of GroupBy.

February 15, 2006 - Mary & Doug

  o Added AOEParse to list of monikers in planio_print_common.ml
  o Fixed bugs with printing and parsing of query plans

February 15, 2006 - Michael

  o Evaluation_expr
    - Added debugging code that counter-checks values against
      inferred physical types at runtime. This is compiled out
      by default.

  o Cs_util
    - Removed unnecessary item cursor conversion in _old_
      build_add_var_xml_value_unsafe_allowed code.

February 14, 2006 - Philippe

  o Bug fix to the inlining rewrite

February 14, 2006 - Michael

  o Added several refinements to physical typing

February 14, 2006 - Philippe

  o Changed the variable inlining rewrite to be restricted
    to XPath expressions by using the SBDO XPath annotations.
    It is still possible to inline all variables by using the
    -inline-variables switch.
    NOTE: The inlinings that happen now are sufficient for the
    introduction of TTPs and *should* not impede the introduction
    of joins. For instance, the expression 
      doc("x")/a[doc("y")/b = c]
    should be compiled/optimized as a join expression.
  o Added scjoin algo's for descendant-or-self, ancestor-or-self
    following and preceding
  o Fixed XPath annotation at the core level to include the
    ddo-calls generated by predicates  

February 13, 2006 - Philippe

  o Rewriting: added optimization rule for streaming plan support
    that eliminates residual MapToItem over MapFromItem ops

February 13, 2006 - Michael

  o Fixed bugs in code building functions that depend on physical
    typing (let, some, every). The test suite now passes with
    -code-selection-by-physical-type on.

February 12, 2006 - Michael

  o Fixed bugs in physical typing

  o Fixed bug in Code_functional_ops.build_let_code

February 12, 2006 - Philippe

  o Rewriting: added backup rule for inlining static uri
    variable bindings. This enables the compiler to introduce
    AOEParse ops, when variable inlining is turned off.

February 12, 2006 - Michael

  o Physical typing
     - Lifted variable and tuple field use count access to physical
       typing
     - Isolated code building functions that deal with variable/tuple
       field bindings in Code_binding. Added separate functions for
       binding sax values, item cursors and item lists
     - Parameterized physical signatures of variable/tuple field
       binding operators by the physical types of the corresponding
       variable/tuple field. This circumvents the need to define
       distinct physical operator versions for the possible kinds
       of bindings that can occur, namely, binding a sax value, item
       cursor or item list.
     - Made various fixes to type_check_physical_op

  o Code selection
     - Established a stronger coupling between physical typing
       and code selection of variable/tuple field binding operators
       by selecting code in Code_binding according to the physical
       type parameters of those operators
     - This behaviour is shielded by the new command line option
       -code-selection-by-physical-type (off by default).

     - It currently affects AOELet, AOESome, AOEEvery, AOEMapFromItem,
       AOECreateTuple

     - Unless turned on explicitly, code selection, variable/tuple field
       use count access and code building functions for those bindings
       are treated as before. The old code is still in place (and needed
       by., e.g., AOENullMap etc.).

        Note: Physical typing is still buggy. The bugs did not show up
	      before, because code selection was not really controlled
	      by physical typing, but by its own logic. Now, the two
	      phases interact more, and pyhsical typing errors result
	      in concrete, possibly wrong, materialization behaviour.

February 11, 2006 - Philippe
  
  o Bug fix in MapConcat over TTP rewrite

February 10, 2006 - Philippe
  
  o Staircase join for Following

February 9, 2006 - Jerome

  o Rewritting:
     - Turned variable inlining off again, but added a switch to turn
       it on (for Philippe).

  o Pretty-printing:
     - Added pretty-printing of physical signatures in physical plans
       (showing the physical operator and the physical types for the
       input and the output).
     - Fixed other minor pretty-printing issues for query plans.

February 9, 2006 - Michael

  o Physical typing of constructors now delegates operator selection
    to Code_constructors. It differentiates between 'virtual' streamed
    and materialized versions of constructors if -streaming on.

February 9, 2006 - Michael

  o The decision wether to stream a parse operator or not
    is now controlled by path annotations (if -streaming on).

  o The decision wether to stream a treejoin operator or not
    is now controlled by its input's physical type (if -streaming on).

  o Prepared decision making for constructors:

    For each physical constructor operator PO*, split up that name
    into PO*_Stream and PO*_Materialized (in analogy to POParse_Stream
    and POParse_Load)

    NOTE: This is a purely 'virtual' distinction; we do not and won't
	  have materialized implementations of constructor operators.

    We need that distinction in order to properly propagate 'non-
    streamability' along data flow paths that start with a constructor.

February 9, 2006 - Philippe

  o Full Axis Feature in place. SC-Join version of following
    and preceding coming up.

February 8, 2006 - Jerome

  o Overloaded functions:
     - Cleaned up normalization/overloaded.ml module. Removed dead
       code, reorganized the tables.

  o FLWOR expressions:
     - Fixed orderby code to deal with type promotion. (Note: this is
       only doing a pairwise comparison, this may actually not give
       the proper result in the case of decimals when all should be
       promoted to float/double and the precision is an issue).

  o Prolog:
     - Added minimal support for collation declaration, at least
       checking the constraints on which declaration is valid or not.

  o Testing:
     - Now properly resetting the base uri processing context.

  o Validation:
     - Fixed propagation of nilled property during validation.

  o F&O:
     - Fixed support for distinct-values, numerous problems with type
       promotion and handling of NaN.
     - Fixed fn:subsequence for negative starting points.
     - Implemented fn:nilled function.

February 8, 2006 - Philippe

  o Started adding full axis feature.

Februari 8, 2006 - Philippe

  o Fixed bug in nested loop ancestor code (dm)
  o Staircase join versions of ancestor and descendant
    are now in place. NOTE: The ancestor algorithm
    is optimized to use the DOM parent pointers. This is
    because an ancestor scan would be suboptimal due to the 
    lack of skipping posibilities in our name indices.
    The sc-join version of ancestor more or less corresponds
    with nested loop + pruning. Early experiments show 
    substantial speedups for ancestor and descendant.

February 8, 2006 - Michael
  
  o Path analysis now produces annotations for the logical
    AST. These are carried over to the physical AST, and
    can be used to guide decisions at code selection.

February 8, 2006 - Michael
  
  o Plugged path analysis into the processing model as an
    analysis preceding code selection.

    It is controlled by the command line flag -streaming on
    (defaut is off) that I want to use as a global switch for
    streaming functionality, later.

    Path analysis currently just prints the analysis result
    _instead_ of producing logical annotations; this will be
    fixed soon.

February 7, 2006 - Michael
  
  o Added code that calculates tuple field use counts on the basis
    of existing returned/accessed tuple fields information (still
    overly conservative, probably).

    Setting -allow-streamed-tuple-fields on enables that code
    to selectively materialize tuple fields if necessary;
    -allow-streamed-tuple-fields off (default) conservatively
    materializes every field.

    This option should always be used with the corresponding
    -allow-streamed-variables option, since otherwise it's hard to
    actually get a stream into a tuple field.

  o Enhanced code_selection_context by an additional field
    global_annotation that is used to communicate tuple field
    use counts from the analysis to code_selection (resp.
    physical typing in the future).

    _Global annotations are not yet stored for global variables._

February 4, 2006 - Philippe
  
  o Fixed bug in new code selection for TupleTreePatterns
  o Added ancestor and descendant code for SC join
  o Added some interface functions for different kind of
    index scanners:
    - scanner over a pre/post window
    - scanner starting at the first node with pre-order >= p
    - scanner starting at the first node with post-order >= p

February 3, 2006 - Mary

  o Fixed bug in function inlining in Optimization_rewrite

    Function inlining is implemented by applying the standard rule of
    replacing formal arguments by let-bound variables.
 
    declare function foo ($x,$y) =  { BodyExpr }; 
    
    foo (Expr_x, Expr_y);
    foo not recursive 
    ==>
    let $x := Expr_x, $y := Expr_y  return BodyExpr

  o Added command-line option "-inline-functions [on/off]". 
    Default is on, as inlining seems relatively well-behaved at the
    moment. 
   
  o Aligned all operators to take (asequencetype option, cvname)
    pairs in that order.

  o Pretty-printing of query plans: Fixed more bugs in AOELet and
    AOEVarDecl.

February 2, 2006 - Jerome

  o Testing:
     - Now catching low-level failures of the system and report those
       neatly as test results.

  o Prolog:
     - Raises proper errors for redeclaration of the xml and xmlns
       namespaces.

  o F&O:
     - Fixed fn:resolve-uri: reversing the order of parameters,
       dealing with the empty sequence properly.

  o Pretty-printing of query plans:
     - Added support for precedence in pretty-printer.
     - Fixed numerous problems with boxes in the pretty-printer.
     - Fixed bug in pretty-printing of 'treat as', 'cast as',
       'castable as', (inverted parameters).
     - Fixed bug in pretty-printing of fn:error (wrong namespace, and
       only allowing one independent expressions instead of several).
     - Made the printing of the algebra more consistent. Some more
       work aligning with the algebra paper and/or the XQuery core...


February 2, 2006 - Mary

  o Physical typing
    - First implementation of physical typing     
    - NOTE!!  Function inlining is disabled as it raises physical
      typing errors.     

    Created Xquery_physical_algebra_ast_util.ml*
            Xquery_physical_type_ast.mli

    Renamed Physical_types.mli  to Physical_values.mli
	    Physical_values.ml* to Physical_values_util.ml*
    
  o Terminology fix-up
    The term "materialized" was overloaded in Physical_types and
    Physical_values: It meant "Dm.item cursor", "Dm.item sequence",
    and "(Dm.item sequence) array".  

    In Physical_value.mli:
       Sax   denotes (SAX-token cursor)
       Dom   denotes (item list) or (item cursor)
       XML   denotes Sax or Dom
       Tuple denotes XML array
       Table denotes Tuple array

    I've tried to clean this up by using "materialize" as a verb in
    functions that take an ('a cursor) to an ('a list) or ('a array),
    instead of using it as a noun to describe values.

  o Note/TODO: 

    We are not compiling some/every into map operators as described in
    paper. 

February 1, 2006 - Philippe
  
  o Added a rule for removing snaps over operator trees
    without actual updates. For now, the rule is only 
    triggered by using '-treejoin-log treejoin'

January 31, 2006 - Mary and Doug

  o Printing plans in XML
    - Started fixing bugs in Planio modules for printing/parsing plans

January 31, 2006 - Philippe

  o Fixed pruning rewrite in optimizer
  o Fixed pruning to support other axes
  o Introduced index windowing for Twig Joins

January 31, 2006 - Michael

  o changed physical algebra AST annotation to pass on
    path annotations from the logical algebra level

January 30, 2006 - Michael

  o added support for streams in variables, namely in
    AOELet, AOESome, AOEEvery, AOEMapFromItem

    - streams are disabled by default; they can be activated
      by setting '-allow-streamed-variables on' (in that case,
      selective materialization is controlled by variable use
      counts as is the case for item cursors -> item lists)

  o added support for streams (and item cursors) in tuple fields

    - streams (and item cursors) are disabled by default;
      they can be activated by setting '-allow-streamed-tuple-fields on'
      (in that case, *no materialization is applied at all*, currently;
      this will be controlled by tuple field use counts, soon)

January 28, 2006 - Jerome

  o Prolog:
    - Raise error in case of multiple boundary space declarations.
    - Raise error in case of multiple default namespace declarations.
    - Raise error in case of multiple construction declarations.
    - Raise error in case of multiple ordering declarations.

  o F&O:
    - Switched the order of the parameters in fn:lang (the node comes
      last not first). Allowed the first parameter to be the empty
      sequence, which is treated as the empty string.

  o Data Model:
     - Fixed bug in node_lang method, not catching the end of stream
       accessing xml:lang attributes properly.

  o Optimization:
     - Fixed bug in function inlining, not checking properly for
       QNames equality, and not dealing with mutually recursive
       functions.

  o Testing:
     - Fixed bug in the test suite, not properly reinitializing the
       processing context.

January 27, 2006 - Mary

  o Added missing code_selection/cs_code_typing_type.ml*.
    compile.

    Still working on physical typing rules...Revealing all sorts of
    nastiness in code selection.

  o More open questions on physical typing: 

     I'm not sure about the physical typing for GroupBy's dependent
     subexpressions. The formal definition says that it has two
     dependent sub-expressions, one that takes an Item value and one
     that takes a Tuple, so each has a different type for INPUT.  

     The creation of a GroupBy in Optimization_rewrite, however,
     indicates that it only has one dependent subexpr, which takes a
     Tuple to an Item.

January 26, 2006 - Mary

  o Physical typing
 
    - Added physical-typing/operator-selection sub-phase to code
      selection in Cs_code_typing_top in code_selection/.

    - Physical typing is still a NOP, but it is in place and doesn't
      break anything. 

    - Removed all the hard-coded physical signatures in
      Cs_util_coercion.

  o Open questions on physical typing: 

    - What are physical types of functions' actual arguments & return
      values?  They can/should be context dependent if we do inlining.
      For now, they are always completely materialized XML values.

    - What is the physical type of an externally declared variable?
      For now, they are always completely materialized XML values.  In
      theory, we could permit users to provide Sax token-streams as
      input.  BIG impact on API. 
    
January 26, 2006 - Philippe

  o Fixed rules to reintroduce treejoins in the optimizer to be more
    robust

January 26, 2006 - Philippe
 
  o Added basic documentation for name indices

January 26, 2006 - Philippe

  o Fixed an incredibly stupid bug in xquery_algebra_ast_util
    that was screwing up the twigstack evaluation

January 25, 2006 - Philippe

  o Added functions for checking applicability of twig joins.
  o Added empty staircase join modules
  o Factorized some AST functionality from twigjoins to algebra_ast_util
  o Fixed rewrite rule in optimizer to cover more cases 

January 24, 2006 - Mary

  o Re-organization of Code_selection modules & directories

    The physical implementations for the algebraic operators are (more
    or less) organized according to the groups in the ICDE compiler
    paper.  The operators with multiple or complex implementations,
    e.g., Join, GroupBy, etc., are in their own modules.  The
    operators with related, simple implementations are grouped
    together.  

    There is room for more improvement/refinement here, but I think
    this re-org meets our current needs.

    In code_selection/code:

    Code_functional_ops : Let, If, Var
    Code_builtin_fn     : BuiltinFn, ConvertSimple, PromoteNumeric, UnsafePromote
    Code_user_defined_fn: UserDefinedFn
    Code_overloaded_fn  : OverloadedFn
    Code_constructors   : All Node constructors, Scalar, Seq, Empty, Error
    Code_type_operators : Treat, Validate, Cast, Castable
    Code_typeswitch     : Typeswitch
    Code_item_tuple     : Some, Every, MapToItem, MapFromItem
    Code_tuple          : Input, Create, Access, Concat, Product, Select, Materialize
    Code_join           : Join, LeftOuterJoin
    Code_map            : Map, MapIndex, MapIndexStep, MapConcat, OuterMapConcat,
                          NullMap
    Code_group_order    : GroupBy, OrderBy
    Code_update         : Copy, Delete, Detach, Insert, Rename, Replace, Snap
    Code_parse          : Parse
    Code_treejoin       : TreeJoin
    Code_tuple_tree_pattern : TupleTreePattern
    Code_prune          : Prune
    Code_nestedloop     : Distinct 
    Code_keyref         : KeyRef

    Utility functions, grouped by the operators they support, are in Code_util_*.

January 23, 2006 - Chris

  o Shredded Changes
    - Fixed main memory btree and record bugs relating to cursor semantic
      o Aligned with BDB's semantic for corner cases
    - Added replace to shredded datamodel. 
      o Only limited testing of shredmm:// (main memory shredded version).

January 23, 2006 - Jerome

  o Constructors:
     - Fixed another bug in handling of whitespace in
       processing-instruction constructors.
     - Fixed handling of boundary whitespace: now defined as per the
       specification, done cleanly at normalization time without
       relying on strange empty text nodes, and dealing character
       references properly.

  o F&O:
     - Fixed bug in fn:substring-before, in case there is no match.
     - Fixed error in fn:matches, fn:replace, fn:tokenize, when
       dealing with invalid regular expressions.
     - Added consistency check fn:replace, for which the replacement
       text shouldn't contain a standalone '\' or a '$' not followed
       by a digit.
     - Fixed bug in fn:compare, not handling empty sequence properly.
     - Fixed fn:codepoints-to-string(), to handle the proper range of
       character sets.

  o Parsing:
     - Implemented the proper normalization of end of lines in string
       parsing.

  o Testing:
     - Fixed bug in test harness, not handling non-deterministic
       queries where both a result and an error are possible.

January 23, 2006 - Philippe

  o Optimization_judge:
    - Fixed bug in order_dup_matters function caused
      by structural/physical equality issue

  o Optimization_rewrite:
    - Fixed a structural/physical equality bug in 
      nested_null_index rewrite
      This was causing the tests fn-node-name-2.xq and
      last-10.xq of the XQTS to fail under weak typing.

January 19, 2006 - Mary

  NB: Note file name changes

  o TupleTreePattern code moved to
    code_selection/code/code_tuple_tree_pattern.ml*
  o TwigJoin code moved to code_selection/code/code_tj_*stack.ml*
  o NestedLoop code moved to code_selection/code/code_nestedloop.ml*


January 19, 2006 - Mary

  o Started changes to support Physical Typing phase that precedes
    Code Selection

    - Added code_selection/code sub-directory
      For each algebraic operator OP, there is/will be a module that
      contains all its physical implementations in
      code_selection/code/code_OP.ml* Any code in
      cs_code_selection_expr and cs_code that is specific to OP
      is/will be moved into code_OP.ml.   

    - Xquery_physical_algebra_ast.mli : names of physical
      operators.

January 19, 2006 - Philippe

  o Added functionality for statically determining the
    relevance of order (optimization_judge)
  o Fixed some rewrite rules to take order-relevance of
    on operator's context into account

January 18, 2006 - Philippe

  o Twig Joins:
     - fixed a bug related to subsequent evaluations of the same
       tree patterns
     - added optimized cursor navigation by using B-Tree lookups
       for skipping nodes

  o Optimization:
     - added code that detects whether order/duplicates matter for
       a certain operator. This avoid the wrongful rewriting of 
       FLWORS in tree patterns

  o Fixed a bug in dynamic_stack.top

  o Added a function to jump to a certain node for a twigjoin source
    in physical_dm_name_index

January 17, 2006 - Jerome

  o Constructors:
     - Fixed bug in lexing of single-quoted strings containing
       character references.
     - Fixed bug in document loading, wrongly discaring text events
       containing the empty string, resulting from text node
       constructors containing the empty.
     - Fixed computed text constructors to deal properly with the
       empty sequence as input. Created
       fs:item-sequence-to-untypedAtomic-optional special function
       which propagates the empty sequence properly instead of
       returning an empty string (which now results in an actual text
       node).
     - Fixed bug in handling of whitespace in processing-instruction
       constructors.
     - Fixed element constructors and element nodes in the data model
       to propagate base_uri properly from the context.

  o F&O:
     - Fixed bug in EBV (effective boolean value), not returning an
       error for multiple items the first of which is not a node.
     - Fixed fn:base-uri() to normalize to fn:base-uri(.)
     - Fixed fn:concat to cast atomic values to xs:string.
     - Fixed fn:normalize-space() to normalize to
       fn:normalize-space(.)
     - Implemented support for fn:normalized-unicode().
     - Implemented support for fn:encode-for-uri().

  o Data Model:
     - Fixed bug in base_uri method, not catching the end of stream
       accessing xml:baseuri attributes properly.

  o Base:
     - Added ./base/glx_camomile to encapsulate some of the calls to
       Camomile.

January 13, 2006 - Jerome

  o Query plans I/O:
     - Removed obsolete algebra printer from ./planio.

  o Dates and time:
     - Fixed all the operators on dates, which should rely on
       dateTimes to deal with starting instants and timezones
       properly.
     - Fixed comparison operators between dates, which was using a
       non-tail recursive function which was not scaling for large
       dates.
     - Fixed treatment of leap years, which are not only computed
       starting 1582. Also XQuery uses a proleptic Gregorian calendar
       (http://en.wikipedia.org/wiki/Proleptic_Gregorian_Calendar),
       for which AD 4 and AD 8 are indeed leap years.
     - Fixed bug in treatment of dates, there is no year zero, the
       year before year one is year -1.
     - Fixed bug in printing of negative years.
     - Fixed implementation of subtraction of duration from a date,
       very buggy.
     - Fixed arithmetic operators on times to preserve the absence of
       timezones properly.
     - Fixed arithmetic operators on times to deal with time overflow
       properly.
     - Added support for gregorian types: xs:gYearMonth, xs:gYear,
       xs:gMonthDay, xs:gDay, xs:gMonth.

  o QNames:
     - Fixed bug in equality for QNames, which should ignore the
       prefix.

  o Constructors:
     - Provided proper XQuery error message when parsing a character
       reference which is out of range. (Was breaking the test suite).
     - Implemented the proper attribute value normalization for
       attribute constructors (normalizing new lines, whitespaces,
       etc.).
     - Implemented the proper xml:id attribute extra value
       normalization.

  o Casting:
     - Fixed casting of xs:decimal to xs:integer to truncate the
       decimal part properly.
     - Implemented casting of xs:float and xs:double to xs:integer.
     - Implemented casting of xs:boolean to numeric values.

  o Syntax:
     - Fixed parsing bug for xquery declaration in the prolog.

  o Testing:
     - Fixed test harness to handle test-cases in non-leaf parts of
       the test hierarchy properly.

January 12, 2006 - Philippe
  
  o Fixed optimization_rewrite to deal with tree patterns
    starting from variable references in a orderly fashion

January 11, 2006 - Michael

  o Purely syntactical change:
    Changed logical_algebra_types to contain path_annotation;
    affected modules (syntax, only):
    Logical_algebra_types,
    Xquery_algebra_ast_util, Xquery_algebra_ast_annotation_util,
    Compile_expr, Compile_util, Compile_top,
    Optimization_rewrite,
    Compile_context, Code_selection_context
    
    Excised a function from Compile_context to its own module
    Compile_context_util in order to resolve a cyclic dependency
    and isolate the transition from generic to logical types.
    

January 11, 2006 - Philippe
  o More small changes to align Twig Join code better with the paper

January 11, 2006 - Philippe
  o Small change to align Twig Join code better with the paper

January 10, 2006 - Mary & Jerome

  **** Began re-organization of Code_selection module to support
       physical types and type-dependent code selection. 

  o Xquery_algebra_ast_annotation: 
    
    Added physical_type to align with physical_value

  o Added new subdirectory code_selection/code

    Do "cvs update -d"!

  o Cs_code_selection_expr :

    Started to move operator-specific code into individual modules in
    code_selection/code. 
 
    E.g., AOEParse => code_selection/code/code_parse
          AOETreeJoin => code_selection/code/code_treejoin

January 6, 2006 - Jerome

  o Testing:
     - Removed obsolete config files for XQueryUnit.

January 6, 2006 - Mary

  o Fixed yet another bug introduced in Monitor on December 9 : 
    when nested document loads occured in evaluation phase, did not
    record actual evaluation time that followed doc load

January 6, 2006 - Jerome

  o Index support:
     - Fixed bug in that indices should be allocated before doing code
       selection for global variables. [Bug reported by Philippe.]

January 6, 2006 - Philippe

  o Xquery_algebra_ast_util: fixed a bug related to restoring
    output fields of a treepattern
  o Some fixes / fine-tunings to the twigstack implementation


January 4, 2006 - Jerome

  o Testing:
     - Added the standard configuration flags to galax-test.

  o Configuration management:
     - Separated "optimization options" into: "logical optimization"
       options, "code selection options", and "runtime" options.
     - Factorized the code which creates sets of options, avoiding
       code duplication (and the corresponding possible
       misalignments).
     - Changed the following configuration parameters:
           -execute-normalized-expr
           -execute-logical-plan
           -execute-physical-plan
       to  -execute [standard/normalized/logical/physical]
       with standard XQuery 1.0 syntax being the default.
     - Changed the following configuration parameters:
           -print-error-code on/off
       to  -print-error-kind [standard/code]
     - Removed obsolete update optimization flags:
    	   -update-hack-oi
  	   -update-hack-bi
  	   -update-hack-sbi
     - Put back a flag to print query plan in XML:
           -print-plan-kind [standard/xml]

  o Query plans I/O:
     - Hooked back the query plan XML printer to the processing model.
     - Fixed bug in printing of plans with snap operations.


January 4, 2006 - Philippe

  o Twig joins are now in place. They need more testing before
    benchmarking can start.
  o Bugfix in AST utility function for treepatterns
    (get_sub_node_indices)

December 29, 2005 - Jerome

  o Arithmetics:
     - Removed native support for xs:int, raising numeric overflow
       problems.

  o Date and time:

     - Fixed canonical lexical value for xdt:yearMonthDuration and
       xdt:dayTimeDuration.
     - Added missing inequalities for date and time types.
     - Fixed division between two xdt:yearMonthDurations to rely on
       the proper integer division.
     - Fixed bugs in xdt:dayTimeDurations division.

  o F&O:
     - Fixed many bugs in fn:number().

December 28, 2005 - Mary

  o Fixed bug introduced in Monitor on December 9 : double reporting
    of document load times. 

December 22, 2005 - Jerome

  o XML Schema:
     - Added preliminary support for derivation by extension. This has
       been tested on complex types with simple content derived from
       simple type (notably used heavily to support elements with
       attributes and simple content). Needs more testing.

  o XPath:
     - Fixed bug in handling of following-sibling and
       preceding-sibling axis.

  o Atomic types:
     - Fixed equality for the base64Binary and hexBinary types.

December 22, 2005 - Philippe

  o Pathstack code cleanup

December 20, 2005 - Philippe

  o Code selection: fixed a bug in nested loop style
    evaluation of treepatterns.
  o Streamlined pathstack algorithm some more. Added
    in-place dulicate elimination to pathstack.

December 20, 2005 - Philippe

  o Code selection: 
      - Removed some inefficiencies in pathstack.
      - split up twigjoins in three modules:
        pathstack, twigstack and a common part
  o Optimization:
      - Added a rewrite rule to clean up the residual
        sbdo optimization around treepatterns.
  o AST: Added a field to twig nodes indicating wether or not
    the output of it's field should be restored. This is used
    by twig joins to determine what part of the matching 
    patterns should be restored. I need to keep the field names
    because nested loop evaluation uses these names as dot-fields.
  

December 15, 2005 - Jerome

  o Physical Data Model. Implemented support for XML streams inside
    tuple fields:
     - Removed unnecessary usage of xml_value in the code (since now
       it is either a materialized or a streamed value, the various
       code bits should distinguish between those).
     - New physical data model types are all grouped in the same
       interface: ./physicaldm/physical_types.mli
     - table materialization forces materialization of the tuple
       fields.

  NOTE: allowing more value in tuple fields has revealed how little
  understanding we have of even item-cursor query plans. I had to
  change the implementation of built-in functions to operate on lists
  for now. This is bothering, but not surprising, hopefully we can
  address that part of query planning as part of the large SAX
  vs. materialized question.

  Things to do:
    (1) reorganize the code more consistently per physical type (e.g.,
  SAX physical operators on one side, materialized operators on the
  other side, etc.) another approach is to have a file per algebraic
  operators with variants at the physical level.
    (2) clean up of physical signature (used in coercion, code, etc.)
    (3) more testing is necessary. This passes the current test-suite
        and the usecases.

December 15, 2005 - Philippe

  o Tweaked name index loading so that both the pre-order
    and post-order b-trees are bulk-loaded in order, eliminating
    the need for sorting the b-trees during finalization.

December 15, 2005 - Philippe

  o Removing Twig index (Will switch to more efficient
    name index instead)

December 13, 2005 - Mary

  o Procmod_phases 
    - Fixed nasty little bug in which _all_ phases were curried and not
      evaluated/forced until serialization!

  o Testing 
    - Added Construct test-group bugs to regress/knownbugs.xml & bugzilla

December 13, 2005 - Philippe

  o Added support for Twig Join indices.
    All over the place.

December 13, 2005 - Jerome

  o Alignment:
     - Fixed bug in support for binary operators, not checking
       properly that the operands are either a node or the empty
       sequence (e.g., () is 100 was succeeding).
     - Fixed normalization of processing instructions, broken (using
       wrong Regexp to check for '?>' content.
     - Fixed bug in fn:node-name(), can now take the empty sequence as
       input.
     - Implemented fn:prefix-from-QName().
     - Re-implemented fn:namespace-uri-from-QName().
     - Fixed serialization/printing of values of type xs:QName.

  o Testing:
     - Added support for handling bugs located in the test
       suite. Those are flagged as 'inspect' instead of doing a real
       comparison with the test-suite expected result.

December 9, 2005 - Mary

  o Implemented weak typing.  

    To use: -static-typing on -typing weak
    In usecases/Makefile: change to TYPE_KIND=weak.  Default is "none".

    Required changes to:
    normalization/overloaded.ml
    rewriting/rewriting_rules_typing.ml
    schema/schema_judge.ml
    typing/typing_call.ml
    typing/typing_call.mli
    typing/typing_errors.ml
    typing/typing_expr.ml
    typing/typing_fn.ml
    typing/typing_util.ml
    usecases/Makefile

    Weak typing tripped bug in NS Uscase, which needs to be reported:
    usecases/ns_context.xq
    usecases/ns_usecase.xq

  o Changed config/Makefile.gen to use camomileunpack.cma (no
    'Camomile.' package prefix) instead of camomile.cma ('Camomile.'
    prefix) to accommodate both O'Caml 3.08 and 3.09.

    code_selection/cs_code_fn.ml
    config/Makefile.gen

  o Fixed bug in printing of string literals
    ast/xquery_common_ast.ml
    ast/xquery_common_ast.mli
    ast_printer/print_common.ml

  o Fixed bug in Galax API -- materialization function not wrapped in monitor
    galapi/caml_api/galax.ml

  o Fixed bug in Monitor 
    - Nested document loads not reported in results. 
    monitor/monitor.ml
    monitor/monitor.mli

    Analysis time not wrapped in procmod/procmod_phases.ml

December 9, 2005 - Jerome

  o Alignment:
     - Fixed bug in serialization of decimal values without
       significant digit after the decimal point.
     - Fixed several bug in serialization of xs:float and
       xs:double. (not handling the number of digits after the decimal
       point properly).
     - Fixed bug in fn:round over xs:decimal.
     - Implemented fn:round-half-to-even.
     - Added proper exception for division by zero in the xs:integer
       and xs:decimal cases.
     - Fixed idiv support, just completely wrong again...
     - Fixed mod support, now aligned with idiv properly.

  o Testing:
     - Added support for multiple output files (i.e., multiple
       possible results for cases where part of the semantics is
       implementation-defined or implementation-dependant).

December 8, 2005 - Philippe

  o Bugfix #1: Restricted inlining of TupleTreePatterns to
    only one selecting child twig.
  o Bugfix #2: Removed infinite looping from the 
    map_concat_pushdown rewrite rule. Apparently, when a 
    MapConcat was pushed down another MapConcat, the optimizer
    ended up switching around the two MapConcats eternally.
    Fixed this by restricting the rewrite to non-MapConcat-ops.

December 7, 2005 - Philippe

  o Bugfix in compile_expr.compile_flwor
    whenever dot is explicitly used (e.g. step1[./step2]), this may 
    result in a core expression of the form
    let $dot := $dot return TreeJoin[axis::nt]($dot)
    If this happens, you need to make sure to compile the independant
    subexpression of the let before hiding the corresponding dot-variable 
    field from the compilation context. This did not happen, but is now 
    fixed
  o Added a small rule to deal with the returned let-bindings in 
    compiled path expressions in optimization_rewrite.

December 4, 2005 - Jerome

  o Data model reorg:
     - Moved support for items from ./dm to ./physicaldm. This is to
       prepare for implementing support for item sequences (including
       cursors) inside tuples.
     - Removed dependency of AST over atomic types. 
     - Turned atomic values into 'protovalues' in the AST represented
       directly as atomic_type.
     - Added special core expressions and algebraic operators for
       fs:functions which uses protovalues (fs:convert-simple-operand,
       fs:promote-to-numeric, fs:unsafe-promote-to-numeric)
     - Propagated the change throughout.
     - Move atomic values support for dm/atomic (from datatypes).

  NOTE: I haven't fully checked that the new support for protovalues
  does not break some of the optimizations (either in rewritings or
  optim/joins). This will require Mary/Chris's review.

  o Derived atomic types:
     - Added support for type annotations on atomic values in the data
       model.
     - Fixed casting and atomic value construction to take a type name
       as input.
     - Fixed serialization to print the proper type constructor based
       on type annotations.

  NOTE: this does not check for facets yet! only supports the type
  name hierarchy. Allowing to construct, and process atomic values
  derived from primitive atomic types (e.g., xs:short("42").

  o Alignment:
     - Fixed fn:boolean, wrong for xdt:untypedAtomic(""), and
       xs:double("nan").

December 02, 2005 - Philippe

  o optimization: clean up of some rules + making some rules more
    robust

  o rewriting: making loop fusion more robust and adding loop merging
    rewrite (makes picking up chained predicates easier at the
    algebraic level)

December 01, 2005 - Michael

  o /projection:

    Added modules Alg_path_struct, alg_path_structutil,
    Alg_path_analysis, Alg_analysis_context that implement
    path analysis functionality over the XQuery algebra
    (in analogy to the corresponding modules without 'Alg'
    prefix; they still coexist).
    NOTE: Algebra AST annotations are prepared,
	  but not yet included.

  o /toplevel:
    Changed Galax-project to employ the new algebra-level
    path analysis instead of the old core-level analysis.

December 1, 2005 - Philippe

  o Cleaned up default (nested loop) code generation for
    TupleTreePatterns + added support for chained predicates

November 30, 2005 - Philippe

  o cs_code_nestedloop.ml:
    Added dedicated module for nested loop code generation

  o Added support for loop fusion over chained predicates in
    rewriting_rules_notyping

  o Support for rewriting chained predicates into tree patterns in
    optimization_rewrite

  o Added support for merging multiple twigs at a time in
    xquery_algebra_ast_util

November 25, 2005 - Philippe

  o Added loop fusion rule for normalized predicates
    in rewriting_rules_notyping.ml

  o optimization_rewrite.ml: 
     - Removed nested MapTiItem rewrite
     - Changed syntactical treepattern introduction to
       exploit the advantages of loop-fusion, disabling
       the need for a complicate 'nested MapToItem'
       rewrite
    NOTE: TreePatterns are now correctly picked up for 
          predicates when static typing is enabled

November 25, 2005 - Mary

  o Fixed bug : Missing file location in error reporting.

November 25, 2005 - Mary

  o Fixed bug in typing of Cast --- code clean-up introduced error.

November 23, 2005 - Philippe

  o Changed some optimization rules to seamlessly interoperate with
    static typing rewrites and distinct-docorder calls

  o Fixed the pathstack algorithm

November 22, 2005 - Mary

  o Unicode support

    *** NOTE! ***

    - Added Camomile Unicode library Version 0.6.2
      Download Camomile from : 
      http://prdownloads.sourceforge.net/camomile/camomile-0.6.2.tar.bz2
      Follow INSTALL instructions.

      (Version 0.6.3 is fine too, but requires O'Caml 3.09.  Use 0.6.2
       if you are still using O'Caml 3.08)
    - Changes to Makefiles for Camomile
    - Implemented fn:string-to-codepoints(), fn:codepoints-to-string()

  o Support for Constructors

    - Numerous changes for testing XQuery Test Suite Construct test
      group.  Currently, passes 385/423
    - Aligned normalization of computed constructor names,
      processing-instructions, and comments with spec.
    - New exception Error.Constructor_Error
    - Propogation of namespaces in core-to-algebra compilation 
    - Added NCNAME to Datatypes_lexer
    - Streaming_constructors: added rules for checking dynamic content
      of computed processing-instruction & comment constructors

  o Cast
    - Added namespace environment to Cast AST nodes in Core/Algebra.
      Necessary for casting strings/untyped to QNames.

  o Parsing 
    - Aligned Constructor productions with spec
    - Added item type:

      document-node { element-type | schema-element-type }

  o Static Typing
    - Subtyping_build_regexp, Subtyping_letter: Added support for document-node types 
    - Typing_step: Added support for kind tests

  o Usecases
    - Updated output of SGML, String, XMark to align with entity ref
      serialization. 

November 18, 2005 - Jerome

  o Compilation:
    - Removed cyclic dependency between ./schema and ./dm modules.

  o Processing context:
    - Added support for schema location hints.
    - Moved the alive_documents to the compilation context. This
      removes a problematic dependency on the data model.

  o Testing:
    - Added support for schema import, based on locations in the
      catalog.

November 17, 2005 - Philippe

  o Near-completion of twig join algorithm. The algorithm
    works fine, but a problem with materialization of result
    tuples persists

November 16, 2005 - Jerome

  o I/O:
    - Fixed long-standing problem with file handles not being closed
      after parsing. (Occurring in SAX parsing, and utilities used to
      bulk-load files).

  o Streaming:
    - Fixed stream-discard operations to fully consume the stream,
      hence closing the corresponding opened files...

  o Testing:
    - Fixed test-harness to properly lookup the source files from the
      catalog.
    - Implemented the "Inspect" comparison (always succeeds, and
      flagged to facilitate actual inspection).
    - Added hashtable lookup for source files, avoiding repeated
      parsing.
    - Added some printing.

  o Toplevel:
    - Fixed back galax-parse to the right code (wrongly committed
      debugging code).

November 16, 2005 - Philippe

  o Optimization
    - Adjusted rewrites to deal with ddo-calls in between
      XPath steps

November 16, 2005 - Philippe

  o Twig Joins
    - Added twig join algorithm in cs_code_twigjoin
    - added some utilities in dm_name_index and 
      xquery_algebra_ast_util for twigjoin support

  o Fixed merging op TupleTreePatterns

November 14, 2005 - Jerome

  o Testing:
    - Small fixes to the text comparison.
    - Slight reorganization of the test harness to support a small
      debugging script.

November 14, 2005 - Jerome

  o F&O:
    - Fixed integer division which was wrongly using idiv.

November 14, 2005 - Mary

  o Static typing 
    - Finished implementing static typing of KindTest
    - Fixed bug in Subtyping_letter -- document types not handled
      correctly. 

November 14, 2005 - Mary

  o Static typing 
    - Approximating interleave (&) types with factored type until
      interleave types supported in Subtyping module.
    - Axis/Node-tests : Still working on kind tests

November 11, 2005 - Jerome

  o Serialization:
     - Fixed sequence normalization to comply with the serialization
       spec.
     - Fixed some dealings with entities again.

  o Testing:
     - Fixed the test harness to properly work with version 0.8.0 of
       the test suite.

  o Type support:
     - Fixed top-level structures for the core schemas (removing
       unnecessary complexity).

  o XPath:
     - Fixed bug in support for the self:: axis, introduced when
       implementing the new DM API with node tests.

November 10, 2005 - Mary

  NB: NUMEROUS changes for static typing & rewriting.  See detailed
  notes in CRAlignmentNotes.txt, checked in separately. 

  o Implemented static typing rules & rewriting rules for all
    expressions except for step expressions, and for all builtin and
    special functions. 

  o Streaming constructors 

    - Aligned dynamic semantics of streaming constructors and typing
      rules with CR spec.  See CRAlignmentNotes.txt.

    - Aligned signature of pi_constructor w/ other nodes. 

  o Constructors 
    - Fixed normalization of computed constructors to align with fixed
      constructors. 

  o General alignment w/ CR working draft 
    cs_code_fn.ml
    namespace_builtin.ml
    namespace_builtin.mli

  o Rewriting rules
  
    - Gutted & rewrote rewriting_rules_notyping &
    rewriting_rules_typing. 

    - Excised all SBDO rewrite rules

  o Static typing

    - Changed Subtyping_letter.subname, which did not handle xs_anyType
    case correctly.  Need to review change with Jerome.

    - Added back type intersection in
      Subtyping_top.intersects_with. Used in simplification of
      typeswitch. 

November 4, 2005 - Mary

  *** NB : Static typing still under construction ***
 
  o Rewriting_rules_typing: 
    - Reinstating all the old typing rules plus new ones for fs:*
      functions. 
  o Align dyn/static semantics of fn:boolean() with CR working draft

November 2, 2005 - Jerome

  o Schema:
    - More simplification of the core schema structure, preparing for
      more changes on the run-time for handling types.

November 2, 2005 - Jerome

  o Cleanup:
    - Removed ast/xpath_treepattern module, not used.
    - Removed unused pattern matching when comparing letters in DFA.

  o Schema and type support:
    - Fixed bug in mapping atomic types names to atomic types. (not
      using proper equality).
    - Changed all names in the core type AST from resolved QNames to
      symbols. Removed the corresponding mapping back from symbols to
      resolved QNames in the run-time.

November 2, 2005 - Mary

  o Errors

    Changed Error.error_with_file_location so that _original_ exception is
    returned & can be matched correctly.

    Working through cs_code_selection to be sure all errors are
    wrapped.

  o Rewriting/Procmod

    - Adding rewriting_rules that depend on typing one at a time

  o More on static typing

October 31, 2005 - Philippe

  o Put twig join related code selection in a separate
    module cs_code_twigjoin.ml(i)

October 31, 2005 - Mary

  o Galaxd (Galax daemon)

    A little daemon for answering distributed queries using "dxq:",
    distributed query protocol in fn:doc(). Implemented by Trevor Jim
    for future use.

    - Added galaxd target to Makefile
    - Added toplevel/galaxd.ml

  o Working on Static Typing
  
    - Align/reimplement static typing rules aligned with FS and reimplemented 
      * fs: special functions
      * Overloaded builtin operators and fn: functions
      * Builtin functions 

    - Cs_code_fn
      * Factored out gads of boiler-plate code.  More work needed to
      stream-line this module
      * Implemented several missing aggregation functions 

October 28, 2005 - Philippe

 o Added SBDO functionality inside TupleTreePattern code selection.

October 25, 2005 - Philippe

 o xquery_algebra_ast.mli
    added sbdo annotation to twig nodes
 o xquery_algebra_ast_util.ml(i)
    some pattern manipulation and checking 
     utilities
 o optimization_judge.ml(i)
    factored out some judgements
 o optimization_rewrite.ml
    major reorganization of the XPath related rewrites
 o processing_context.ml
    changed default sbdo behavior to SBDO_Preserve

October 19, 2005 - Philippe

 o Changed the AOETupleTreePattern operator
   to support all axes

October 16, 2005 - Chris

 o Bug in materialize annotations
   - incorrectly determining safe/unsafe w.r.t materialization


October 16, 2005 - Chris
 o The return of AOELet
   - added to all the usual phases
   - added an inlining rule in phase 6 of the optimizer
   - fixed a bug in the compile context
   - fixes empty iteration bug

 o (untyped) predicates now have the form: 
   some $x in E1 satisfies
    some $y in E2 satisfies
      op:equal(fs:untyped($x,$y),
               fs:untyped($y,$x))

October 16, 2005 - Chris
 o Restoring rules about inling and phase ordering  
   in the optimizer
 o Documenting some code
 o NOT DONE: Need to restore Let in the algebra


October 14, 2005 - Jerome

 o Test suite:
    - Hacked query file names (added '.xq' to work with the new
      catalog).

October 14, 2005 - Mary

 o Native type testing
   - Added processing-instruction & comment to type_lexer, xquery_type_ast,
     normalization in schema_norm
   - Fixed type printing in print_type_core
 o Procmod_compiler
   - Replaced call to exit() with an exception
  
   ** NB: PLEASE -- never call exit() directly! **

 o Website
   All pages refer to galax-*@research.att.com mailing lists

   TODO: Still need to fix viewtarget.cgi which is broken

October 14, 2005 - Philippe

 o Refactoring of the TupleTreePattern datamodel to make
   twig join implementation easier and more efficient.

Ocotber 6, 2005 - Philippe

 o Small fix in rewriting rules to restore the ability
   to generate a streaming query plan (clean_fn_doc rule)

October 5, 2005 - Philippe

 o Added support for TupleTreePatterns (i.e., a logical
   tree pattern operator that consists of single
   child/descendant twigs and that operates over tuples).
 o Added rewrite rules for introducing TupleTreePatterns
   from combination of TupleTreeJoin operators
 o Added default evaluation (matching TupleTreeJoin eval)
   for the TupleTreePattern operator

 NOTE : TupleTreeJoins (and thus TupleTreePatterns) are
	only introduced properly when evaluating
	cleaned-up normalized expressions

September 27 - Philippe

 o Some more tweaking of the rewrite rules + cleanup
   The rules that are now used for XPath cleanup are
   - Cleanup 0 - Remove redundant Concat of MapFromItem 
                 with sequence construction
   - TJ 1 - Introduce TupleTreeJoin for common steps 
   - TJ 2 - Introduce TupleTreeJoins for predicate steps
   - Cleanup 2 - attempt to clean up MapToItem
   - TJ 3 - Remove MapFromItem over MapToItem
   Other rules are not used any more (at least for common 
   XPath expressions).
 o Rule TJ A & TJ B are still used to map TupleTreeJoins
   to bulk TreeJoins

September 21 - Philippe

 o Rule for removing unnecessary fs:node-sequence calls
   in rewrite_rules_notyping
 o Added a rule for removing a MapConcat over a MapFromItem
   in case of sequence construction (map_from_concat_bis)
 o Moved rule for removing unnecessary MapConcats over
   MapFromItem with singleton tuple access to the 
   TupleTreejoin rewrite phase to avoid conflict with
   selection pushing
 o Replaced rule for introducing TupleTreejoins in
   predicate steps to work with the renewed optimizer


September 30, 2005 - Mary

 o Test target mailing addresses for commit

September 18, 2005 - Chris

 o Shredded Store Changes
   - Added new descendant method
     o controlled by a flag
       -shredded-descendant-hack
   - minor code cleanup
 o Optimizer Changes
   - added singleton products to concatenation rewrite
   - added singleton MapToItem inlining. 
 o Small interface changes for testing harness
   - additions only

August 31, 2005 - Jerome

  o Alignment:
    - Removed rule that functions declared in the prolog must always
      have a prefix. This is not the case anymore. (this became only a
      recommendation to developers).

September 11, 2005 - Chris

 o Optimizer changes:
   - optimizer works in mini-phases 
     o rules in a phase are applied to a fix point
   - removed potential race condition (doc'd in code)
   - optimization judgments to prolog
   - added function inlining

 o Algebra Utils:
   - added function to do deep copies on algebra ops

 o Code Cleanup
   - removed unused compile_annotation.ml
   - this had been moved to xquery_algebra_ast_annotation.ml 
   - reference from planio cleaned up

September 9, 2005 - Chris

 o Added ordered determinstic semantic
   - correctness required adding update_ordering
   - parsers takes "snapo" instead of snap determinstic { } 
     because of lexer problems (will be fixed by Jerome?) 
   - modified all the necessary phases

 o Changed annotations 
  - removed options from annotations 
    o fixes some phase ordering problems
  - Added logical_algebra_types.mli for logical ops
  - algebra_type.mli only contains physical op types

 o Annotations (physical_annotations)
   - added has_snap, has_nested_snap to annotations
     o needed for determinstic ordered semantic
     o resides in algebra_type.mli 

 o insert normalization now has explicit copy { Expr }    

 *******************************************************************
 *** There should be a bug in the way cursors are handled in the ***
 *** deterministic ordered semantic, but I haven't found a query *** 
 *** to trigger it. Please send along if you have one.           ***
 *******************************************************************
                                         
August 30, 2005 - Chris

 o Optimization changes
   - judgments for dealing with side-effects are now tied into
     optimization.
     o optimization_judge.ml(i)
   - changes for optimization (these are documented in an optimization
     note) which will be latex'd soon.
   - optimizations semantics should not be effected without side-effects
     though some rules, particular those involving tree-joins, are not 
     fired when side-effects are present. 
     o Someone more in the know about these semantics or formal
       statements of them would be helpful.
  
  o XQuery! Changes 
    - Modified parser/lexer to allow:
      o "snap" "nondeterministc"? { Expr } 
      o "revalidates" are gone because of automaton
         overflow in camllex 
    - Modified places were ast changes were necessary 
      o normalization, factorization, planio, compilation, 
        typing, algebra
    - Implemented conflict detection for insert-insert
      conflicts
    - Modified galax_dm.ml to handle replace_value on attributes
    - printer modified as well 


August 28, 2005 - Chris

 o Shredded_main_memory is up
   - main memory version of shredded representation
   - use shredmm:// - uri in the document function 
     o handles loading
   - new (handcoded) btree. 
     o It has debugging support and internal tests
     o Btree only pushes to the left (to avoid splits)
     o Does not implement merge on delete

 o Code Movement
   - lots of little unnecessary shredded_jungle_* modules removed
   - common modules factored into shredded/shredded_common/ directory
   - Just the basic structures should be required now.

 o Factorization fixes
   - added judgement for side-effect free expressions
   - changed factorization to use these

 o Rewriting fix (can_fail) 
   - CECall and CEOverloadedCall did not
     look at their arguments

 o Added support in the store for rename

August 22, 2005 - Jerome

  o Parsing:
    - Removed obsolete native Galax XML lexer/parser. We only have PXP
      now.

August 17, 2005 - Jerome

  o AST annotations:
    - Removed stream annotations from the core, not used
      anymore. Removed computation of stream annotations, not called
      anymore. This should be done at the algebraic level.

August 19, 2005 - Mary

 o Typing_expr.ml: Added trivial typing rule for CESnap:
     | CESnap ce -> type_cexpr typing_ctxt ce 
 o Fixed bug in typing of atomic constructors

August 17, 2005 - Mary

 o Namespace_symbols.mli
   - Exported 
     val rtype_symbol_of_unit_symbol   : resolved_unit_symbol -> rtype_symbol
     val ranon_symbol_of_unit_symbol   : resolved_unit_symbol -> anon_symbol

August 17, 2005 - Philippe

 o Added module for the tree patterns in the algebra AST

August 17, 2005 - Chris

 o Syntax fix 
   - no need for parens inside snap
   - removed print statements

August 17, 2005 - Chris

 o Bug fix in snap:
   - was not materializing input cursor.

August 16, 2005 - Chris

 o Implementation of Deltas and Snaps
   - *ESequencing -> *ESnap
   - snap {Expr}
     o takes only a sigle expression
     o application of updates at the end of scope.
   - Normalization puts a top-level "snap"
   - update evaluation yield "delta"(s)
   - apply in cs_code_update.ml 
   - *not* all update operations in place
 o *ONLY LIMITED TESTING* 
   - so start writing queries!

August 16, 2005 - Chris

 o Factorization 
   - Factorize_globals disabled
     o This disables join pickup in 
       many cases.     
   - Factorize_iteration rewritten
   - not ready to be turned on by default.

 o Fixes:
   - infinite loop bug on empty left outer joins
   - group annotation bug 
 
 o Rewriting 
   - let_clause inlining off


August 15, 2005 - Jerome

  o Data model clean up:

    - Dm#node_type is now only defined on elements and attributes, and
      returns a type annotation (non optional anymore). This
      simplifies access to type annotations.

August 13, 2005 - Philippe

  o Operators:
    - Added a tuple-distinct operator
    - Added functions to access the document order and document id
    - Added u pruning operator (over tuples)

  o Options:
    - Added a pruning option
    - Added prototype rewrite rule that introduces pruning
      WARNING: This rule is triggered by the -prune option but
      needs additioal work to e sound over predicates

August 12, 2005 - Mary

  o C & Java APIs

    Aligned C and Java APIs with change 20 July 2005 change to data model that
    removed 'item' object as super-type of node and atomicValue. 
    
    C API requires that all Datamodel values be items:
    galapi/c_api/galax_wrap guarantees that all values returned to C
    API are wrapped as items, which enforces strong typing on C side.

August 11, 2005 - Jerome

  o Another push on subtyping:
    - Implemented pre-processing of types which allows proper
      treatment of wildcards, the type name hierarchy and substitution
      groups in subtyping.
    - Reorganized subtyping directory. Split the regular expression
      handling in separate sub-modules.

  o Some clean-up of the indexing support:
     - Fixed the dependency issue between loading, datamodel, dm,
       physicaldm and index. The loading is now split in three parts
       ./dm/dm_load contains signatures (not implementation) for
       loading function that can be used across the
       system. ./physicaldm/dm_load_index contains loading operations
       which also populate name indices, finally
       ./datamodel/galax_load contains actual loading code for the
       main memory data model, and registers itself to dm_load and
       dm_load_index.

August 10, 2005 - Chris

  o shredded-jungle-load:
    - typed/untyped load works
  o Shredded_store:
    - fixed bugs with typed values
    - fixed bug in encoding of values
    - some minor signature cleanup

August 9, 2005 - Jerome

  o Fixed small bug in code selection for name indices.

August 9, 2005 - Jerome and Philippe

  o Hooked up the code for index building together.
      - Added dm/dm_name_index module
  o Fixed bug in compilation which was only creating the Parse
    operator call when streaming is on.

August 8, 2005 - Jerome

  o Indexing:
     - Added support for index syntax, replacing the key/keyref
       stuff. Allows the definition of name indices as well.

August 8, 2005 - Jerome

  o XML updates:
     - Updates are now first class expressions and compositional with
       the rest of the language, following the proposed syntax of
       XQuery!.
     - Added parsing, normalization and compilation support for:
       rename, detach, and copy expressions.
     - Cleaned up AST/Normalization to stop distinguishing expressions
       from updates. Statements are still in place as
       expressions. This simplifies the code greatly.
     - Copy is implemented physically just as export from DM to
       stream.
     - Cleaned up DM API for updates, added call for detach, different
       from delete. Delete is currently implemented as a detach.
     - Serious clean up of the update code in cs_code_update, used
       just above the DM API. Removed all the stuff related to
       'concrete' updates not necessary anymore. Calls to update
       algebraic operators now directly call the corresponding DM API
       methods.

August 5, 2005 - Jerome and Philippe

  o Added support for name indices in the code selection context. This
    can now be used during code selection for building indices during
    loading and during staircase join.

August 4, 2005 - Jerome

  o Not giving up on cleaning up Chris's join code :)
     - Cleaned up outer-join support.
     - Factored out more code in various util modules.
     - Factored out the building of conditions used in both hash and
       sort joins.
     - Factored out code selection of predicates up to the
       code_selection_expr module!!
     - Fixed a couple of bugs introduced during cleanup...

August 4, 2005 - Chris
 
 o Additions/Changes to base/
   - Added Shredded_Error to the base/error.ml
   - Added Register_handlers module, to deal with closing and
     syncing.

 o Changes to Shredded_store_sigs.mli
   - Added Cursor dm wrappers for both directions (prev/next)
   - Added in duplicate flags for hash tables and btrees
   - Added in appropriate accessors for hash
   - Aligned get_set, get_both, get_set_range, get_both_range
       o This convention is now documented in the signature file

 o Shredded Representation
   - Added in detach semantic
   - Fixed insert support (passes XQueryUnit update tests)
   - Cleaned up Store signature to move loading inside
   - Renumbering is in place
   - Code Factored

 o Toplevel
   - galax-run: calls close handlers

 o Added in some comments in the config/Makefiles.* about linking

August 4, 2005 - Jerome

  o A bit more clean up of predicates code, as preparation for
    Philippe's work tomorrow.

August 4, 2005 - Jerome

  o Fixed bug in datatypes/btree bulk-loader.
  o Minor clean up of some of the optimization/code_selection modules
    related to predicates.

August 3, 2005 - Jerome

  o Mode clean up of the main-memory Btree support for Philippe:
     - Made the interface to the cs_btree_array module polymorphic in
       the value stored, this still relies on the fact the that value
       is a pair whose first item is an atomic value.
     - Moved total order for atomic types and values into the
       respective util modules in ./datatypes.
     - Moved Cs_btree_array into Datatypes_btree in ./datatypes, so
       this occurs before streaming.

August 3, 2005 - Jerome and Philippe

  o Join again:

      - More cleanup of the sort join.
      - Broke the huge sort-join code into modular pieces.
      - Added sort-join variant which takes a table/btree index array
        as input instead of the right branch.

August 2, 2005 - Jerome

  o Datamodel: clean-up of the dm:typed-value() accessor.

      - Limited the use of the export_typed_value() node accessor to
        element and attribute nodes only (since this is the only place
        this is used).
      - Fixed dm:typed-value() accessor for attributes, now looking up
        the type annotation properly.
      - Factorized the dm:typed-value() code for comment, text nodes
        and comments in the generic DM.

  TODO: Still need some more work for element nodes, notably we need
  to fix the support for mixed content.

August 1, 2005 - Jerome

  o Moved int32/int64/bigint conversions in Gmisc.
  o Some minor fixes to the testing infrastructure.
  o Fixed type coercions to impose left-to-right evaluation order.

July 29, 2005 - Mary

  o Makefile:
     - Reinstated byte-code targets, all except shredded-jungle-load,
       which must be native code.

  o examples/jungle
     - Hacked the Makefile so that jungle.cmxa and all dependent
       libraries are linked in correctly.

July 29, 2005 - Jerome

  o Join:

    First set of changes to clean up join support (ouch)...

     - Factored out the "BTree" code used in the Sort join in
       ./code_selection/cs_btree_array
     - Factorized Ridlist support in ./code_selection/cs_ridlist

July 29, 2005 - Jerome

  o Testing:

     - Fixed again some whitespace processing issue during 'text'
       comparison for the results.

July 28, 2005 - Mary

 o Makefile:
   - Eliminated byte-code targets, which complicated installation of
     Jungle, which depends exclusively on native-code targets

 o Galapi:
   - Updated examples/caml_api to align w/ new Dm APIs (item is now
     union, not object)
   - Changed Dm_functions to align w/ new Dm APIs
   - Updated examples/jungle/ to align with new Shredded_jungle_dm

 o Lots more on static typing & testing

  *** TODO ***

   1. Write expands_to_union_interpretation_of, which yields union type 
   2. Hacks for supporting xs:integer until named typing is in place:
      Typing_util.ml : added a hack for xs:integer in can_be_promoted_to
      Schema_builtin.ml : added xs:integer to xdt_anyAtomicType_decl
                          added cxtype_integer to cxtype_atomic and cxtype_numeric
			    
   3. In Cs_code, need to review dynamic lookup of overloaded
      functions and value promotion.  Make sure it coincides with
      static rules.

   4. Critical subtyping checks: 
      t <: item() 

July 28, 2005 - Philippe

  o Rewrite rules
    - Added rewriting rules that remove most of the clutter in
      compiled path expressions.
    - Fixed many bugs in the other rewrite rules.

  o Command line options for XPath evaluation
    - Replaced the '-treejoin' option with '-treejoin-log' and
      '-treejoin-phys' to control physical and logical operators
      separately.

  o Changed CEForwardAxis and CEReverseAxis to operate over arbitrary
    variable names instead of only the context item glx:dot.

  o Fixed SBDO bug (bugzilla #36)

July 28, 2005 - Jerome

  o Casting: Made several fixes to casting of numeric values.

     - Fixed precision lost when casting float/double to decimal.
     - Now calls the proper serialization operation when casting
       float/double to string.
     - Make disctinction between -0 and 0 when casting float/double to
       string.
     - Fixed several issues with precision when casting negative float/double
       to string.
     - Fixed bug in extraction of exponent during casting of
       float/double to string.
     - Fixed serialization of NaN, due to wrong NaN comparison.

  o Literals:

     - Fixed handling of character reference and predefined entity
       references in string literals.

  o Datatypes:

     - Now relies on Int32, which gives the correct upper and lower
       bounds for xs:int.

July 25, 2005 - Chris

 o Other Changes
   - Made string pool contents public
     (i.e. exposed the triples in base/pool and namespaces)

 o Shredded Changes

   - Design changes 
     o Nodeids used instead of preorders
       - Internal/External ids now used 
     o Start of renumbering and module for inserts
     o XPath pushed into datamodel
     o Functors cleaned up in the store
     o Multiple documents can be opened in the store

   - New Features
     o Datamodel has nodetest
     o Comments and PIs are now stored
     o Validated documents (typed loading)

   - Performance Changes
     o IDL fixes 
       - copies removed (flags -> enums)
       - fixed interfaces for cursors
         o updated code appropriately

     o EQNames stored (versus triples)
  
  - Fixes in compilation
  - Updates are completely excised in shredded   

July 25, 2005 - Philippe

  o Optimization

    - Selection pushing rewrite destroyed a rule that is supposed to introduce
      the TupleTreeJoin. Simply reordering the rules fixes this.

July 22, 2005 - Jerome

  o Testing:

     - Fixed testing output to compy to the W3C test suite result
       descriptions guidelines, and the corresponding XML schema.
     - Fixed bugs support to report when a bug seems to be fixed!
     - Made text comparisons even more permissive, removing leading
       and trailing whitespace, to account for the variations found in
       some test results.

July 22, 2005 - Mary

  o Xquery core AST & Normalization

    - Changed all occurrences of (csequencetype) to (csequencetype *
      cxtype) so that all normalization of exprs and types is
      contained in Normalization module.

  *** Work-in-progress ***

  o Typing modules

    - All static typing rules in place, but utterly untested

    - Typing_util :
      [FS Section 7.2.6]  data_on 
      [FS Section 8.5.1]  can_be_promoted_to
      [FS 8.6.1 Elements in validation mode]

    - Typing_step : rules for typing step expressions
      [FS Section 8.2] Judgments for step expressions and filtering

  o Schema modules

    Slight reorganization

    Schema_builtin : built-in types only
    Schema_util    : functions for building types
    Schema_judge   : ALL judgments from FS Section 8.1
       Some of these should probably move to a typing_judge module,
       but they're all in one place for easier debugging.

July 21, 2005 - Jerome

  o XML Schema:
      - Fixed long-time missing feature in schema import:

         <xs:attribute name="foo"/> now works
           (= <xs:attribute name="foo" type="xs:anySimpleType"/>)
         <xs:element name="foo"/> now works
           (= <xs:attribute name="foo" type="xs:anyType"/>)

July 20, 2005 - Jerome

  o Datamodel:

      - Semi-serious reorganization of the DM APIs. The items are now
        back to a union type, between atomic values and nodes. The
        node hierarchy is now independant from the atomic values
        hierarchy.

      - Provided a built-in implementation of
        descendant/descendant-or-self/ancestor/ancestor-or-self at the
        generic level. This can be overridden by a specific
        implementation.

      - Changed the data model node API so that each methods takes an
        optional pair of nodetest with a schema (used as context to
        check type tests).

  o General code organization:

      - Fixed numerous dependency issues, and compilation ordering
        issues. As a result:

          Created a new directory ./dm/base which contains some basic
          operations used in several other part of the data model,
          including streams.

          Moved XML parsing back to the ./streaming directory.
          Moved DTD import to the ./schema directory
          Moved atomic type hierarchy to the ./datatypes directory.

 *** TODO: THE SHREDDED DATA MODEL DOES NOT HANDLE NODETESTS PROPERLY!

July 19, 2005 -- Philippe

  o ast/
    xquery_algebra_ast.mli
    xquery_algebra_ast_annotation.ml
    
    - Changed the signature of the TupleTreeJoin to 
      "TupleTreeJoin (inv, outv, axis, nt)"
      The TupleTreeJoin behaves like a MapConcat. The axis/nt 
      is evaluated for every item bound to the field denoted 
      by inv. The result is appended to the tuples in the field
      out_v.
    - Fixed the annotations so that the TupleTreeJoin correctly
      reports that it accesses the tuple field out_v. This fixes
      the problem with the erroneous introduction of the natural
      join over TupleTreeJoins inside predicates.
    
  o ast_printer/print_xquery_algebra.ml 
    - Improved pretty printing for TreeJoin and TupleTreeJoin.
    
  o code_selection/
    cs_code.ml, cs_code.mli, cs_code_selection_expr.ml
    
    - Added default code selection for the TupleTreeJoin. 

  o compile/ 
    compile_expr.ml
    
    - Moved the introduction of the TreeJoin into the algebraic
      rewrite phase. The "-treejoin allow" command line switch now
      triggers a rewrite rule that has the same effect.

  o optimization/
    optimization_rewrite.mli, optimization_top.ml, optimization_top.mli 
    
    - Added rewriting rules that introduce bulk TupleTreeJoin operators
      into compiled XPath expressions.
      
  o rewriting/rewriting_rules_notyping.ml
    
    - Fixed Bugzilla bug number 36 -- related to a buggy sbdo-related
      rewrite rule. The rewrite rule "eliminate_axis_ddo_call" has been 
      diabled.


July 14, 2005 - Jerome

  o Updates:

     - Fixed node identity bug in the main memory update
       implementation. [Introduced when putting postorder back in
       place].

July 14, 2005 - Jerome

  o Subtyping:

     - Fixed one more bug in DFA negation (missing transitions for the
       special final state).

July 13, 2005 - Jerome

  o Testing:

     - Made the test harness configurable. It can now support multiple
       testing units (e.g., W3C tests, XQueryUnit test), assuming each
       test unit follows the same structure as the W3C test suite.

     - How to use:

         (1) download and install the W3C XML Query test suite.
         (2) edit the ./regress/testconfig.xml file appropriately
         (3) Run the tests using make test-suite
       (this runs the command line ./galax-regress regress/testconfig.xml)

     - TODO:
         - perform more tests
         - provide better error messages when something goes wrong
           (e.g., in the case of a wrong configuration)
         - compare error codes properly (the error code is ignored for
           now, until Galax can generate conformant error codes).

July 13, 2005 - Mary

  o Static Typing

    - Started re-implementation of static typing

     Notes:
     - We're missing sub-typing on scalar types. 

     - We don't do constructor-mode preserve ; only strip

July 12, 2005 - Jerome

  o Testing:
     - Very first version of test harness is operational. It is
       located in ./regress.

     NOTES:

       -- You still have to set up parameters in the script itself
     (see the Config section). Goal is to replace this by the
     ./regress/testconfig.xml file.

       -- The file ./regress/knownbugs.xml is used to list bugs in
     bugzilla and the corresponding test cases which trigger the
     corresponding bug.

       -- The test harness requires a copy of the W3C test-suite. (See
     internal W3C XML query home page).

       -- You can execute the tests by running 'make test-suite' in
     main galax directory.


July 12, 2005 - Jerome

  o Static typing:
     - Added hooks to parse and normalization content models, useful
       to test subtyping in isolation.
     - Fixed bug in type simplification: was removing non-none types
       from unions instead of removing none types.
     - Fixed pretty-printing of NFA's and DFA's.

  o Subtyping:

     - FIXED OLDEST GALAX BUG EVER:

       Negation for deterministic automatas was missing some final
       states (when the original automata does not have a transition
       for a letter, which is a short-cut for a sinc state). This was
       breaking subtyping for examples like (a*,b) <: (a+,b).

       According to the changelog, the last time this code was touched
       was July 2001. :-( or :-) ?

     - Moved negation outside of the DFA module since it needs to
       create a new state, and that is done on integer states only for
       now.

  TODO:

     - Add support for named typing. Notably, there is no support for
       subtyping of atomic types.

July 11, 2005 - Jerome

  o Constructors, alignment and bug fixes:
     - Now element node constructors properly extract the content of
       document nodes.
     - Now document node constructors properly extract the content of
       document nodes, reject attribute nodes, and perform the right
       kind of erasure.

  o Streaming: Small reorganization of the streaming modules.
     - Created a streaming_constructor module which contains
       operations for constructing new nodes. Move corresponding code
       from Stream_ops and Code_util.
     - Renamed Sax_types into Streaming_types, and consolidated
       utilities in Streaming_util.

  o Parsing:
     - Bug fix: Comments and PIs were not returned by the parser!!!

  o Base:
     - Added netstring Netbuffers as a new kind of Galax I/O.
     - Fixed Gmisc.load_file to use Netbuffer, much more efficient.

  o Testing:
     - First step toward a simple test harness for W3C test
       suite. Still work in progress.

July 7, 2005 - Jerome and Philippe

  - Compilation:

      o Fixed bug in TreeJoin generation when context item is a global
        variable (e.g., external variable).
      o Fixed the default evaluation code for TreeJoin to align with
        the previous evaluation code for forward and reverse axis.
      o Fixed bug in step evaluation when the context item is an
        atomic value, now is raising the proper type error.
      o Removed algebraic operators for Let, ForwardStep, ReverseStep,
        removed corresponding printing, code selection, rewritings,
        and actual evaluation code.
      o Removed hack that Michael introduced to merge sbdo calls
        inside treeJoins. ISSUE: is that preventing streaming in
        certain cases?

July 6, 2005 - Philippe

  - Compilation:

      o Changed compilation of step expressions in compile_cexpr.ml to
        use TreeJoin operator instead of AOEForwardAxis and
        AOEReverseAxis. This gets rid of the AOELet-hack.
  
  TODO: remove AOELet related material and verify that there
        is actually always a #dot field in the tuples on which
	the TreeJoin operates.

July 6, 2005 - Philippe & Jerome

  - Streaming:

      o Removed call to Michael's stream analysis code in
        factorization.

July 6, 2005 - Philippe & Jerome

  - Streaming:

      o Changed compilation of parsing calls to be consistent at the
        logical level by introducing the logical operator AOEParse.
      o Moved decision about streaming vs. materializing parsing to
        the code selection level. 
        WARNING: the latter introdices a bug. When '-treejoin allow' is
        used, the parsing always happens in a streaming fashion, which
        generated wrong results when streaming is not applicable.
      o TODO: remove Michael's code from factorization


July 6, 2005 - Chris

 o Jungle Changes:

   - By default Jungle uses the 4.3 Version now
     o There are instructions on how to use 4.1 as well 
       in tools/Jungle/Notes

   - Jungle no longer deals over strings (char arrays instead)
     o allows idl to be compiled again
   - All get operations return None when not found instead of 
     throwing exceptions - except for "_unsafe" variants.
   - Cursor operations for "getall"
   - Changed record layout

 o Shredded

   - Visible Changes
     o new uri jungleshred://  
     o toplevel: shredded-jungle-store to shred xml documents
     o new directories: shredded, shredded/shredded_jungle, 
	shredded/shredded_main_memory/

   - Internals 
     o shredded_dm.ml: A functor that given a store provides a datamodel
     o shredded_store.ml : Given a collection of (typed) indexes, 
                           provides a store.
     o shredded_load.ml: A functor that given a store, provides loading code.
     o There are registration functors as well

   - Implementation
     o shredded/shredded_jungle, communicate with the new jungle store
     o New children index (for getall)

July 6, 2005 - Philippe

   - Fixed slash-slash rewrting rule to rewrite occurences of steps
     as follows:

      o descendant-or-self::node()/child::nt -> descendant::nt
      o descendant-or-self::node()/descendant::nt -> descendant::nt
      o descendant-or-self::node()/descendant-or-self::nt -> descendant-or-self::nt
      o descendant-or-self::node()/self::nt -> descendant-or-self::nt

July 5, 2005 - Jerome

  o Processing model:

     - Added a pre-processing phase for static analysis before each
       processing phase. The new architecture looks like the
       following:

        Pre-normalization 	  -> Normalization
        (Nothing)
        Pre-rewriting     	  -> Rewriting
        (Typing + SBDO)
        Pre-factorization 	  -> Factorization
        (Freevars)
        Pre-compilation   	  -> Compilation
        (Nothing)
        Pre-optimization  	  -> Optimization
        (Nothing)
        Pre-codeselection 	  -> Codeselection
        (Nothing)

     - Moved the processing model data structures in
       ./procmod_types.mli

     - Factorized many of the processing model code to avoid code
       duplication and inconsistencies in the way each phase is
       processed.

     - Fixes many inconsistencies in the various interfaces uses for
       the top-level operations used in the processing model. Notably:

         -- now parsing uses the parsing context properly instead of
            the processing context.
         -- code selection now returns the ASTs as well as the
            compilation context.


July 1, 2005 - Mary

  o More More More Testing CVS

June 29, 2005 - Jerome

  o DTD support:

     - Added back support for DTDs. Hooked up PXP DTD parser
       and implemented a mapping to the XQuery type system.
     - Reenabled the -dtd flag in galax-parse and galax-mapschema.

  o Streaming:

     - Added simple diff operation over streams. We should be able to
       use this to compare testing results.

June 29, 2005 - Philippe

  o Reverted to an atomic execution of TreeJoin. Got rid of Michael's
    messy code selection code that aggregated multiple TreeJoins into
    one algorithm.

  o Removed the function that was labelling the top-level stream as
    entry into the tree join, which was breaking compositionality.
     [Question: WHY WAS THIS NEEDED??]


*******************************
June 29, 2005 - Mary and Jerome
*******************************

Move to new CVS server: cvs-graphviz.research.att.com

*******************************

June 20, 2005 - Jerome and Philippe

  o Code organization:

     - Moved ./streaming/dynamic_buffer to ./base/dynamic_buffer since
       it is a generic module for buffer management.

  o Streaming:

     - Fixed bug in composing streaming XPath operators yielding
       labeled stream with stream-based constructors. Now doing the
       buffering properly.

  This fixes the following issues in XPath streaming:

    (6) Fix constructors to deal with labeled streams (and buffer the
        fragments of the streams which are labeled).

  Remains to be done:

    (1) recompiling and testing simple cases with Michael's existing
        code  (BENCHMARKING/TESTING)
    (3) fix the text node bug reported by Michael (DONE)
    (4) revert to an atomic evaluation of consecutive Tree joins (no
        need to coalesce consecutive Tree joins)
    (5) look at how to expand Michael's work for more complex cases
        such as:
    (7) Fix the treejoin option to be independant of factorization.
    (8) Align optimization options for the command-line to be more
        consistent.
	   (e.g., -treejoin [stream/staircase/nl]
	          -join [hash/nl]).

June 15, 2005 - Jerome and Philippe

  o Datamodel:

     - Fixed the Tree data model implementation to use only pre/post.
     - Fixed data model loading to deal with stream labels, fixing the
       nodeid/copying bug.
     - Fixed data model loading to deal properly with pre/post.

  o Streaming:

     - Added ordered streams back, to support post order (which is an
       additional information only on closing events, while preorder
       is on opening events.
     - Propagated that change back in the rest of the code.
     - Removed node id information from the annotation module.
     - Fixed serialization to deal with labeled streams (and buffer
       the fragments of the streams which are labeled).

  This fixes the following issues in XPath streaming:

    (0) Change load_xml_node_sequence_from_typed_ordered_stream to
        examine labels and yield in result those nodes whose
        stream-label is set.
    (2) fix the streaming/loading bug related to copying

  Remains to be done:

    (1) recompiling and testing simple cases with Michael's existing
        code
    (3) fix the text node bug reported by Michael
    (4) revert to an atomic evaluation of consecutive Tree joins (no
        need to coalesce consecutive Tree joins)
    (5) look at how to expand Michael's work for more complex cases
        such as:
    (6) Fix constructors to deal with labeled streams (and buffer the
        fragments of the streams which are labeled).
    (7) Fix the treejoin option to be independant of factorization.
    (8) Align optimization options for the command-line to be more
        consistent.
	   (e.g., -treejoin [stream/staircase/nl]
	          -join [hash/nl]).

June 13, 2005 - Jerome

  o Streaming:

    - Fixed Jungle to work with the new streaming interface.
    - Fixed Streaming XPath code to support other kind tests (the
      document-node(a) variant is still not supported since it
      requires advanced lookup in the stream which I am not sure how
      to deal with).
    - Factorized node tests code between the streaming XPath
      evaluation and the DM evaluation. Common part of the code is
      located in the new directory: ./code_util.

  o Alignment:

    - Implemented all the missing kind tests:
       element(), element(a), schema-element(a), etc.
    - Factorized the code between type matching and kind tests.

  o Bug fixes:

    - Fixed code selection bug for key declarations. Using the safe
      variable allocation mode instead of the 'unsafe' one, which
      doesn't work currently for prolog declarations.
    - Fixed detection of duplicate attribute creation in element
      constructors. Aligned support for detecting duplicate attributes
      in queries and documents.

June 6, 2005 - Mary

  o Schema_import.mlp -- More cleanup and addition of file locations. 

June 5, 2005 - Mary

  o Schema_import.mlp -- Erroneoulsy made changes to schema_import.ml
    instead of .mlp.

June 5, 2005 - Mary

  o Sax_types

    Added file locations to SAX/RSAX/TSAX events.  A SAX event (like
    all other AST objects) now contains two parts : a descriptor and a
    file location.

  o Finfo 
   
    File locations produced by Pxp parser are (entity,line,column)
    triples.  Added new file-location kind for (entity,line,column)

  o Schema import and Schema validation

    Now report file locations on errors! 

    NB: Error handling in Schema_import is in need of restructuring.
    I could use Vladimir's help on this.

June 3, 2005 - Mary

  o Overhaul of errors and error reporting 
    --------------------------------------

    Summary:
    - No errors, except lexing or parsing errors, should contain file
      location info.
    - Top-level functions that recursively walk over ASTs catch
      internal errors and re-wrap them with file location
    - Avoid tramping of file locations arbitrarily throughout code and 
      eliminates most uses of Finfo.bogus

    General principle for reporting errors:

    File locations are associated with top-level expressions, which
    are represented, in turn, by AST nodes, Core AST nodes, and
    algebraic ops, each of which stores the top-level expression's
    location.

    To correlate an error with the location of a top-level expression,
    each phase that recursively walks over a AST/Core AST/Algebra tree
    (e.g., normalize_expr, compile_expr, rewrite_expr,
    cs_code_selection) "wraps" each error generated within a recursive
    call with the file location associated with the current AST/Core
    AST/Algebra node that is being processed.  For example, the
    general structure of the top-level recursive call is:

      let fi = node.file_loc
      try 
        match node.desc with
        | ... dispatch on each kind of AST/Core AST/Algebra node ...
      with 
      | exn -> raise(Error.error_with_file_location fi exn)

    Where 
    
      let error_with_file_location fi exn = 
        match exn with
        (* Error already wrapped -- reraise it *)
        | Query(Wrapped_Error(fi, msg)) -> Query(Wrapped_Error(fi, msg)) 
        (* Otherwise, wrap it with the location of the current
           AST/Core/Algebra node *)
	| _ -> (Query(Wrapped_Error(fi, bprintf_error "" exn)))

    This technique avoids the "tramping" of file locations throughout
    arbitrary levels of code, which may or may not raise errors.

May 30, 2005 - Mary
  o Errors

    - Started clean-up of error handling.

    - Partitioned errors into groups by processing-model phases.

      *** NB *** Every phase now has a default error whose name is the
      phase's name, e.g., "Compilation", "Factorization" etc.
      These names should be used instead of "Internal_Error", so that
      we can more easily track the source of errors. 

May 26, 2005 - Mary

  Integration of Streaming XPath Evaluation 
  -----------------------------------------

  o In Sax_types, added new typed_annotated_sax_event (similar to the
    annotated Core AST nodes).  An annotated typed SAX event is a record that
    contains a typed SAX event and an annotation field. 

    In new Sax_annot, there are two kinds of annotations: one for
    document-order labels and one for streaming-xpath labels. 

    Ordered sax events are implemented as typed_annotated_sax_events
    in which the document-order annotation is set to a global document order.

    Labeled sax events are implemented as typed_annotated_sax_events
    in which the stream-label annotation is set.

    NB: We could generalize the annotations to include the node-types
    as well, which would coalesce resolved streams with typed
    streams. 

  TODO:
  ----
    (0) Change load_xml_node_sequence_from_typed_ordered_stream to
        examine labels and yield in result those nodes whose
        stream-label is set.

    (1) recompiling and testing simple cases with Michael's existing
        code

    (2) fix the streaming/loading bug related to copying (See 6)

    (3) fix the text node bug reported by Michael
    (4) revert to an atomic evaluation of consecutive Tree joins (no
        need to coalesce consecutive Tree joins)
    (5) look at how to expand Michael's work for more complex cases
        such as:

        declare variable $d := doc(); $d//a//b
        or
        let $d := doc() return $d//a//b

    (6) Serialization -- move buffering code here. 

May 6, 2005 - Jerome

  o Bug fixes:

    - Fixed code selection bug for key declarations. Using the safe
      variable allocation mode instead of the 'unsafe' one, which
      doesn't work currently for prolog declarations.

April 28, 2005 - Michael

  o Streaming XPath evaluation
    - Added new source code folder streaming_xpath. It contains both code for
      evaluating XPath expressions in a streamed fashion and the additional
      data flow analysis necessary to decide where streaming is applicable.
    - Factorize_top
      * Added stream analysis as a (last) factorization phase for XQuery
	modules; adds XQuery core ast annotations
    - Xquery_core_ast_annotation
      * Added new field indicating wether an operator can be evaluated in a
	streamed fashion or not
    - Compile_expr
      * Added compilation of certain FLWOR patterns into logical tree joins.
      * Added compilation of certain fn:doc calls into dedicated parse stream
	operators.
    - Cs_code
      * Provided streamed / materialized physical variants for logical tree join
    - Cs_code_selection
      * Selects streamed / materialized tree join code
    - Top_*, Processing_context
      * Added new fields that specify wether tree joins and streamed XPath
	evaluation is considered during compilation
      * Added new fields that specify buffering behaviour for streaming XPath
      * Added support for printing a data flow graph in Graphviz dot format
    - Galax-run
      * Added serialization directly from a stream (without intermediate
	materialization) whenever tree joins are alllowed - this is a HACK!

April 26, 2005 - Philippe
  o Rewriting
    - rewriting_ddo_annot.ml
      Aligned the SBDO removal code with the replacement of CELet and CEFor
      by FLWOR. Removed a few bugs.

April 25, 2005 - Philippe
  o Rewriting
    - Restored remove_fs_node_sequence rewriting rule

April 14, 2005 - Mary
  o Schema validation 
    - Fixed bug in local element name lookup 

April 14, 2005 - Jerome

  o Processing model alignment:
    - Renamed 'cleaning' phase to 'rewriting' phase, aligning with
      papers et al.

April 13, 2005 - Mary

  o XML Schema : Added support for <import> 
    - Parse_xquery, Norm_*_top
      * Added support for Schema/Module/Service location hints 
    - Schema_util
      * Added resolve_location_hint for Schema/Module/Service location hints 
    - Print_type 
      * Fixed bug when printing nested schemas 

  o Parse_io/Parse_xml
    - Removed processing_context from interface to
      pull_parser_from_input_spec/open_xml_stream_from_io

  o Fn_error: Exposed implementation-defined Galax error type, which
    contains:

    (qname : item, description : item option, error-objects: item list)

  o Documentation
    - Various updates
    TODO: Fix API Doc

April 13, 2005 - Jerome

 o Bug fixes:
    - Added missing lexers.
    - Fixed boundary-space calls in the Caml,C,Java APIs.

April 12, 2005 - Jerome

 o Toplevel:
    - Removed unused namespaces serialization parameter in Top_config.

 o XQuery prolog:

    - Changed the prolog grammar one more time to separate
      setters+other declarations according to the April 4th spec.
    - Added support for default order and copy-namespaces
      declarations.

 o Alignment:

    - Added parsing support for extension expressions with
      pragmas. Extension expressions are always ignored, i.e., they
      return the value in the expression following XQuery's normal
      semantics.

April 4, 2005 - Jerome

 o F&O:
    - Fixed bug in fn:doc() function opening the same file multiple
      times (Bug reported by Emiran).

April 4, 2005 - Jerome

 o Added a -aggressive SBDO removal operation.

 o XQuery core:
    - Removed CEFor and CELet from the XQuery Core AST, now all
      replaced by actual FLWOR expressions.

 o XQuery algebra:
    - Removed Map over items (AOEFor) which isn't used anymore. Kept
      the AOELet binding which is used to bind the context item,
      still.
    - Removed AOEEnclosed, which is a no-op. Compilation of CEEnclosed
      now is just removed from the plan.
    - Renamed SepSequence to MapIndexStep, aligning with the VLDB
      paper.

 o Module reorganization:
    - Moved ./code_selection/predicate to
            ./optimization/optimization_predicates.
    - Removed unused ./rewriting/annotate_walker.mli file.
    - Moved ./planio/algebra_printer to its natural location as
     ./ast_printer/print_xquery_algebra.
    - Moved ./compile/compilation_annotation to its natural location
      as ./ast/xquery_algebra_ast_annotation.
    - Now compiling the data model and streaming support before the
      ast and XQuery compilation. 
    - Moved the XML SAX parser into the ./parsing directory.
    - Renamed ./parsing/xquery_parser.mly to
              ./parsing/parse_xquery.mly


March 31, 2005 - Mary

  o Command-line interface
    - Changed command-line interface as follows: 
      -var <varname>=<untypedAtomic value> : binds global variable
       $varname to untypedAtomic value
      -doc <varname>=<filename> : binds global variable $varname to
       document in specified file

  o Import schema/module
    - Target of AT hint is now URI -- previously a local file name. 

March 28, 2005 - Mary

  o Cursor.ml
    - Fixed bug in rev_list_of_cursor (reported by Philippe)

March 28, 2005 - Philippe

  o Added remove_fs_node_sequence rewrite rule to the
    set of standard rewrite rules in cleaning_rules_notyping

March 18, 2005 - Philippe

  o Corrected the rule remove_fs_node_sequence 
    in cleaning_rules_notyping to work with the
    new normalization
 
March 18, 2005 - Mary

  o Union, intersect, except --
    Fixed bug in code_selection/cs_code_fn : operators must sort by
    document order.

    *** See following remarks (also in code) ***

    We should probably change the normalization of op:{union,
    intersect, except} so that the SBDO operator is explicit (right
    now, it is implicit in the implementation).  Then if we know the
    inputs are already in document order, we can use faster
    sorted-merge implementations for each operator and also avoid the
    explicit SBDO.  These rewritings should be added to the rewriting
    rules.

March 18, 2005 - Philippe

  o Added two rewriting rules in cleaning_rules_notyping.ml
    (1) A rule that removes all unnecessary distinct-docorder ops
        that get introduced by normalizing predicate steps
    (2) A rewrite rule that replaces normalized occurences of
          expr1//axis::nt
	where no predicate occurs and where the axis is a 
	forward axis, different from "attribute" into
	the normalized equivalen of
	  expr1/descendant::nt

March 18, 2005 - Philippe

  o Small fix for the sbdo optimization
    NOTE: the state of declared variables in the prolog is
    not correctly propagated to the SBDO optimization of 
    the expressions (yet)

March 17, 2005 - Philippe

  o Changed annot_ddo.ml to support the new normalization rules
    - cleaning/annot_ddo.ml
    - normalization/norm_expr.ml
    - ast/xquery_core_ast_annotation.ml(i)
  o Extended rewrite rule elide_node_sequence_ddo_hack
    to obtain cleaner expressions
    - cleaning/cleaning_rules_notyping.ml
  o Added fs:docorder-or-atomic-sequence
      and fs:distinct-or-atomic-sequence
    - namespace/namespace_builtin.ml(i)
    

March 14, 2005 - Mary

  o Changed normalization of Path expressions 
    normalization/norm_expr
    - Changed normalization of path expressions into FLWORs
    - Added fs:distinct-docorder-or-atomic-sequence to check for
    homogeneous sequences of nodes or atomic sequences

  o Added rewriting rules for for/let Clauses
    cleaning/cleaning_rules_notyping
    - The rewrite rules for individual for/let clauses are not as
      clean/pretty as those for for/let expressions. 

  o Added fs:distinct-docorder-or-atomic-sequence and 
          fs:node-sequence-or-atomic-sequence 

  ** NB : SBDO rewritings have not been modified yet ** 

March 11, 2005 - Philippe

 o Cleaning:

    - Another bug fix for the duptidy automaton. On unsupported syntax
      constructs, the automaton shifted into the sink state, which was
      wrong because in the duptidy automaton, this implies that nodup
      holds, which is not necessarily true (for instance for CESeq's,
      this will cause trouble).  Fixed by introducing a dummy state
      that has the not-nodup property.

March 11, 2005 - Philippe

 o Cleaning:

    - Small bug fix, related to the use of FLWOR expressions in XPath
      expressions that caused the result to be out of document order

February 23, 2005 - Jerome

 o Compilation:
    - Fixed bug in the tools makefiles when compiling with Jungle.
    - Fixed bug in make depend in the galapi Caml Makefile (how come
      we never picked it up?)

 o XML Updates:
    - Fixed the parent pointer bug.
    - Implemented insert in document nodes.
    - Move code wrappers from Cs_code to Cs_code_update

February 21, 2005 - Chris
 
 o Factorization 
   - Fixed a problem with factorization of path expressions.
     o Conservative approach to pulling out FLWORs.
       Eventually we should relax this but requires a bit of work

 o Optimization
   - Added check to deal with termination if rules commute
     o NOT GENERAL, PLEASE, IF YOU ADD PUSHDOWN RULES READ THE 
       COMMENTS IN OPTIMIZATION REWRITE.
       - A cost model is one way to solve this problem.

  o Top Level/Conf:
     - Added -output/print-logical-rewrites flags
       o uses explicit flushes in the formatters.

  o Changed link order in Makefile so algebra printer
    is accessible from optimization_rewrite
      
February 16, 2005 - Jerome

 o Web site:
    - Small fixes to main Web site page.

 o Materialization:
    - Removed hidden calls to list_of_cursor from Physical_util. All
      calls should now occur in appropriate evaluation code
      operations, plus a few data model operations.
    - Fixed -print-materialize flag to indicate the correct size of
      the materializat collection, and the origin of the call.

February 15, 2005 - Mary

 o Web site & API tests

***** Galax Release 0.5.0 February 15, 2005 *****

February 15, 2005 - Jerome

 o Factorization:

    - TURNED IT BACK OFF!!!! :-(

February 15, 2005 - Mary

 o Documentation for 0.5.0

 o API Tests
   - Removed base-uri() applied to document node.  Currently
     undefined.

 o typed_value/string_value/fn:data in Dm and Datamodel
   - Aligned with January WD

 o Testing F&O 
   Cs_code_fn, Norm_expr:
   - Aligned fn:substring w/ F&O 

 o Usecases
   - Removed old -unsafe-join-hack option

February 15, 2005 - Jerome

 o Monitoring:

    - Added a printing operation for cursor->list materialization. 

 o XML parsing:
    - Fixed printing of PXP error messages.
    - Added error for XML document of a version different from 1.0.

 o XML updates:
    - Fixed parsing bug in complex updates introduced by the last fix!

February 14, 2005 - Jerome

 o Compiler:

    - Made sort/hash joins the default algorithm.
    - Replaced -unsafe-join-hack flag by a -nested-loop-join flag
      which turns all joins into nested loops.
    - Put factorization back on by default.

 o XML parsing:

    - Made PXP's pull parser the official XML parser. The XQuery
      parser is still based on the previous lexers, since we have to
      have more control over how parsing is done.
    - Removed all XML parser.

 o DTD support:

    - Removed DTD support for now. We need to re-implement the DTD to
      XQuery type system mapping to work over PXP's DTD data
      structure.

 o XML updates:

    - Grammar change: all updates start now with the update
      keyword. This fixes a serious parsing bug where simple updates
      where confused with a function call.

       insert --> update insert
       delete --> update delete
       replace --> update replace

    - Fixed bug in "insert after" when the update location is the last
      sibling.

 o Documentation:

    - Some fixes to the contributor's list, release notes, and
      documentation.

February 11, 2005 - Jerome

 o XML parsing:

    - Removed attribute content expansion in the PXP parser. Was
      triggering a bug in one parsing test.

February 11, 2005 - Jerome

 o MinGW port:

    - Fixed several makefile so that it compiles with PXP's lexers.

February 10, 2005 - Chris

 o Factorization fixes
   o added test to make sure factored expressions are 
     side-effect/failure free
   o This requires static context - this change is 
     propagated throughout
   o Factorization no longer pulls terms out of  
     branches (If or TypeSwitch)

February 10, 2005 - Jerome

 o XML parsing:

    - Fixed bug in PXP hook for DTDs. Now works properly.
    - Added ./usecases/docs/company.dtd so the company-data.xml is
      well-formed.
    - Removed hack to deal with missing start and end document events,
      made unnecessary by the above bug fix.
    - Fixed bug in PXP hook for parsing of processing instructions.
    - Fixed bug in PXP hook not setting up the correct internal
      encoding.

 o Updates:

    - Added back materialization in compilation of FLWOR expressions.

 o Static typing:

    - Fixed parsing bug in ./typing/typing_fn.ml.
    - Added a place-holder for static typing which indicates that this
      feature is disabled in that version.
    - Put appropriate error messages for incomplete patterns due to
      the on-going implementation.

 o Toplevel:

    - Aligned options for galax-parse with the other command-line
      tools.

 o Tests:

    - Emptied obsolete ./usecases/expect directory.
    - Turned makefile echoing off during testing.
    - Fixed makefiles for usecases and regress.

February 10, 2005 - Mary

 o Java API
    - Added compilation options in JAVAOPTS to Makefile.unix

 o Installation
    - Removed byte-code executables from Binary distribution 
      They just bloat the size.

February 9, 2005 - Mary

 o Testing
    - Usecases and API tests now use same error messages
    - Removed Makefile.release


February 8, 2005 - Jerome

 o XML parsing:

    - Added prototype hook to PXP's pull event parser. Note: This
      seems *a lot* faster than the hand-made Galax parser and should
      be much more compliant to the XML 1.0 specification.

       NOTE: This feature can be turned on with [-use-pxp on] on
       galax-parse and galax-run.
       NOTE: There is no support for DTD's at this point.

February 8, 2005 - Mary and Jerome

 o Renamed 'regress' target to 'regression'

 o APIs
   - Fixed API tests and examples to use imported library modules
     correctly. 

 o Normalization & Rewriting 
   - Fixed bug in which rewriting was not applied to prolog
   - Fixed bug in normalization in which order of variable
     declarations was reversed in var_simplify applied by List.fold_left 

February 8, 2005 - Mary and Jerome

 o Doc & Query Parsing

   Fixed nasty bug in Sax_parser and in Parse_top: must set encoding
   after getting a handle on the lexing buffer.  

   PXP interface requires that input encoding be set explicitly after
   reading the first 2 bytes of the document or after reading the XML
   declaration; The encoding is necessary to populate the lex buffer
   more than one character at a time.  The auto-detected encoding is
   chosen if "" is passed as the encoding.

February 8, 2005 - Mary and Jerome

 o Code selection
   Fixed bug in HashJoin and SortJoin (same as OMapConcat below)
   - Fixed bug that we introduced on Feb 4.  When outer join occurs
     and result is empty, we now return table with one empty tuple
     instead of table with no tuples.
   
February 8, 2005 - Chris

 o Fixed bug in OMapConcat
   - on empty input, it should return a place holder tuple.
     This is the "outer" portion
   

February 7, 2005 - Mary
 o API tests
   - Added rules for checking API tests -- these rules disappeared
     from Makefile.gen!

February 7, 2005 - Michael

 o Projection:

    - Path_analysis
      Changed XQuery core ast pattern matching for document uri
      extraction. Pattern matching now assumes being performed after core
      level rewritings.

    - Stream_project
      Fixed treatment of a (non-empty) set of projection paths, where each
      individual path is empty and has the subtree flag turned on.

      Introduced testing the underlying input stream for being exhausted in
      order not to query an empty environment stack for projection paths.

    - Projection_context
      Added val project_stream_is_empty to the interface in order to implement
      the behaviour mentioned above.

    - Galax-project
      Changed to make use of the stream_project interface as expected, namely
      to treat main modules as a unit.

      Projected documents can now be output to files, instead of just printing
      the projection paths.

February 7, 2005 - Mary
 o Solaris Port
   - Some changes to support Solaris -- still working on this

February 7, 2005 - Jerome

 o Factorization:

    - TURNED OFF BY DEFAULT UNTIL IT'S FULLY DEBUGGED.

 o Pretty-printing:

    - Recovered the proper behavior for printing statements
      independantly of the XQuery prolog when compiling a whole
      module.


February 7, 2005 - Jerome

 o Optimization:

    - It's back!
    - Added optimization of statements from within the module
      optimization call.

     ** NOTE: Why don't we do any optimization in the XQuery prolog?
     ** NOTE: Parsing has become impossibly slow. This might be
        related to the new netstring/PXP versions we are using and
        must be checked.
     ** NOTE: the last relational query triggers another bug in
        factorization!!

 o Base:

    - Added a Debug module. Minimalistic, but we should start use it
      consistently for debugging messages.

February 4, 2005 - Jerome

 o Core AST:
    - Removed obsolete usage information in let expressions. (follow
      up of removing it from the algebra).

February 4, 2005 - Jerome, Mary and Michael

 o Physical values:
    - Pulled support for cursors and SAX token streams up to the Galax
      API. (Caml only in separate functions).
    - Uses physical values in the processing model now, instead of XML
      values.

 o Algebra and code selection:
    - Removed obsolete usage information in let expressions (subsumed
      by the annotation in the Algebra AST).


February 4, 2005 - Jerome, Mary and Michael

 o Cleaned up physical values:

    - Removed internal support for "functional" tuples, which are now
      implemented as 'unit'. Renamed tuple into tuple_unit, clarifying
      they do not contain any value.
    - Removed all operations on lists of tuples (or materialized
      tables). Only cursors of tuple_unit are now in use.
    - Removed many obsolete pieces of code in:
       -- execution_context: removed current tuple

       -- coercion functions: removed all operations on tables and
          tuple lists.
    - Changed the output type of Tuple construction to return a single
      tuple instead of a table.
    - Removed PT_table and PT_tuple_list from physical_type in the
      Algebra AST.
    - Removed unused Evaluation_update module.

February 4, 2005 - Jerome

 o Context:
    - Turned preserving boundary whitespace, PIs and comments on by
      default.

 o Compilation:
    - Fixed a problem in compilation of Forward and Reverse steps when
      'fs:dot' is bound explicitely within a FLWOR expression.

 o Core AST:
    - Added operation to set an annotation for a core expression.
    - Split core expression construction between fmkacexpr which
      takes an annotation as input, used notably during rewritings,
      and fmkcexpr which always builds and empty annotation and is used
      consistently during normalization.

 o Normalization:

    - Removed all uses of empty_annot() during construction of the
      core AST. Uses fmkcexpr instead and sets the annotation through
      the proper operation instead.
    - Serious clean up of Norm_util. Notably isolating core AST
      creation operations and using them systematically.
    - Added new normalization 'ident' which takes the AST and
      generates an identical XQuery core for it. This can be triggered
      using the new [-executed-normalized-expr on] flag.
       ** Note: This still requires some debugging **.

January 30, 2005 - Nicola

 o Wsdl stub generation:
    - aligned default namespace management with the last changes in 
      Galax
 o Apache module:
    - rewritten for the current Galax API, not yet fully tested

January 28, 2005 - Jerome

 o XQuery pretty-printing:
    - Fixed pretty-printing of schema declaration to use a semicolon
      at the end("declare schema {};").
    - Fixed bug in 'cast as', reversed type and expression.
    - Fixed pretty-printing of QNames, using '*' not '*:*'.

 o XQuery Core pretty-printing:
    - Fixed pretty-printing of global variables declarations.
    - Fixed bug in 'cast as', reversed type and expression.
    - Fixed bug in 'treat as', reversed type and expression.
    - Fixed pretty-printing of atomic type to be in the built-in xs:
      namespace instead of the xsd: namespace.
    - Fixed pretty-printing to print a newline after each function
      declarations.

 o Core bootstrapping fixes:
    - Added the overloaded op: function signatures back into
      pervasives.xq so that they can be called directly from the
      query.
    - Pretty-printing axis steps using $fs:dot/Step instead of Step.

 o Normalization:
    - Split normalization into separate modules:
        Norm_expr for expressions
        Norm_update for updates
        Norm_top for prolog and top-level function
    - Moved predicate_kind support in Norm_util.

January 27, 2005 - Mary

 o Web demo
    - Aligned with new Netcgi library 
      NOTE: Netcgi module requires well-formed HTML! 

January 24, 2005 - Jerome

 o AST:
    - Cleaned up the XQuery and Core ASTs for function declarations:
      duplicate output types are removed, duplicate function names in
      external function removed, argument names added to external
      functions, common parts pulled out from the body.

 o XQuery parsing:
    - Fixed bug in element constructor syntax, now allows an empty
      content as in 'element a {}'.

 o XQuery pretty-printing:
    - Fixed pretty-printing of global variables declarations.
    - Fixed pretty-printing of external function declarations.
    - Fixed pretty-printing of treat as (reversed arguments).
    - Fixed pretty-printing to use the correct precedence (January
      2005's grammar).
    - Added bootstrapping tests for the usecases (pretty prints then
      executes and compares with the expected result). This is now
      automatically run when doing 'make' in the usecases directory.

 o XQuery Core pretty-printing:
    - Complete overhaul of the pretty-printer for the XQuery core.
    - Aligned pretty-printing on the XQuery one.
    - Fixed pretty-printing of function calls to be a proper XQuery
      function-call (i.e., removed type assertions).
    - Added support for precedence.

January 23, 2005 - Jerome

*** Chris: please check the last fix. ***

 o Started cleaning up and aligning the options for the command-line
   tools (only galax-run and galax-compile for now).
 o Changed the xqt: namespace to the err: namespace, following the
   latest XQuery specs.
 o Fixed parsing bug in attribute content, missing the escaped quotes
   ('' and "").
 o Fixed serialization of character data not to escape '%',
   unnecessary, except if serializing DTDs, which we never do in
   Galax. NOTE: This fix should be made part of PXP. (See:
   ./base/encoding.ml).
 o Moved compilation of the Galax Caml API before the toplevel
   directory.
 o Fixed a serious code selection bug related to scoping of global
   variables in function declarations, tripped when a function
   parameters had the same name as a global variable.

    Test should return 89:

      declare variable $high := 1;
      declare variable $high := $high+10;
      declare variable $low   := 100;

      declare function local:f($low,$high) { $high - $low };

      local:f($high,$low)

January 19, 2005 - Mary

  o galax/Makefile

    - Changed reference to pr_o.cmo to $(OCAMLHOME)/camlp4/pr_o.cmo
      because caml4po is broken on MacOS.

January 19, 2005 - Chris

 o Fixed bug in Factorize_unique

    - result of renaming was not used.

January 13, 2005 - Chris

  o Factorization: 
    - fixed bug in factorize_globals causing function
      evaluation (Bugs #7 and #8)
    - added factorize_unique.ml(i)
      o These are currently unnecessary but,
        many parts of the code can be simplified
        by using this mini-phase.

January 11, 2005 - Jerome

 o Documentation:

    - Fixed PXP reference to 1.1.95 required for compilation.

January 11, 2005 - Jerome

 o Grammar alignment with January 2005 editor's working draft.

    - Cleaned up some old grammar code: aligned non-terminal with
      XQuery's document grammar, removed obsolete 'qnamekeyword'
      non-terminal.
    - 'where' clause now takes an expr_single.
    - Fixed precedence. (many bugs fixed there)
    - Replaced xmlspace declaration by boundary-space declaration.
    - Separated the prolog declarations into two sets, following the
      new grammar.
    - Moved the default namespace declarations from the setters to the
      import/namespace declarations.

January 6, 2005 - Jerome and Mary

*** WARNING: New packaging approach. Beware that this new version now
requires you to install third-party packages (pcre + ocamlnet + pxp)
before to work with Galax source. ***

 o Complete new compilation approach for Galax:

    - removed 'tools' pcre, netstring and pxp. The new version is
      based on, and requires prior installation of the latest version
      for those packages directly from their maintainer.
    - Cleaned up the makefiles to avoid duplication of variables,
      notably, the location for PCRE, NETSTRING, PXP is only in
      ./config/Makefile.gen

 o URL support:

    - New version of netstring (ocamlnet) and PXP fixes serious
      problems with URL support, notably on windows.
    - Now uses recent support from PXP and netstring, which should
      provide more robust support for URLs in general.

** NOTE: someone should check the F&O functions for URIs **

 o MinGW:

    - Fixed compilation chain to work on MinGW. This requires serious
      hacking beforehand to obtain a version of
      Caml+finlib+pcre+netstring+pxp.

 o MacOS:

    - Fixed compilation and linking on MacOS X.

January 2, 2005 - Chris

 o Changes to how fs:untyped-to-any style semantics are handled
   in sort and hash joins.
   o In particular, correct promotions for doubles
   o Bug in bounds
   o Added some comments to make conventions in the joins more clear.

 o NOTE: the sgml usecase "passes" but gives an EOF Error. This was
   the case before this checking - there is a line of whitespace in
   the output but, not in the .expect.

December 7, 2004 - Mary

  o Constructors
    - Added computed processing-instruction constructors 
  o Makefiles/Port
    - Updated several Makefiles to support Solaris (from Ben Yao)
  o F&O 
    - Aligned fn_error w/ Oct 2004 draft; Four alternatives.
      We need to decide what our implementation-specific error format
      should be. 

December 7, 2004 - Amelie
  o Updated Projection 
    - Path_analysis seems ok (both regular and optim)
    - Function analysis needs to be done on a per-function basis for builtin functions
    - To Do: Prolog analysis

December 3, 2004 - Mary

  o Alignment with Oct 2004 WDs:

    13. The div operator can now be used to divide two
        yearMonthDurations or two dayTimeDurations. In either case,
        the result is of type xs:decimal.

    15. The to operator now accepts an empty sequence as an operand,
        and returns empty sequence.

December 2, 2004 - Mary

  o Added -output-err to galax-parse tool

December 2, 2004 - Mary

  Regression tests:

  o Added -output-err to redirect error output to clean-up testing
    - Replaced all occurrences of stderr by
      Conf.glx_stderr/Conf.glx_err_formatter
    - Do NOT use stderr directly. 
  o Cleaned up regress/Makefile
    - Aligned with new galax-regress package (replaces xqueryunit)
    - Added API tests to standard regression tests
 
November 24, 2004 - Chris

 o Type Assertions

   - Groups and Tuple Creations now have type assertions.
   - Serialized in XML Query Plans.
   - Printer change to reflect this.

 o Usecases

   - changed the expected namespace to "http://www.galaxquery.org"
     in the .expect files.

November 22, 2004 - Mary

  o Alignment with 0.4.0 ... again
    Pervasive.xq:
    - New function: glx:stem

November 20, 2004 - Chris

  o Bug fixes:

    - Fixed scoping bug in factorization that was causing
      both reported sequence and what seemed to be a function 
      call bug.
      o Seperated structures that were passed and returned into
	different types within factorization
      o Seperated out common code.
	   *** Are there any issues with namespaces here when 
	   *** we pull up variables? ***

  o Cursor Changes
     - commented out old array and list fold code in the cursors
     - put in new code with tiny amount of state that should be 
       faster and cleaner. 

  o Changed docs/xmark.xml
    - changed back to Factor 0.1 from 0.25
    - xmark_usecase.expect still expects this
    
November 15, 2004 - Jerome

  o Added glx:deep-distinct() function to remove duplicate
    based on deep equality. Used to run Clio queries.

November 10, 2004 - Mary

  o Monitor
    - All phases are now printed, even if value is 0.0
    - Increased precision of elapsed time to microseconds (6 decimal digits)
    - Changed memory monitor to do full_major collection (recover
      unreachable blocks) instead of minor collection. 
      NB: *** full_major is expensive! ***

November 9, 2004 - Mary

  o Monitor/Galax API

    FINALLY, found bug in monitoring of eval_statments*.
    Materialization of sequence was not wrapped in monitor code:

    let v' = wrap_monitor proc_ctxt
     Monitoring_context.Query_Evaluation_Phase list_of_sequence v in

    Monitor output partitioned by : prolog, statements, and
    serialization.

  o Caml API
    - Changed serialization to be monitored as "Serialization"

November 8, 2004 - Mary

  o Caml API

    - Changed serialization to be monitored as "Statement" -- lazy
      evaluation means that serialization may actually be evaluating
      query.

November 8, 2004 - Mary

  (NB: It appears that -monitor-time on gives more accurate results
       than does -monitor on, which includes memory usage)

  o Galax-run

    - Changed toplevel galax-run so that all calls go through Galax
      API (and are therefore monitored).  Removed all direct calls to
      Procmod_compiler.

  o C API

    - Added external calls for monitoring eval_statement functions
    - Added galax_procctxt_from_{compiled,prepared}_prolog

  o Caml API

    - Moved compile_serialized_{logical,optimized}_main_module into
      Galax API so that they are monitored correctly 
    - Fixed bugs in which list_of_sequence was not wrapped in
      monitor. 
 
  o Monitor/Monitoring_context

    - Changed to support nested calls to start/end monitor
    - Added stack for resumed, monitored calls

November 3, 2004 - Mary

  o Normalization

    - Changed normalization of treat-as and validate expressions to
      propagate annotations from child expressions.  XPath annotations
      were not propagated to these expressions.

November 2, 2004 - Mary
  o Printing options for command-line
    - Fixed command-line options so that -pretty-print-module + other
      printing options prints complete module.

  o Rewrite/Cleaning modules
    - Added ability to apply rewrite rules to prolog 
    - Prolog rules are selected in Procmod_phases

  o Static_context/SBDO 
    - Static_context: (Temporary) addition of var_maxone_property
      table to static_context so that maxOne properties of global
      variables are accessible during rewriting.

      We have to use the static context, because currently, the
      rewrite context is not returned between API calls.

    - Annot_ddo
      If a global variable has the maxOne property, new prolog rule
      adds variable to var_maxone_property of static context. 


October 29, 2004 - Chris
	
  o Code factorization

    - requires free variable annotation, which is added to all the
      usual places.
    - Iteration Factorization rule placed in factorize_iterate.ml
       o Factors expressions under for loops
    - Global Variables factorization module (factorize_globals.ml)
    
	********************************************************
	PLEASE CHECK THAT THE UPDATE CEXPRS ARE WALKED CORRECTLY
                               FOR UPDATES
	     THERE ARE NOTES IN THE NEW FILES IN BOTH WALKERS
	********************************************************	

  o New Algebraic Operation: AOESepSequence
    - allows us to separate sequences in a non abusive way. Certain
      rewrites with map index were not possible, because it has a
      stronger numbering semantic. Namely it numbers the absolute
      position of items in a stream. We don't want that all the time.

  o Group By
    - fixed to allow early termination of some recurring loops

  o Optimizer changes
    - utils and rewrites corrected and updated to deal with this new
      operation, and the corresponding push-down rewrites
    - added rules (and explanations) of how to deal with combinations
      of outer maps and outer joins

  o Updated PlanIO parser
    - still outputs oddly because new namespace problem

  o Bug Fixes
    - Fixed bug regarding normalization of the where clause of 
      flower expressions
    - Fixed bug about pushing down on item based fields
    - Fixed bug having to do with safe and unsafe variable
      settings for cursors
 
October 28, 2004 - Mary
  o Toplevel/galax-run
    o Changed calls to monitor to be more descriptive: 
      Galax-run.compile_standard_library_module
      Galax-run.compile_main_module
      Galax-run.execute_compiled_prolog
      Galax-run.execute_statement

October 27, 2004 - Mary
  o Monitor 

    Added functions for monitoring external functions :
    start_monitor_external_call & end_monitor_external_call.
    Usage: 
    start_monitor_external_call (processing_context, "descriptive string")
    ... program text ...
    end_monitor_external_call (processing_context)

    Useful for getting monitor output that covers both application and
    Galax time. 

  o C API 
    Added galax_start_monitor_call & galax_end_monitoral_call to C
    API. 

October 26, 2004 - Avinash
  + Added an optimization rule for updates.
     - Now the query plan has a Materialize (by default) for
       every return clause.
     - The Materialize is removed for all the queries except
       of updates (to implement the snapshot semantics of updates).
     - -update-hack-bi on flag would remove the Materialize from
       the query plan all together (even for updates). 
     - The ordering of the optimization rule matters now, hence
       this particular rule is placed before the rule, which removes
       the MapToItem from the query plan in case of nested FLWR 
       expression.

October 25, 2004 - Mary
  o Command-line
     - Fixed misspelling of -optimization
     - Added switches to usecases/ & regress/ Makefiles/scripts

October 25, 2004 - Jerome

  o Processing model and optimization flags.

     - Put all optimization experimental flags in ./base/conf (see the end).
     - Added flags to turn on/off optimization and factorization to
       the standard processing context. Removed corresponding
       temporary hacks from Chris's code.
     - Turned loop fusion off by default.

October 25, 2004 - Jerome

  o Processing model:

     - Added a new factorization processing phase, after rewriting,
       and before compilation. It used to be a pre-step within
       compilation, but needs to be expanded to support factorization
       of updates (to support the snapshot semantics), and to support
       further code factorization for optimization purposes.

       Result of the factorization phase can be printed using the
       usual flags:

  	 -print-factorized-expr  Print optimized expression
  	 -output-factorized-expr  Output optimized expression to file.

  o Factorization:

     - Added a new directory ./factorization
     - Moved old ./compile/normal_form* into ./factorization
     - Cleaned up a bit in there.
     - Added a sub-factorization module to deal with FLWOR updates in
       order to prepare the compilation in a way that facilitates the
       application of the snapshot semantics.

  o Updates: Moved updates as core expressions. This cleans up a lot
    of existing code and simplifies maintenance. This also is a step
    closer toward first-class updates in XQuery.

     - Moved atomic updates to be part of the XQuery core. Added
       "sequencing", i.e., two sub-expressions evaluated one after the
       other, as part of the core. This is similar to the ';' in Caml,
       and used in XML updates only.
     - Removed previous updates AST from the core, statements are now
       like expressions.
     - Propagated the changes to the rest of the compiler: pretty
       printing, normalization, cleaning, ddo analysis, compilation.

October 25, 2004 - Roel, Philippe

  o Implemented vertical loop fusion and rule for removing
    node-sequence calls (when applied to axes).

  o Added new parameter loop-fusion (default = on) and changed default value
    for sbdo parameter to duptidy.

October 25, 2004 - Philippe

  o Even more bugfixes in the DDO code. As it appears, order by
    clauses were skipped by the algo and (worse) predicate steps did
    not get handled off in a proper manner, due to the passing of a
    wrong boolean value

  o Added the loop_fusion rule set which pushes node-sequence calls
    surrounding normalized step expressions into the CEFor return
    statements.  After that, the loop_fusion rule itself (which will
    be added later today) will apply the vertical loop fusion
    technique in order to obtain pipelined execution of path
    expressions where possible. This should point out the real benefit
    of the sbdo optimization.

October 22, 2004 - Philippe

  o Fixing small bugs in the annot-ddo module.

October 22, 2004 - Philippe

  o Fixing small bugs in the annot-ddo module.

October 21, 2004 - Roel

  o Fixed bug in the final state that is returned by the DDO
    automata. This final state is used for the rest of the execution
    of a query when a path expression is bound to a variable and
    stored in the DDO symboltable.

October 20, 2004 - Roel

  o Implemented the sloppy and duptidy DDO automata. The duptidy
    automaton is expected to be the best and most useful, since it
    leaves only intermediate DDO operations in place when duplicates
    can occur in that step.

October 19, 2004 - Mary

  o API, again
    - Fixed bug in which processing_context arguments to most API
      functions was simply ignored (!)
    - Processing context no longer passed to all API functions.
      Create processing_context once, modify as needed, then _only_
      pass to load_standard_library.

October 15, 2004 - Jerome and Avinash

  o XML update compiler, bug fixes:
 
     - Fixed bug for delete in algebra pretty printer.
     - Fixed lexing bug for atomic updates within FLWOR updates. Was
       not putting the lexer back into default state.
     - Fixed compilation of FLWOR updates completely wrong (was not
       aligned with the new FLWOR compiler for expressions).
     - Fixed materialization code bug. The MaterializeTable operation
       in the algebra was using the old tuple list materialization
       instead of the new materialization code.
     - Fixed annotation code for MaterializeTable.

NOTE: This still requires a lot more testing.
NOTE: We will have to factorize the FLWOR update code to avoid
duplication.

October 11, 2004 - Philippe

  o Perarations for the implementation of other automata
    i.e., detecting the beginning/end of a path expression
    when processing the AST

October 13, 2004 - Chris

 o Two bug fixes
   o Fixed MapIndex to start at 1 instead of 0
   o Fixed Group operator to correctly return empty on empty groups

October 11, 2004 - Mary

  o Alignment of Caml, C, and Java APIs with compiler's new processing
     model
    - Removed "quick start" functions as they do not align well with 
      processing model.
    - See Galax modules in 
      galapi/caml_api/galax.mli, galapi/c_api/galax.h,
      galapi/java_api/Galax.c for new interfaces

  o Updated all API Examples to align with new API
    - Added config/Makefile.api 
    - In galax/, run "make tests" to run all usecases and API
      tests. 

  NB: **** config/Makefile.unix has changed ****
   
October 8, 2004 - Jerome and Mary

  o Static typing:
 
     - Started putting back static typing in place, expression per
       expressions. Function calls are back, but not built-in
       functions. Big missing pieces are path expressions and
       constructors.

October 8, 2004 - Roel

  Fixed bug in DDO rewrite rule: in stead of removing
  distinct-docorder we now replace it by node-sequence to guarantee
  type correctness

October 7, 2004 - Philippe

  Repairing small bug in automaton implementation

October 6, 2004 - Philippe

  Cleaning up debug output in the SBDO optimization

October 6, 2004 - Jerome and Mary

  o Bug fixes:

     - Fixed wrong bug fix for external variable declarations.
     - Fixed real bug in support for the external context item.
     - Fixed lexing bug on lexing of "<a" tokens in the operator
       state.

  o Processing model and API:

     - Forced passing of the external context at the time of
       compilation for a prolog or main module. The external values
       are put in place only at execution time still, but that removes
       some unnecessary additional call to set the context later on
       and enables proper support for an external context item.
     - Added a default external context.
     - Propagated the changes to the Caml, C and Java APIs.

  o Code cleanup:

     - Did a serious renaming for typing module to make it more easy
       to organize.
     - Remove dead code in ./algebra.

October 6, 2004 - Philippe

  * The DDO rewrite is now completely in place.
    The only thing left to be done is testing it.
    Files affected
      o cleaning/annot_ddo.ml
      o cleaning/ddo_symboltable.ml
      o cleaning/ddo_symboltable.mli

October 5, Mary

  o Schema_util, Schema_builtin

    - Changed element/attribute/type declarations to use a Hash table
      instead of lists.
    - Changed merge_cxschema to check for duplicate declarations.
    - Moved builtin schema attributes from Schema_util to Schema_builtin
  
October 4, Mary

  o Branch Alignment

    Updated Dev branch with changes made to 0.4.0 after release. 
    No more changes to 0.4.0!

    August 2, 2004 - Mary
    - Changed examples/Makefile* to only run API examples if the API libraries
      are built and installed 
    August 3, 2004 - Mary
    - Fix to config/Makefile_jungle.def
    August 5, 2004 - Mary
    - C and Java APIs : 
      o Changed set_xquery_whitespace to set_xmlspace_kind
      o TODO: add construction mode and ordering mode to processing contexts.
    - More on rationalization of Makefiles
      o !!!config/Makefile.unix has changed!!!  
        Copy to config/Makefile and personalize!
    - New directory: examples/tutorial for documentation
    September 1, 2004 - Yitzhak
    o C API
      Removed GALAXCOMPILE conditional compilation from galapi/c_api/itemlist.h
    o Node Ids
      Added 64-bit pre-order integer node id to support PADS nodes

October 5, 2004 - Jerome

  * Re-implemented subtyping for the new type system. This still does
    not deal with substitution groups and nillable flags.

October 5, 2004 - Philippe
  * Added recursive calls to non XPath ast nodes in the ddo
    annotation module. 
    TODO: o adding the rewrite code which removes/replaces ddo calls
          based on the ddo annotations (which are now in place)
	  o generate tests from the automaton code to test all
	  possible paths in the automata
    The ddo annotation code is now much more robust and elegant.
    On top of that the state of variables is recorded in a
    kind of symbol table, which enables us to remove even more
    DDO calls

October 5, 2004 - Philippe
  * Fundamental changes to the ddo cleaning algorithm. The algorithm
    is now node based and does not rely on pattern matching any more.
    Not all AST types are supported yet, but simple XPath expressions
    are correctly annotated
    
    Files affected:
      ast/xquery_core_ast_annotation.ml(i) 
      cleaning/annot_ddo.ml(i) 
      cleaning/ddo_automaton.ml(i) 
    
    Files added: cleaning/ddo_symboltable.ml(i)
 
October 3, Mary
  o Item type atomic() replaced by built-in atomic type
    xdt:anyAtomicType
    - Declaration added to schema_builtin

October 1, Mary

  o AST
    - Replaced polymorphic 'a cexpr type by monomorphic annotated
      acexpr type everywhere in code: effects Normalization, Algebra,
      Cleaning, Code selection, 

  o Split Rewriting_walker into Rewriting_walker and Annotation_walker
    - Rewriting_walker rewrites AST in purely functional bottom-up
      walk of AST.
    - Annotation_walker annotates AST by imperatively updating each
      node's annotation in bottom-up walk of AST. 

      See Typing_notyping module for example use of Annotation_walker

  o Removed analysis phase entirely -- no longer necessary 
    - Effects primarily Procmod_phases and Procmod_compiler.


September 30, 2004 - Avinash

  * Added new namespace support into Jungle.
	- 2 new indices to store nsenv and namespace bindings.
	- 2 new id's, (namespaceid and uri_prefix_mapid)
	- modified main record to store index of nsenv for element record.
	- For creating nsenv from stored information we use a per store 
	  cache of nsenv.

 September 29, 2004 - Philippe

  o cleaning/annot_ddo
    - added basic framework for ddo annotating the ast
  o cleaning/ddo_automaton
    - new implementation for automata
  o cleaning/cleaning_rules_notyping
    - added the ddo-optimization to the non-type-based rules for now,
      since the type based rules do not compile anymore
  o procmod/procmod_phases
    - added the new rewrite rule to the sbdo phase
  o procctctx/processing_context
    - removed the requirement that strong typing should be enabled for
      the ddo optim to work
  o Makefile
    - uncommented the building of ddo_automaton and annot_ddo in 'cleaning'
    
September 28, 2004 - Philippe
  o xquery_core_ast_annotation.ml/.mli
    - added the sbdo_optim_annot type
  o print_top.ml
    - support for printing the AST XPath SBDO annotations

September 28, 2004 - Philippe

  XPath AST annotations are now in place
  o xquery_core_ast_annotation.ml/.mli
    - added the xpath_sbdo_annot type
    - added a xpath_sbdo_annot member to the ast annotation type
    - added accessors and mutators for the new member
  o norm_top.ml
    - replaced the empty annotations in xpath normalization rules
      slash, axis and step_qualifiers with the correct xpath_sbdo
      annotations
  o norm_util.ml
    - modified the normalization of function calls to produce correct
      annotations for fn:doc(), fn:document() and fn:root() calls
  o print_top.ml
    - support for printing the AST XPath annotations
 
September 26, 2004 - Mary

    o Added config/Makefile_stemmer.conf to bootstrap Makefile

September 24, 2004 - Mary & Emiran
    o tools/stemmer
      - Added Porter's English Stemmer Algorithm 
      - New function: glx:stem
      - Added config/Makefile_stemmer.def ; stemmer is required tool

September 24, 2004 - Jerome

  o Added Namespace_names.string_of_uri function for Avinash.

September 24, 2004 - Chris
 o Code Moves
   - Algebra directory is removed
   - Code having to do with the code selection portion of the algebra
     moved to the code_selection directory
     o There are new prefixes cs_code and cs_util to eventually split 
       these as well but, this requires a larger redesign
   - planio directory created to handle serialization of XML Query plans

 o Introduced predicates and predicate_branches
   - compile Select statements resulting from and, or 
     in where clauses in to trees with conjuncts at the leaves.
     This allows us to split predicates and push down portions:
     $x = 5 and $x = $y, we can now push down the $x = 5 condition.
   - Switched all joins to use predicates for evaluation
     o TODO: Add simple heuristic about order to evaluate 
   - Optimizer changed to push down predicates
     o Notable exception: Does not combine predicates in joins
   - Introduced evaluation code for predicates 
   - Predicate branches are a technical condition of when you
     can make joins (i.e. when you can seperate into two independent
     branches). These are documented in the code, there are a host
     of these types of conditions.
   - prediates work by rid list operations. This should change a bit.
  
 o Join Changes
   - updated all joins as above to work with multiple predicates.
   - NOTE: 
       ********************************      
       *** THESE JOINS ARE NOT SAFE *** 
       ********************************
     o They may not handle all the implicit type promotions properly
       and will need to be revisted when static typing comes back.
   - cs_util_join_helper now 'compiles' predicates into join_information
     structures.     

September 23, 2004 - Mary
  o Core AST Annotations
    - Added Xquery_core_ast_annotation module.  All core expressions are
      now annotated. 
    - Removed tcexpr, tcstatement, ..., tcmodule
    - Added acexpr, acstatement, ..., acmodule
    - Effects every module that previously operated on ucexprs and
      tcexprs
  o Procmod phases
    - Normalization is now optional phase.  Disable using -normalize
      off  (Necessary for testing parser stand-alone)
  o Bug fixes
    - Decimal.ml: Fixed float to decimal conversion for floats w/
      negative exponents
    - Schema_util.ml: Fixed definition of is_none_type

September 23, 2004 - Nicola
  o Wsdl:
    Fixed the type annotations for the encoded case of an RPC Soap Body.

September 19, 2004 - Mary

  o Datatypes_util:
    Moved bt_can_be_promoted_to here; was missing xdt:*Duration types

September 19, 2004 - Mary

  o Monitor:
    - Fixed monitor so that calls to wrap_monitor not wrapped in
      start/end_monitor_calls work 
    - Monitor times are emitted as xdt:dayTimeDurations

  o Function & Operators Alignment:
    - Added aggregates over xdt:dayTimeDurations and xdt:yearMonthDurations
    - Changed op:node-equal to op:is-same-node
    TODO: fn:error is still not aligned

September 16 2004 - Chris

 Bug Fixes:
  o distinct-values was not promoting types according to the
    F&O semantic for comparison. This is fixed in both places
    it occurs (distinct-values and the groups)

  o Hash_join: 
    o  The same equality semantic problem also occured in groups
       and their hash functions. The promotion for fs:untyped-to-any
       is not handled by trying to cast to both potential outcomes. 

 Algebra Changes:
   o Changed AOEMapItemTuple to AOEMapFromItem
   o Changed AOEMapTupleItem to AOEMapToItem
   o Changed AOEGroup to AOEGroupBy
   o Updated printer and optimizer
   Grouping Operator:
    o Added field (induced_groups) to group_desc
      o deals with projection/scoping of tuple fields
    o Removed distinct-values

 Code tweaks:
  o distinct-values now performs over cursors and uses
    hash functions instead of O(n^2) list compares.
  o ocaml 3.08's hash fn seems to be multilinear on the words it hashes.
    - seems to be identity on ints & enumerated (Union) types.
    - string_hash function added to base
  o AtomicValueHash: to hash on atomicValues added to dm.ml(i)

 Hash join: 
  o correctly work with fs:untyped-to-any semantic
    should now be safe to pickup.   
  o Materialization changed to correctly support this
    semantic.
  o No longer assumes homogenous sequences.

 Compilation:
  o Tuple names are now compiled to be unique. 
  o FLWOR Normal form before compilation begins
    o Element construction and function application
      pulled out

 Optimization Rules:
  o Updated rules to work with new semantics
    of grouping 
  o Updated group creation rules

September 14 2004 - Chris

 Bug Fixes:
  o Fixed bug dealing with external variables in code selection.
  o Fixed procmod_compiler to properly handle context items and
    external variables
    - Variables need to be added, and then have code selection 
      done. We are skipping intermediate phases.
    - examples seems to work but, there are problems with
      namespace serialization

 Procmod_compiler:
  o Fixed above mentioned bugs
  o Added functions to get_*_context to return copies 
    of the precompiled functions 
    - needed for contexts with mutable contents (ex: hash tables)
 
 Code tweaks:
  o Changed can_be_promoted helper function for efficiency in
    algebra_code_util.ml
    - caml calls polymorphic equality function on (=) 
      between Union types. 'match' is faster.

  o Cleaned up Sort Join code to be used on >,>=,<=,<,=
    - improved check to pick it join.
    - still only picks up one pattern (double nested Some)      
    - should only pick it up if it is safe now but, 
      the flag remains.
      *******************************
      ** Hash join is still unsafe **
      *******************************

    - extended to work on all atomic values


  Algebra Pretty Printer:
   o Another modification to bring the printer inline with the  
     notes.
     - added [()] notation for parameters in the algebra

  Algebraic Operations:
   o New: AOENullMap 
     - Used to determine Nullity of sequences in groups
   o Changed: AOELeftOuterJoin, AOEOuterMapConcat
     - Now use new method of setting Nullity of their
       inner relations
   o Changed: AOEGroup
     - Distinct-Values support added in aggregation
     - empty behavior changed 
 
     ***************************************
     *** Possible bug in Distinct-Values ***
     *** This follows the same semantic  ***
     *** As already Noted on Galax-dev   ***
     ***************************************  
 
  Optimizer Changes: 
   o Updated rule set to work with the new operations
   o Small simplification in the rule set
   o Added Rule to pick up multiply nested joins 
  
  Algebraic Plan Parser:
   o Updated to work with new operations and changes

  Normal Form For FLWORs:
   o Compiler now puts expressions in a normal form 
     documented in normal_form.ml
   o This allows simplification of optimizer rules.
   o added_files: normal_form.ml(i)
   o flags added to toplevel/galax-run.ml and 
     toplevel/galax-compile.ml to toggle this option
     (default: on)
     

September 11, 2004 - Nicola

  o WSDL

    - added the Wsdl_maker module which builds a wsdl AST from an
      XQuery AST. For the time being, it is untested.
    - added Wsdl_printer as an empty stub for a future WSDL pretty
      printer

September 10, 2004 - Mary

  o Function & Operators Alignment
    TODO
    - fn:error is not aligned
   
  o Galapi

    - Updates to align API tests with new prolog syntax
    - Changed Quick start calls to eliminate nested monitor calls

September 9 2004 - Chris

  Galax-run:
  o New Flags
    - algebraic-optimization: 
        ******************************************************
        *** Allows algebraic optimization to be turned off ***
        *** since this code changes rapidly. (default: on) ***
        ******************************************************
	
    - print-comp-annotations:
        Toggles the printing of compile annotations
        in algebraic pretty-printed plans (default: off)

  o Fixed bug dealing with multiple input files
    - Module being compiled several times
    - Need a fresh compile context each iteration

 Pretty Printer:
  - Updated to use same notation as notes. 
    o Op{Dependent}(Independent}

  - Changed breaking to make more readable
  - Modified to work with new/changed operations

 Plan Printer/Parser:
  - Updated to work with new/changed operations
    ** Still an issue with namespaces in the output ***  

 New Algebraic Operation: OuterMapConcat
 - Operates like MapConcat.
   o If an input tuple returns a sequence, no effect
   o If an input tuple is empty, return a sequence
     with input tuple fields having their value
     and dependent returned fields = ()
   o ** NOTE ** Is not 100% safe yet because
     of tuple scoping issues. Requires all
     tuples names be unique -> flatten scopes
     -> allow it to correctly set the return fields
     see note in code for more details

  Group Operator Semantic Change:
  - Removed value support
  - Extended to multiple grouping sets
  - Added "Exclusions" for optimization rule
    o These allow for correct handling of valid rule
  - Split out group code into group_code.ml(i)
  - Many improvements can be made; some are noted in the code.

  Optimizer Rules:
  - Map through group rule to introduce outer joins and maps
  - Group combination rules and flattening
  - Rules dealing with MapIndex
  - Old product rewrite is safe again
  - Changed Pushdown-selection to be more aggressive
  - Rule dealing with Maps through products

August 27, 2004 - Chris

 Added Groups, LeftOuterJoin, Hash Joins and Sort Joins
 - Optimization Rules changed
        o Product rewrite has been simplified
          Now called natural_product_rewrite
        o Old Product rewrite may still be useful
          but currently disabled.
        
 - Optimization Rules added
	o group_apply_flatten_rule 
	 (flattens the apply clause of Group Bys)

        o map_singleton_join_or_product_rule
          replaces Map(Join(InputTuple,..)(E_1) 
          Join(E_1,..)

        o First step to unravelling multiply nested
          groupings

        o Rule to push maps of joins, through groups
          allows denesting of queries.

 - Code clean-up
	o Removed annotate_algop since it was misleading
	  (no annotation is actually done in this call)
        o Code factorization in cursor_peek 
          - it saves 20% on peek's cost
            by pulling out the constructor

 - New algebraic operation AOEGroup
 	o Currently only supports single groups
        o Optimization Rewrite Rules to write into groups
        o Sort Based grouping only
        
 - New Operation AOELeftOuterJoin
        o Optimization Rewrite rules to recognize 
        o Code to implement it
        o Modifications to pick up the code correctly
        

 - New physical joins
        o Disabled by default, flag below.
        o *** THEY ARE UNSAFE ***
        o sort_join.ml(i) and hash_join.ml(i)
        o Currently sort_join only picks up op:gt 
          (in xmark) but, will be shortly extended
          to other comparison functions.
        o Do pass usecases but still unsafe        
        *** USE AT YOUR OWN RISK ***

 - Flag in galax-run.ml : -unsafe-join-hack
        o When set allows the pickup of string based
          hash joins. It is unsafe for many reasons,
	  which are documented in algebra/hash_join.ml
	  and algebra/sort_join.ml. 

  - XML Query Plan Parser and Printer
        o Updated to work with new operations and
          namespace support
        *** The serialization is still not 100% correct ***
        **** NOT WELL TESTED - PROBABLY BUGGY ***            

August 27, 2004 - Jerome

Namespaces:

 - Added a Namespace_util module which contains specific support for
   hashtables and association lists for resolved QNames. Since
   resolved QNames now contain their namespace prefix, Caml equality
   should *not* be used for resolved QNames.

Alignement with XQuery /July 23rd draft:

 - Fixed semantics of computed element and attribute constructors, now
   doing atomization on the element name and supporting
   xdt:untypedAtomic. Also fixed namespace resolution for the element
   and attribute name.

Bug fixes:

 - Fixed the C API to compile properly with the new namespace support.
 - Fixed Jungle to compile properly with the new namespace
   support. However, this will still not work as we must find a
   different way to store and recall namespace environments for nodes.
 - Fixed the fn:doc() function to cache the input document so that
   multiple calls to the same URI does not recomputes the document.

External variables:

 - Now allow also input atomic values directly from the command line
   for external variables.

August 26 2004 - Chris
 
o Fixed bug in compilation to AOEMapIndex 
	- used wrong arugment in construction

August 25, 2004 - Jerome

Namespaces overhaul. Phase 2.
Preparation for new static typing.

** WARNING **

 - Static typing is excised entirely in that version, and all modules
   depending on it.
 - Significant merge had to be done with WSDL support. Please check
   new namespace support.
 - There is still some bugs in handling of rqnames due to the triple
   proposal. I am doing more testing now.

** WARNING **

  - Reorganized namespace modules:
     o Namespace_names definitions and basic operations on names.
     o Namespace_builtin replaces Funcs_and_ops, now contains also
       built-in names previously in Namespaces, Datatypes_util, and
       Schema_builtin.
     o Namespace_context replaces Namespace, with support for
       namespace environments.
     o Namespace_resolve replaces Resolves.
     o Namespace_symbols replaces Sym.
     o Removed the old prefix/URL tables entirely.

  - ASTs:

     o Cleaned up core AST to use only rqnames and never symbols.
     o Cleaned up algebra AST to use symbols, still using rqnames for
       variable, tuple fields and function names.

  - Started tuning of the run-time for names comparisons:

     o Re-implemented node tests to use symbols and not access
       resolved QNames anymore.
     o Factorized sequence type matching to use symbols except for
       type lookups (core types still are based on rqnames).

  - Schema import:
     o Fixed some issues with interactions between namespace
       declarations in the query prolog, schema import and type
       declarations.
     o Type declarations in the XQuery prolog should now be in a
       "declare schema { ... };" declaration and allow nested
       namespace declarations.

  - Bug fixes:
     o Fixed bug with serialization of the built-in namespaces.
     o Fixed bug with serialization of qnames using the default
       function namespace in documents.
     o Fixed bug with target namespace declaration in a library
       prolog, not taken into account during normalization.
     o Removed all usage of xsd: namespace, not used.

August 18, 2004 - Chris

- AST -> Compilation Changes 
	o New operations: MapConcat, Join (and Product redone)
	o updated printer, annotations, code-selection, 
	  compilation to use them	
	o added printing of optimized expression flag to galax-run
	o added code and coercions for these operations 	
	o Compilation of FLWOR blocks changed to use Concats
	
- Materialization interface
	o Cleaner interface
	o Allows materializations functions to have portions
	  done at runtime
	o Updated code that used the materialization features
	o Interface for hash materialization (not yet used)

- Rewrites in Optimization Phase
	o Generic code to handle rewriting
	  - optimization_util.ml(i) 
	  - tree walking code
	o Rewrite from Independent Maps -> Product
	o Rewrite from Product + Selection -> Join
	o Rewrite to remove Empty Maps
	o Rewrite to push down selection
	** a fixed point ordering, first cut
	   rule semantic will change ** 

- XML Serialization of Query Plans
	o Deserialization using procmod_compiler.ml
	o Plans obtained through galax-compile
	o Exposed processing model functions 
	  to deal with reserialization

	** Not well tested. Small bug with namespaces ***
	** will be fixed soon, then more testing      ***

- Changes to galax-run.ml/galax-compile/conf.mlp
	o Flags to print plans before and after optimization 
	o Flags ro reload and execute query plans 
	   - ("-load-logical-plan", "load-optimized-plans")
	   - Note: Only whole modules can be run	

- Compile Variable-Tuple Field Annotations	
	o Modified cannotations to allow rewalking of the tree
	o These are used during optimization rewriting

- Cleaned unsafe/safe variable setting
	o unsafe set is used when the variable is used only
	  once, safe implies the variable is materialized

August 16, 2004 - Nicola

  - WSDL:
     o Fixed namespaces in import/export.
     o Stub generated by xquery2soap uses "import module".

August 13, 2004 - Jerome

  - Processing model:
      Long due reorganization and clean up of the processing model,
      notably of the code inside ./galapi/caml_api/galax.ml.

     o Cleaned up interfaces for the various processing phase
       context. Now all context modules have a default_context()
       operation and a replace_previous_context() operation.
     o Added ./procmod/procmod_compiler module to handle combination
       of multiple processing phases.
     o Added an analysis phase in ./analysis. This phase performs
       various kinds of static analysis useful to prepare query
       optimization. This phase operates over the XQuery core and
       takes place after rewriting, and before compilation.
     o Added an optimization phase in ./optimization. This phase
       performs algebraic query optimization and takes place after
       compilation, and before code selection.
     o Removed obsolete Query_Path_Analysis_Phase from the monitor.
     o Fixed monitor to deal with the changes to the processing model.

  - API:

     o Changed galax.ml to use the new new processing model support.

  - Toplevel:

     o Changed galax-project to use the new processing model support.

August 13, 2004 - Jerome

  - Parser:

     o Fixed parsing error for 'stable order by'. [Bug reported by
       Chris]
     o Fixed syntax of variable declarations to use := instead of { },
       forgotten alignment with the new XQuery syntax.

August 12, 2004 - Nicola  

  - WSDL: Changed the obsolete call to import_embedded to
    import_schema_element in order to maintain consistency with
    Vladimir's code.

August 11, 2004 - Vladimir 
  
  Schema import: adjustments after the recent namespaces ovehaul and
  some code clean-up.  The top-level import functions have new names,
  hopefully more intuitive ones.


August 11, 2004 - Mary

  Recovered following changes, which were lost:
  [[ July 26, 2004 - Mary and Emiran
    
      - Added new built-in function: glx:file-exists:
    
        declare function glx:file-exists($arg1 as xsd:string?) as
        xs:boolean? external; 
    
        If argument is empty, returns empty.  Otherwise, checks if local
        file exists and returns true, otherwise false. 
    
        (NB: Checking that an arbitrary resource exists is tricky, so
        we're just handling local files for now) 
    ]]

  - Fixed following bugs:
    o fn:substring() used string functions to convert floats to ints;
      now uses appropriate float functions
    o fn:substring-after() and fn:substring-before() are now
      symmetric; previously used two different ways to split strings; 
      Also, occurrences of regular expression special characters are
      now quoted before calling split functions
    o fn:tokenize() was returning matches of delimiters when specified
      as non-trivial regular-expressions.  Now correctly only returns
      strings between delimiters.

  - Makefile_pcre.def
    o Removed custom -cc options

August 11, 2004 - Chris

- XML Serialization of the algebra

   o Updated XML Serialization/Reserialization *** untested ***
       - Still not hooked in, just updated the code

- New annotation and its use

  o Added annotation for bound variables in the algebra ast (and usage
    counts).
  o Added "unsafe" code to avoid materialization for variables 
    used only once.

- Code cleaning and fixup

   o Fixed bug in tuple position map.
   o Removed InputItem from the AST (it was unused).
   o Updated Makefile for usecases to use correct flags.

- Note on xmark usecase:

   o currently, xmark usecase 19 will fail incorrectly this is due to
     non-determinism in the query. This will be fixed shortly.


August 11, 2004 - Nicola

  The only purpose of this commit is to make Galax compile again, by
  eliminating the call to build_ns() in wsdl_load.mlp
  Therefore, the WSDL tool is still unusable (please see also my last
  post on galax-dev concerning WSDL related problems). 

 - WSDL:
   o small progresses on type mapping; printing problems for URIs
     still persists
 - Apache extensions:
   o the 2 Apache modules now have a common structure, separated from
     the calls  to the 1.3 or 2.0 APIs; more testing needed 


August 9, 2004 - Chris

- Added compile_annotations and code cleanup
  o Free variables, tuple fields accessed and tuple fields returned
  o Added a function in algebra_code.ml to perforam faster checks
	if a base_type can be promoted to an atomic type
  o compile annotations are a field in the AST
  o Modified pretty printer to print the annotations on the plan

- Tuple compilation
  o Tuples are now indexes into arrays handled in tuple_context_manager
  o Tuple scoping is handled in the code_selection_context
  o Materialization is now an explicit function call 
    - a default implementation is provided in algebra/materialization.ml
  o Tuple code is now based on using *side-effects* to avoid copies
    - there are no longer algebra_context copies during execution
  o Some signatures were changed to reflect
  o This method renders certain operations as nops (for example concatenation)
    **a code review is needed to decide how far we want to push these signatures **
	
- The XML plan generator and XML plan parser still do not work.
  o Updates to deal with new style prolog but, need to revise 
    to deal with namespaces

- Fixed bug previously fixed in release version
   ( AOEError compiled to AOEAnyAttr )  
 
- Added bt_can_be_promoted_to algebra_code_util.ml
  o Removes a List.mem (this was the association list 
	access during execution)

August 5, 2004 - Jerome

*** WARNING ***
  The following update makes deep changes to the Galax namespace
  support. Only limited testing has been performed, so please report
  any problems ASAP!

  The following features are impacted by that change:
    - Jungle (namespace support will have to be changed)
    - XML Schema import (namespace resolution should be checked/fixed)
    - WSDL import (namespace resolution should be checked/fixed)
*** WARNING ***

Namespaces overhaul. Phase 1.

[ Note: the following changes fix the validation bug reported by Mary,
  as proper namespace environment are now produced through export and
  can be used for validation. ]

 - Namespace:

    o Completely re-implementation of the namespace environment
      support. Namespace environment are references to binding tables,
      and are scoped. The support is made efficient by avoiding
      copying tables when unnecessary.
    o Removed obsolete code from Namespace, Resolve, Norm_top which
      was keeping track of top-level element/attribute/type
      declarations from the schema.
    o Simplified the interface for namespace environment, removing
      several unnecessary calls. Operations on namespace environment
      always rely in namespace binding tables, not individual
      bindings.

 - SAX:

    o Fixed resolved streams and typed streams to use a reference to
      the namespace environment instead of delta bindings.

 - Data model:

    o Removed obsolete namespace nodes from the data model.
    o Added namespace_environment() method for element nodes in the
      data model. This returns a pointer to the namespace environment
      for a given element node.
    o Fixed export of a data model instance to generate the right
      namespace environment.
    o Added support for namespace environments in the main-memory data
      model.
    o Fixed loading of the main memory data model to populate the
      right namespace environment for element nodes.
    o Put place-holder for namespace support in Jungle. [[ THIS WILL
      HAVE TO BE CHECKED BY AVINASH ]]

 - XML Schema import:

    o Propagated the change streams into XML Schema import. [[ THIS
      WILL HAVE TO BE CHECKED BY VLADIMIR ]]

 - WSDL import:

    o Propagated the change streams into WSDL import. [[ THIS WILL
      HAVE TO BE CHECKED BY NICOLA ]]

August 5, 2004 - Jerome

 - Validation:

    o Fixed bug introduced a few days ago when fixing another bug with
      the validation context (sic).

August 5, 2004 - Jerome

 - XQuery context:

    o Added support for xmlspace, construction mode and ordering mode
      in the context.
    o Now effectively removing boundary whitespace during
      normalization when xmlspace is set to strip.
    o Removed obsolete -xquery-whitespace flag.
    o Fixed bug in attribute construction, wrongly removing boundary
      whitespace.

August 4, 2004 - Mary and Amelie

 - Projection:

    o Recompiled galax-project to compute and emit path-projection
      structures.

 - Galax API:

    o Added static_ctxt_from_xquery_module.

August 4, 2004 - Jerome

 - Cursor support:

    o Added cursor_of_stream function to create a cursor from a Caml
      stream.
    o Removed all usage of try...with in cursor operations, in an
      attempt to remove a supposedly expensive idiom.

August 3, 2004 - Chris

 - Fixes and Clean-up
   o Fixed missing portions of match in algebra_print_common.ml
   o Fixed external variable handling.
   o Backtracked compilation of tuple creation with empty-tuple
       - again generate the Map in [] portion
   o Hide the variable_context_manager a bit more 

August 3, 2004 - Jerome and Chris

 - Printing:

    o Fixed printing of validation expression when validation mode is not
      present.

August 3, 2004 - Jerome and Chris

*** WARNING ***
  The following update makes deep changes to the Galax run-time,
  notably in the way FLWOR expressions are compiled, and the way
  variables are compiled. Some features might be unstable and static
  typing has been disabled.
*** WARNING ***

 - Changed compilation for FLWOR expressions to use tuples:

    o Added a FLWOR block in the XQuery core.
    o Removed tuples expressions from the XQuery core.
    o Propagated the removal of tuple from the core and addition of
      FLWOR expressions to: printing, parsing, and rewriting.
        ** NOTE: The change was NOT propagated to static typing, which
           must be fully revised anyway. - Jerome ***
    o Added basic operations over tuples in the algebra:
	Tuple construction
	Tuple field access
	Tuple concatenation
    o Added (nested) relational algebra operations:
        Map
	Cartesian product
	Selection
	Projection
        Projection with duplicate removal
	Join (only implemented as a nested loop for now)
    o Implemented compilation of FLWOR expressions into the (nested)
      relational algebra.

 - Compiler:

    o Extended the algebra compiler to support XML updates and the
      query prolog.
    o Added a special evaluation function for operations in the prolog
      that do not return a result, but change the context.

 - Improvements to the evaluation and compilation context:

    o Pulled out as much things as possible from the evaluation
      context to the compilation context, enforcing the computation of
      that information at compile-time rather than at run-time.
    o Complemented the dynamic context with "special" support for the
      the tuple context (used for pipe-lined tuple based
      evaluation).
    o Added support for variables to be just references into arrays
      after compilation. The method in which functions are done 
      is *NOT* re-entrant.

 - Code reorganization:

    o Moved printing of the AST(s) into a separate ./print directory.
    o Split Xquery_ast_util into separate modules, one for each ast.
      (e.g., utilities for the core type ast are now in
      Xquery_type_core_ast_util.)
    o Added code_selection_context.ml(i), algbra_code_update.ml(i)
            variable_context_manager.ml(i), algebra_code_top.ml(i)

 - Bug fixes:

    o Fixed pretty printer for PI's kind tests and comment
      constructors.
    o Fixed bug in the static analysis of free variables for FLWOR
      expressions with a an index variable, and in the default case
      in a typeswitch.
    o Fixed a bug in the compilation process for statements and
      library modules. (The compilation context was not propagated
      properly).
    o Fixed up algebra pretty-printer to deal with updates.

August 3, 2004 - Mary

 - Compilation fixes:

    o Fix to config/Makefile_jungle.def.
    o Changed examples/Makefile* to only run API examples if the API
      libraries are built and installed.


***** Galax Release 0.4.0 August 2, 2004 *****

August 2, 2004 - Jerome

 - Aligned the XQuery prolog with the new working draft.

    o Imposed module declarations on library modules.
    o Added support for construction declaration, xmlspace
      declaration, ordering declaration.
    o Added parsing for validation mode. Currently, only strict
      validation is supported.


August 2, 2004 - Mary

 - Examples 

    o removed base-uri example that is file-system dependent and
      changed expected output.

 - More Makefile changes: 

    o Now distinguish between SHARED link options for CC and
      OCAMLSHARED link options for ocamlmklib

August 2, 2004 - Mary

  - SOLARIS port 
  
    o config/Makefile.gen, Makefile.unix, Makefile_pcre.def , Makefile_jungle.def
      examples/Makefile.config
      Added link options for building shared libraries

    o C and JAVA API Makefiles
      ocamlmklib raised errors from 'ar' when constructing libraries; reverted
      to using gcc by hand and this worked. 

July 30, 2004 - Vladimir

  - Schema import: some improvements in error messages.

July 30, 2004 - Mary and Avinash

 - Tested Jungle installation and added tests in examples/jungle
   o Fixed config/Makefile_jungle.def to build two jungle libraries:
     libjungle.so and libjungleopt.so 
   o Fixed examples/Makefile.config to include correct link options
   o Added load: target to examples/jungle
   o Added (preliminary) documentation for installing and testing jungle

July 30, 2004 - Mary

 - Installation scripts
   o Added libpcre.so* dynamic libraries to binary distribution
     directory : Galax/lib/c 
   o Updated corresponding documentation

 - Usecases
   o Updated calls to F&O functions with new names

 - examples/ and usecases/ Tests

   o Eliminated dependency on LIBPCRE location BUT running these tests
     REQUIRES 'make install' beforehand.

July 30, 2004 - Doug

 - Dates/Times

   o Updated the date/time/duration functionality to correspond to the
   new functions and operators document

   o All 'get' functions have been renamed such that the 'get_ has
   been removed. (ie get_years_from_date is now years_from_date.)

   o New arithmetic functions on durations have been added

   o Most of these changes were actually installed last update, but
   not documented in the changes file.

 - URI's

   o Attempted to allow windows file URI's to be parsed using the old
   URI parser in tools/http/http.ml. Netstring was not parsing them
   correctly.
   
   o The change was made in glx_http. AnyURI still relies on netstring
   however in all cases as it did before.

July 29, 2004 - Jerome

 - Compilation:

    o Fixed dependency problem in the config Makefile for PCRE. Was
      forcing a recompilation of the pcre_stubs library each time make
      was called.

 - Example:

    o Fixed syntax error in ./examples/Makefile.release. Was not using
      TABs properly.
    o Changed the makefiles to make the error messages more visible.

July 29, 2004 - Doug
     
  - DTDs
   o Extended the sax_parser to now parse DTD's, either embedded or
     standalone.

   o Galax internal types are now created from the element
   declarations within the DTDs

   o Support is included for elements, attribute lists, processing
   instructions, comments, and entities are resolved within the xml
   document.

   o There is no support for entities within the DTD itself. If the
   parser encounters an entity within the DTD it will print a warning,
   and consume the rest of the tokens until the xml document begins

  - galax-mapschema
   o Now takes an optional argument '-dtd' which allows it to take a
   dtd as input rather than an xml schema. -validate will validate the
   document against the dtd.

  - galax-parse
   o Also now takes an optional argument '-dtd', in case the user
   wants to perform validation against the internal dtd.

July 28, 2004 - Mary

 - Usecases/Examples
   o Finalized tests. In galax/, execute "make tests" to run usecase
     and API tests 

 - Website
   o Fixed several bugs in formatting queries
   o Changed output format to Serialize_As_XQuery because very long
     strings were causing printing problems in browser.

 - Print_xquery 
   o Fixed bug -- '{}' delimiting enclosed expressions in attributes
     were not printed correctly,

July 28, 2004 - Doug

 - Made the 'jungle specific fix' in dealing with URI's given in the
   fn_doc() global to all backend registered data sources. Made the 
   change in glx_http.ml.

July 27, 2004 - Mary

 - Java API and examples/java_api
   o Added date, time, dateTime, yearMonthDuration, dayTimeDuration
     atomic types.    
   o Added tests to examples/java_api/Test.java


July 27, 2004 - Chris
 - Code Clean-up

  o Added access subexpression functions to xquery_ast_util.ml(i) 
    for algops
	- switched code selection to use these functions instead
	
  o Added a pretty printer for the algebra (algebra_printer.ml(i))
	- Hooked into procmod/galax-run
        - Removed XML printer
	- Split code into utilities for printers

  o Identified location of change for symbol compares 
	- eval_node_test needs to be changed)
	- There is a comment - eventually should switch to all 
          symbols/integer compares during execution until serialization. 
          Probably with unified string pools too.

  o Switched pool.ml from using exceptions instead of .mem
  o Switched datatypes can_be_promoted_to to a switch 
    instead of an association list lookup

July 26, 2004 - Doug

 - URIs

   o Fixed the bug which wasn't recognizing jungle URI's correctly

   o The fix is currently "jungle specific" and will need to be
     generalized in the future for other backends, but this should
     serve as a short term solution. 

     The change was made in base/glx_http.ml. The reason for the bug
     was that the Neturl module doesn't consider a fragment (ie
     anything that comes after the '#' in a URL) to be considered as
     part of the path. Therefore in a jungle URL,
     jungle:///tmp/Galax#test it was returning the path as /tmp/Galax.

     The 'jungle specific' fix on line 75 of glx_http was to concat a
     '#' and the fragment onto the local path of the URL. If there is
     a consensus that the fragment should always be included in the
     path, then the Neturl module can be modified. Input?


July 26, 2004 - Mary and Emiran

  - Added new built-in function: glx:file-exists:

    declare function glx:file-exists($arg1 as xsd:string?) as
    xs:boolean? external; 

    If argument is empty, returns empty.  Otherwise, checks if local
    file exists and returns true, otherwise false. 

    (NB: Checking that an arbitrary resource exists is tricky, so
    we're just handling local files for now)

  - (doc/ under constant construction)

July 23, 2004 - Chris

 - Algebra code cleaning:

    o Changed build_default_builtin_function to build function 
         during code selection
    o Changed build_default_overloaded_function to build possible
         signatures at code selection time

 - Algebra printing:

    o Namespace problem fix: Changed algebra_uri and algebra_prefix
      back to glx to prevent printing of names. Serialization of the
      plans will not be correct.

 - Fix: Changed algebra_uri and algebra_prefix back to glx to 
	prevent printing of names. Serialization of the plans
	will not be correct.

July 22, 2004 - Mary 

  - doc/ 
    Removed install: dependencies on manual.pdf manual.html

    They should already be in the CVS repository.

  - Fixed bug in config/Makefile.unix.  Accidentally made changes to
    config/Makefile instead of Makefile.unix.

July 21, 2004 - Mary

 - New HTML & PDF Documentation (Check it out!)

    o Deleted STATUS, RELEASE_NOTES, SOURCE_CODE, STATUS files

    o Created doc/ subdirectory 

      To access: cvs update -d doc

      If you want to build the doc yourself, you need HeVeA
      (http://pauillac.inria.fr/~maranget/hevea/)

      All the generated doc is in CVS so you don't have to build it
      yourself. 

 - Updated website to refer to new documentation


July 21, 2004 - Chris
 
 - Added a printer and parser for the algebra:

    o Added XML serialization for both logical and physical
      plans. This is in:
         algebra/algebra_print_common.ml(i)
         algebra/algebra_print_util.ml(i)
         algebra/algebra_print_xml.ml(i)
    o Current limitations:
        - Updates are not supported.
        - Physical plans do not have indication of which code has been
          selected.
    o Added top-level flags for printing of algebraic plans.
    o Fixed processing model to trigger printing of algebraic plans.
    o Preliminary parser for serialized plan, but it is still
      incomplete and is not hooked into the rest of the system.
      This is in:
        algebra/algebra_parse_plan_xml.ml(i)
    o Made string_of_axis visible (included its sig in print_common.mli)  

 - Removed obsolete modules:
      algebra/dynamic_context.ml(i)
      algebra/algebra_ast.mli

 - Bug fixes:

    o Fixed Bug in code selection - signature of element and attribute
      code was incorrect.
    o Fixed Bug in compile (text constructor had incorrect number of
      dependent expressions).
    o Fixed Bug in small_stream_context.mli that prevented comments
      from being output.

July 20, 2004 - Mary and Emiran

 - Added LOCALPREFIX to all Makefiles -- some contained it, others
   didn't.

July 19, 2004 - Mary

 - PCRE tool

   o Updated config/Makefile_pcre.def to include correct options for
     linking 

 - Galapi, API Examples, Website
   
   o Updated Makefiles and Makefile_*.def files for Galapi, examples/,
     and website so they link correctly with PCRE library

 - Galapi
   galax.ml : commented out debugging print statements 

 - The Return of Nestring
   
   o Needed tools/netstring to support Web site
   o Deleted tools/glx_netstring and restored tools/netstring

 - Namespace

   o Fixed bug in printing of unresolved QNames with empty prefixes.
     Was printing ':NCNAME'. Now prints 'NCNAME'

July 16, 2004 - Jerome

 - Added support for computed comment constructors.
 - Alignment with July XQuery 1.0 working drafts:

    o Renamed xdt:untypedAny to xdt:untyped.
    o Checked that value comparison for two empty sequences returns an
      empty sequence.

July 15, 2004 - Chris

 - Modified algebra to support semantic annotations and add signatures
   for the operations.

    o Propagated the change to compiler, algebra, procmod, ast and
      execution.

 - Changed code selection to also close off dependent expressions at
   that level rather that at execution time.

 - Changed the the expression portion of the algebra to used arrays
   instead of lists. That should result in a more efficient run-time.

    o Added get_array_paramx to base/param to deal with arrays (for
      builtins).
    o Modified algebra_code.ml to deal with arrays imperatively.
    o Created Cursor.cursor_array_fold, now used in constructors
      instead of the Cursor.cursor_list_fold.
    o A few underlying functions still use lists (notably: Overloaded
      function matching and tuple materialization).

July 12, 2004 - Chris and Jerome

 - New algebraic compiler for Galax!

    o Added two incarnations of the algebra AST.

         -- The first incarnation is in ./ast/xquery_algebra_ast.mli,
         and is polymorphic in one parameter which corresponds to the
         actual code for the operation. This is similar to a 'logical'
         query plan.

         -- The second incarnation is in ./algebra/algebra_type.mli
         and instantiates the former with specific evaluation
         code. This is smiliar to a "physical" query plan.

    o Added two new processing phases:

         -- The "compilation" phase compiles a core expression into a
         logical query plan. This phase is located in ./compile.

         -- The "code selection" phase selects actual evaluation code
         for each algebra operation, producing a physical query
         plan. This phase is located in ./algebra.

    o Extended the algebra AST and compiler for updates. The compiler
      for updates is mostly a place holder for now, until we know what
      we want as an algebra for XML updates. Fixed the evaluation code
      for updates accordingly.

    o Extended the algebra AST and compiler to deal with the prolog
      and modules. Fixed the evaluation code accordingly.

    o Removed the old compiler!

    o Extended the monitor, the processing model, and the API to use
      the new compiler.

July 8, 2004 - Doug and Mary

 - Removed tools/netstring and replaced with tools/glx_netstring

    *** NB: You must do 'cvs update -d tools/glx_netstring' to get new
    directory. 

 - Compilation problems: 
 
   There is a CVS versioning problem with schema/schema_import.mlp.
   'cvs update' may not return the head of the branch for this file.
 
   If you have problems compiling schema/schema_import.ml, then do a
   fresh checkout 'cvs checkout galax' and you will get the right version.

July 8, 2004 - Chris and Jerome

 - First step to the algebra. Algebra operations, compilation and
   execution for expressions is in place. This is not hooked up to the
   rest of the system yet.

    o Created a new AST structure for Algebra operations. In
       ./algebra/algebra_ast. Utilities for the algebra AST are in
       ./algebra/algebra_util.
    o Split the evaluation code in ./evaluation/eval_expr into
      evaluation code in ./algebra_code and coercion operations in
      ./algebra_coercion. The old evaluation/eval_expr is still in
      place until the new version can take over.
    o Updated ./compiler/compiler_expr to compile into the new
      algebra. compilation for updates and the query prolog is still
      missing.
    o Added a new ./evaluation/algebra_execute module as the new
      evaluation module for expressions working over the algebra.
    o Reorganized part of the code:
       Moved ./evaluation/eval_util --> ./algebra/algebra_code_util

July 6, 2004 - Doug

 - URI's

    o Fixed the resolve-uri function in anyURI module such that it now
      gives an absolute URI rather than a relative one
    o Hooked up netstring underneath glx_http so that there is
      consistent URI processing throughout galax. Previously, URI's
      were handled differently depending on whether they were used to
      load documents or used as a datatype.
    o Eliminated half of the netstring library so that now it only
      contains the modules which are necessary to galax.
    o Added tests for anyURI functions and operators in xqueryunit

July 6, 2004 - Jerome

 - Fixed bug in pretty-printing of contructors (document, text and
   elements).

July 6, 2004 - Jerome

 - Fixed bug in namespace resolution when using schema import schema.
   Namespace declarations in the schema import statement was not
   propagated properly. [Bug reported by Nicola]

June 30, 2004 - Jerome

 - Fixed typo in physical_value module (how did it get there???).
   [ Thanks to Emiran for notifying me of the problem! ]

June 28, 2004 - Jerome

 - Type system:

     o Added built-in definitions for xdt: types (xdt:untypedAtomic,
       xdt:untypedAny, xdt:yearMonthDuration, xdt:dayTimeDuration)
       which were missing. [Fixes a bug reported by Emiran].

June 25, 2004 - Doug

 - Documentation:
      
     o Update the STATUS document to reflect new functions and
       operators implemented

     o Internal documentation of dateTime, regularexp, and algebra_fn
       using ocamldoc format

June 25, 2004 - Jerome

 - Jungle:

     o Fixed loading not to stop whenever encountering PI's or
       Comments, but ignoring the event instead.
     o Fixed bug in make install, copying libjungle.so into
       libjungle.a.

 - Namespaces:

     o Fixed a HUGE bug in namespace support, which was accumulating
       uri/prefix binding each time it was registering a new QName,
       resulting in stack overflows pervasively throughout the system.

 - Data model:

     o Changed the DM API to distinguish between nodeid and document
       order methods. node equality is checked through nodeid() while
       other document order operations are checked through docorder.

 - Cursors:

     o Replaced the caml stream library by the new cursor library,
       more efficient (slightly). This is effective except for the
       modules that use the Caml stream patterns: Schema import and
       WSDL import.

 - Updates:

     o Fixed bug in main-memory implementation of delete for
       attributes.
     o Fixed insertion of element nodes buggy when occurring at
       the beginning of the element content.
     o Fixed insertion of attribute nodes. The insert location is
       ignored as attributes are not ordered. This also checks that
       the newly inserted attribute does not already exists so that
       it does not create two attributes with the same name.

 - PCRE:

     o Cleaned up some of the makefiles and removed targets from the
       repository.

June 22, 2004 - Doug

 - More string bug fixes
    
    o Substring now takes doubles rather than ints to correspond to
      the spec.

    o Added new tool, PCRE, to handle perl compatible regular
      expression matching.

    o Added datatypes/regularexp.ml, mli to implement the string
      functions dealing with regular expressions.

    o Many more string tests added in xqueryunit which exposed more
      minor bugs in string functions. Fixed all that were found.

    o Allowed the specified variadic string functions to take varied
      number of arguments.

    o Updated INSTALL file to instruct users on how to obtain PCRE 4.5
      which is a requirement for regular expression
      functionality. This comes standard with the majority of linux
      distributions these days, so it shouldn't cause too much worry.

 - Bugzilla
    
    o It's installed. Look for it to be configured and useable
      shortly.

 - Date/Time
    
    o Added support for fn:adjust-time-to-timezone,
      adjust-date-to-timezone, and adjust-dateTime-to_timezone

    o Added and ran more tests for dates/times to xqueryunit

    o Date/Time library now includes support for every function and
      operator in F&O

    o Began adding internal documentation to datetime.ml and
      datetime.mli

June 11, 2004 - Doug

 - String functions
   
    o Fixed the substring function in algebra_fn.ml

    o Implemented variable number of parameters to the variadic
      functions

    o Created a number of tests for string functions and operators
      exposing a number of bugs

    o Fixed a number of small incorrect behaviors in different
      functions

    o Allowed the () empty sequence when passed as an argument to be
      interpreted as the empty string "", as specified in the
      functions and operators document.

June 8, 2004 - Doug

 - Xqueryunit updates

    o Made updates to xqueryunit and updated the the appropriate files
      in the usecases and regres directories

 - More date/time bug fixes.

June 7, 2004 - Mary

  - SIGMOD Demo Configuration

     o Added detailed instructions for setting up demo in
       examples/wsdl/yoohoo/README

  - Website
     o Updated code to align with current Galapi 
     o Added YooHoo Web service to Galax demo Web site

  - Usecases
     o Removed type declarations in *_context.xq so that demo would work

June 7, 2004 - Jerome

  - WSDL:

     o Small clean up of the mod_xquery.c code, after the big bug fix.
     o Some more testing of the jabber support. Fixed the examples a
       bit.

June 7, 2004 - Jerome

  - WSDL:

     o Cleaned up module and service import, really done now as a
       pre-processing step before normalization. Makes the code easier
       to maintain and makes supports for namespaces clearer.

     o Fixed bug in mod_xquery, which allowed multiple initializations
       to the main module context and processing context, causing
       SIGSEV in some cases.

     o Fixed Jabber dynamic linking problem, compilation of the
       mod_xquery.so library *have* to have the jabber libraries
       passed during compilation otherwise it SIGSEV during calls to
       the glx:jabber functions.

June 4, 2004 - Doug

  - DateTime

     o More Date/Time/Duration bug fixes

June 4, 2004 - Mary

  - Jabber

     o iksemel_wrapper.c: 

      Changed unprefixed QNames to be in the empty namespace.
      Removed reference to glx:jabber:wrapper namespace.

  - Module/service import
 
     o As temporary hack, explicitly add all namespaces in imported
       service or module into context of importing query

  - Apache 

     o mod_xquery.c Put error-checking guards around _all_ calls to
       Galax API.

  - YooHoo service

     o examples/wsdl/yoohoo
 
       Added README file explaining current problems with running
       Jabber and Apache at same time

June 3, 2004 - Jerome

  - Base URI:

     o Fixed a huge bug in the default base uri, breaking all file
       access. [ Bug found by Avinash ]

June 3, 2004 - Doug

  - Date/Time

     o Tested and fixed a number of bugs in the DateTime module
       concerning manipulations of dates/times/durations.  A few bugs
       remain and will be fixed tomorrow

June 3, 2004 - Mary and Jerome

  - WSDL Module

     o Implemented import module to WSDL export 
     o Fixed namespace hack to align WSDL target namespace and
       server implementation 

  - Makefile

     o Fixed compilation dependencies between Galax, tools, and APIs
       (linking hell!)

June 2, 2004 - Mary and Jerome

  - Added "import module" statement, which does XQuery source inlining

  - WSDL

     o Fixed bug in argument order 
     o Fixed bug in extracting content of soap parameter on server
       side

June 2, 2004 - Jerome

  - Processing model:

     o Fixed bug introduced when fixing the stack overflow problem,
       optimization phase was not properly called anymore during
       compilation of statements.

June 2, 2004 - Jerome

  - Support for base-uri:

     o Added base-uri as a parameter to the processing context.
     o Added support for 'declare base-uri' in the query prolog.

  NOTE: Some clean-up for URI support (lower levels) are still
  necessary.

June 2, 2004 - Doug

  - DateTime

     o Fixed get_timezone_from (date/time/dateTime) functions
     
     o Corrected bug in valid_date dealing with the difference between
       Unix time and galax time representations
     
     o Fixed comparisons in dateTimes in order to account for
       timezones being represented as durations

     o Changed default settings for local_timezone and
       current_dateTime contexts. If an error is raised, rather than
       galax crashing from initialization, it sets the current_dateTime
       to January 1st, 1970 at midnight, and the local timezone to UTC

June 2, 2004 - Nicola

  - Wsdl
 
     o corrected an old bug in the processing of the xquery2soap
       parameters: the program failed when it was given a module from
       another directory

  - Jabber

     o changed the namespaces in the output of the Jabber wrapper

June 1, 2004 - Nicola

  - Wsdl

     o updated the contexts for the WSDL usecases
     o added an Apache2 XQuery module in extensions/apache

June 1, 2004 - Avinash

  - Update

     o Rewrote the update code to allow optimizations. The concrete
       update generated are now aligned with the data model update
       operations.

     o Added 3 flags for 3 update optimizations, OI, BI and SBI.

  - Jungle

     o Fixed the code to work with modified nodeid.

May 31, 2004 - Mary

  - DateTime

     o Fixed bugs in valid_date() and local_timezone() 
  
May 31, 2004 - Nicola

   - Wsdl
     o Catch the lexing exception generated by 
       Namespace.uqname_element_of_string for an invalid QName
     o Modified the return statement in a WebService client call:
       if a SoapEnv:Fault is received, fn:error() is called

May 28, 2004 - Jerome

   - XQuery parser:

      o Fixed bug in parsing of wildcard names of the form 'a:*' and
        '*:a'.

May 28, 2004 - Jerome

   - Type system:

      o Changed OLD built-in types in pervasives.xqp not to use
        recursive groups anymore.
      o Removed dependency of the normalization, and evaluation phase
        on any of the old type system. One last OLD module added to
        normalization which only contains normalization from the new
        to the old type system, which is still necessary for typing,
        and cleaning.

   - Normalization:

      o Removed the use of subtyping during normalization of function
        calls. This is now done directly in a syntactic way by looking
        at the sequence type structure.

   - Type matching:

      o Implemented the new sequence type, including matching based on
        named typing.
      o Fixed printing of the type system to put attributes after the
        content model, which is aligned with parsing.
      o Changed the error for type matching not to use the value
        anymore since the value may be a stream already (partially)
        consumed.
      o Fixed bug in sequence type matching for document-node(ncname).

   - SAX cursors:

      o Made node-identity allocation a separate stream
        operation. This gives more flexibility to the query compiler
        in case one needs identity operations, but does not want to
        materialize a DOM tree.

   - Loading and fn:doc():

      o Re-implemented loading for the main-memory data model based on
        an input stream with node identity. This significantly
        simplifies the loading phase.
      o Changed the 'document buffering' approach that is required to
        make sure multiple calls to the same 'fn:doc' function return
        the same node identities. Now the document is not buffered in
        memory, only the doc id is kept in the processing context.
        Multiple calls to the fn:doc() function will results in
        parsing the document multiple times.

May 27, 2004 - Mary 

  Preparation for SIGMOD Demo

  - Apache server

    Moved tools/apache to extensions/apache
    Removed tools/apache
    Apache configuration info in extensions/apache/README

  - Jabber server

    Configuration info in examples/wsdl/jabber/README

  - XQuery Jabber extensions and library

    Configuration info in tools/jabber/README

  - WSDL/Apache tests
    Added tests for deploying Galax WSDL service under Apache in:

      examples/wsdl/apache/
     
    Configuration info in examples/wsdl/apache/README
   

May 26, 2004 - Mary

  - Parser/Normalization

     o Added import WSDL service declaration to prolog.  Syntax:
       import service prefix = "filename"
       Implemented by inlining output of WSDL tool

     o Fixed DateTime.local_timezone() to correctly compute timezone
       when local and UTC days differ
  
May 26, 2004 - Vladimir

    - Schema normalization: attribute and element groups, including
      checking for cycles; somewhat tested.

    - Schema import: some improvements to error messages.

May 25, 2004 - Jerome

   - Cursors:

      o Fixed a bug in the cursor implementation of fn:subsequence.

   - DateTime:

      o Turned global variables in Dynamic_context into functions.

May 25, 2004 - Jerome

   - Fixed the Makefile to compile Galax+Jungle on solaris.

May 24, 2004 - Mary

   - DateTime

      o Whoops -- Fixed error in dateTime.ml which didn't acount for
        the year starting at 1900 from the Unix System call.

   - Galax-run toplevel:

      o Default timezone set to local timezone

May 24, 2004 - Doug

   - Dates/Times:

      o Fixed bug with string_of functions from last update.
      o Began to implement multiplication and division of durations to
        complete arithmetic operations.
      o Fixed previous error in dateTime.ml which didn't acount for
        daylight savings time.
      o Fixed error in dateTime.ml which didn't acount for the year
        starting at 1970 from the Unix System call.

May 24, 2004 - Mary

   - Algebra_fn, Funcs_and_ops:

      o Added fn:current-dateTime,date,time functions

   - Dm:

      o Added supported casts between dateTime/date/time

   - Galapi

      o Added date/time/duration times to C API
        Still working on Java API

   - Galapi Examples

      o Added date/time/duration constructors to Caml and C API
        examples
      o Added exit_on_error guards to all API calls in examples/c_api
      o Updated prolog syntax in examples/seq_context.xq

May 24, 2004 - Doug

   - Dates/Times

      o Finished the dynamic context integration into dates and times
        with an implied local timezone.
      o Corrected the string_of_ functions on dates/times and
        durations
      o Fixed lack of default matching cases in date/time library to
        raise errors where appropriate

May 21, 2004 - Jerome

   - Schema import:

      o Put back schema import into place, using the new type system.
      o Added some support for location hints. Location hints are
        always used if present.
      o Fixed the syntax of schema import for default element and
        location hint.

May 21, 2004 - Jerome

   - Syntax:

     o Fixed lexing bug, where '*' was recognized still as a node test
       in the operators state, and whitespace was forced before a '*'
       standing for multiplication.

   - Namespaces and types:

     o Hacked something in the namespace support to deal with the
       empty prefix, since Vladimir is using "" instead of
       NSDefaultElementPrefix.

   - Validation:

     o Added xsd built-in types in the new type system, missing.
     o Added support for the validate expression. Currently, the only
       form that is supported is the validate { Expr } form which
       validates a single document, element or attribute node against
       any declared top-level declaration.

May 21, 2004 - Mary and Doug

  - More date/time library integration.

May 21, 2004 - Mary

  - Caml API

     o Added date/time/duration times to Caml API
       Still working on C/Java APIs
     o Added implicit timezone parameter to external context.

May 20, 2004 - Doug

  - The date, time, and dateTime constructors now take
    a dayTimeDuration option type to represent the timezone.
    If the timezone is given as 'None' then the dynamic
    context, as defined by the user (defulat localtime)
    will be used instead.

May 20, 2004 - Mary

  - WSDL Examples

     o Removed wsdl_usecases/
     o Moved content of wsdl_usecases to wsdl/examples/usecases

May 20, 2004 - Doug

  - Fixed syntax error in pervasives.xqp

May 19, 2004 - Doug

  - Arithmetic operations as well as comparison functions are
    implemented for dates/times/durations.

    Extraction functions are implemented for dates/times/durations

    All operations on dates/times/durations have been finished
    except for multiply and divide on durations, and timezone
    adjustments. Waiting on context situations for these.

    Next up will be testing this library.

May 18, 2004 - Mary & Jerome

   - Updated jabber configuration:

     Galax-dependent Jabber code in extensions/jabber
     Galax-independent Jabber code in tools/jabber

     Created extensions/jabber
     Removed tools/jabber/*.mlp

May 14, 2004 - Doug

   - Date/Time/Duration library

     Added support for above types: lexing, semantic checks,
     comparison functions

     Working on integrating w/ rest of system and also arithmetic
     functions

May 11, 2004 - Mary

   Updated wsdl/examples/README

   - Info on how to set up Jabber server and Gaim clients

   Added galax/SOURCE_CODE 

   - Contains description of processing model and high-level
     description of all source code directories

May 7, 2004 - Jerome

  - Syntax:

     o Fixed various bugs in the new XQuery parser related to the
       distinction between 'expr' and 'expr_single', the former being
       a sequence of the latter.

       Now a top-level sequence (i.e., without parenthesis) is allowed
       in the following:
         - Function body
         - Where clause of a FLWOR expression
         - Input expression of a typeswitch
         - Input expression of a conditional
         - Inside an XPath predicate
         - Inside a validate expression

     o Changed the syntax of the validate expression to use "{ .. }"
       instead of "( .. )".


May 4, 2004 - Mary
   
   Bug Parade 
   (It's time to install Bugzilla! http://www.mozilla.org/projects/bugzilla/)

   - idiv 
     Default type changed from double to integer   

     Reported by: Peter Coppens <Peter.Coppens@datadirect.com>

   - Optimization 
     Long-standing bug : Substitution of Expr1 for Variable in Expr2
     did not check whether Expr2 captured any free variables in Expr1

     Reported by: Philippe Michiels <philippe.michiels@ua.ac.be>

   - Short-circuited Boolean operators
     
     Changed implementation of boolean and/or expressions to have
     short-circuit semantics by normalizing them as if-then-else
     Removed all occurrences of op_or and op_and

     Reported by: Avinash Vyas <vyas@research.bell-labs.com>

   - Static typing and dynamic evaluation of value comparisons 
     Value comparison operators should not permit comparison of
     non-matching types, either statically or dynamically!

     Reported by: Peter Coppens <Peter.Coppens@datadirect.com>

   - Sibling axes
     Changed evaluation/eval_util so that
     sibling axes on attribute and namespace nodes should now yield
     empty sequence.

     Reported by: Philippe Michiels <philippe.michiels@ua.ac.be>

   - Fixed open-file descriptor leak in base/io module 
     
     Reported by: Prolific 2000 <prolific2000@hotmail.com>

   - Java API/galax_jni_stub.c
     
     Changed Java_galapi_Galax_importMainModuleFromString
     to 
     Java_galapi_Galax_nativeImportMainModuleFromString

     Reported by: Wim Le Page <Wim.LePage@ua.ac.be>


May 1st, 2004 - Jerome

  - Syntax:

     o Changed 'define key' to 'declare key', more consistent with the
       rest of the XQuery syntax.
     o Added a parser for the new XQuery type system. Please
       look at use cases or use the galax-mapschema tool to see
       example of the new sytnax.

  - Type system:

     o Added a pretty-printer for core XQuery types.
     o Changed the XQuery core AST to remove the old type
       system. Changed sortdecls to issds
     o Distinguished between complex and simple types in the core
       XQuery ast pretty printer, and the new parser.
     o Fixed a namespace problem for the wildcard built-in types
       (xs:anyType).
     o Added an option to print the normalized XQuery types to the
       galax-mapschema command-line tool.
     o Removed TSpecRef in the XQuery type AST, redundant with
       TTypeRef for complex types and STypeRef for simple types.
     o Temporarily moved
     o Update pervasive.xq and usecases with the new type system
       syntax.
     o Removed the following 'OLD' modules:
        ast/OLD_xquery_type_ast
        ast/OLD_print_type

April 26, 2004 - Jerome

  - Fixed a few more things in the (new) type system pretty-printer.

April 28, 2004 - Avinash

	- Changed the implementation of Jungle. Now it uses first child and 
	  next sibling index instead of a children index.
	- Added a metadata db to the store (used for storing max_preorder
	  and user specified buffer size). 
	- Modified jungle to work with cursor interface.
	- Added partial support for updates (delete and insert). Deletes 
	  are *preferred* over inserts in the sense that no renumbering is 
	  done for deletes. Insert is implemented as two step process where
	  in first step we compute all the atomic events (e.g. inserting 
	  element node) and in second step apply them. This allows us to 
	  avoid the process of renunbering for each node when inserting a 
	  subtree. Added new file jungle_load_update.ml.
	- Added support for user specified buffer size.
	- Removed a bug in sync method in jungle.  

April 24, 2004 - Jerome

  - Compilation fix for the Jabber optional tool. Removed
    ./tools/jabber/glx_jabber.cmo from the standard targets in
    ./Makefile. Moved the jabber call built-in function back into the
    ./tools directory. Now uses dynamic registration of the function
    at linking time. This makes the code more modular.

    [Thanks to Mary for adding the registration function in the
    Algebra_fn module]

April 23, 2004 - Jerome

  - Named typing: Cleanup and fixes to the printing module for the new
    XQuery type system. [Named typing, as written by Vladimir]

     o default element namespace --> declare default element namespace
     o Removed print_uqname, already in Print_common.
     o Turned all define's into declare's more consistent with the new
       XQuery syntax.
     o Fixed some unwelcome nested boxes and whitespaces making the
       pretty printer not so pretty.

     o OLD/NAMED type modules remaining: This is a step for
       preparation of the coming migration to the named type system
       throughout the code. The goal is going to remove all of the
       code marked as 'OLD'. Hopefully this will simplify
       identification of what has to be removed and helped in
       idenfying sub-components which can be done independently.

        - Marked the old type system AST modules, and old schema
          support as 'OLD'. Propagated this renaming to the rest of
          the code, without any other change.
             in ./ast:
               Print_type           -> OLD_print_type
               Print_type_core      -> OLD_print_type_core
               Xquery_type_ast      -> OLD_xquery_type_ast
               Xquery_type_core_ast -> OLD_xquery_type_core_ast

             in ./schema:
               Namer              -> Schema_namer
               Prefix_manager     -> Schema_prefix_manager
               Validation_context -> Schema_validation_context

        - Renamed the 'named' modules to the proper 'type' or 'schema'
          modules.

             in ./ast:
               Print_named                -> Print_type
               Xquery_type_ast_named      -> Xquery_type_ast
               Xquery_type_ast_core_named -> Xquery_type_core_ast

             in ./schema:
               Glushkov       -> OLD_glushkov
               Schema         -> OLD_schema
               Schemautil     -> OLD_schemautil
               Simplification -> OLD_simplification
               Subsumption    -> OLD_subsumption
               Theta          -> OLD_theta
               Typeutil       -> OLD_typeutil

         - Minor fixes to Schema_namer: Now uses fs: instead of glx:
           as a namespace for anonymous types. Now uses the already
           existing Id module in ./base as a counter.

  - Toplevel:

     o Added -verbose option in the galax-mapschema command-line.

April 20, 2004 - Nicola

  - added a new tool, in ./tools/jabber, that acts as a wrapper for
    Jabber client functions  
  - Algebra_fn: added the glx:jabber-buddies function which retrieves
    the roster and the list of available buddies for a given user
  Note: if ./config/Makefile_jabber.conf, the function only outputs 
        an error message 

April 19, 2004 - Philippe

  - C API:
  
      o Work around for the other API functions added (see below).

April 18, 2004 - Philippe

  - C API:
  
      o Work around for the galax_set_sbdo_kind API function.
        A problem with accessing the hash_variant from the C 
	code caused the API to crash. A temporary fix, consisting
	of passing an integer instead of the hash_variant, was added.

April 16, 2004 - Jerome

  - Added an internal optimization to cursor support to prevent
    consecutive streaming/materialization overhead when applied
    directly on a list.

    This is similar to having a built-in optimization equivalent to:
      Materialize(Stream(list)) == list

April 16, 2004 - Jerome

  - Toplevel:

      o Added a 'galax-compile' function which can be used to compile
        a query, generate the XQuery core, etc. This command never
        performs evaluation, and by default, it prints the final
        result of compilation (optimized core).

April 16, 2004 - Philippe

  - Reflection of C API functionality in Java API (ProcessingContext).
    
April 15, 2004 - Jerome

  - Re-implemented F&O to use cursors as much as possible.
  - Additional changes to F&O occurred, as part of this rewriting:

     o Removed fn:item-at, not in F&O anymore.
     o Changed op:to to return the empty sequence in case the starting
       integer is strictly greater than the ending integer, instead of
       creating a decreasing sequence.

  - Fixed the static analysis rule that computes the number of uses of
    a given variable within a query (was *completely*
    wrong!). Reinstated the full variables inlining rewriting rule
    based on this.

  - Fixed evaluation of treat as, which was a no op! Now treat as
    effectively performs dynamic type matching. In case static typing
    is on, then treat should be removed during rewriting time.


April 15, 2004 - Mary

  - Monitor
  
     o Fixed monitor so that start/end monitor calls are completed for
       API functions that raise exceptions
     o C-API. Fixed bug in galax_monitor_last_call
     o Java-API. Added functions ProcessingInstruction.setMonitorMem
       and ProcessingInstruction.setMonitorTime to initialize monitor.
     o examples/c-api,java-api Added tests for monitor last call and
       monitor all calls.

  - Cleaning

     o Cleaning_judge. Exported free_variables in .mli
     o Working on some bugs in this module

  - Other

     o Algebra_fn. Exported add_bltin_fctn in .mli

April 14, 2004 - Philippe

  - Changed two typos that made the Java API fail to compile
    (galax_wrap.ml).

April 8, 2004 - Jerome

  - Re-submitted the Export_dm module in the CVS repository (not
    updated for some reason).

April 7, 2004 - Jerome

  - Complete rewriting of the physical datamodel, evaluation engine
    and of the data model interface to introduce the use
    of cursors. (I.e., some kind of a lazy sequence).

  - Fixed export *not* to generate the string-value of nodes that do
    not have a real atomic type associated to them. This fixes a bug
    in export that would not generated the correct output stream
    (although this was never tripped), and significantly 

  - Fixed a bug in normalization for (PrimaryExpr[$v]).

  - Removed variable in-lining optimization which slows everything
    down, when passed through an iterator!!!

March 23, 2004 - Nicola

  - WSDL:

     o Temporary implementation of service deployment using 
       the C API and the Apache API. The exported module is
       fully copied in the stub and the WSDL interface must
       be provided

  - C API: fixed compilation problem (Caml monitor functions 
      returned sequences that had to be converted into lists)
     

March 22, 2004 - Mary

  - fn:normalize-string
    Aligned with F&O spec

March 16, 2004 - Jerome

  - XQuery parsing:

     o Short-term fix to bug in parsing for updates. Will have to be
       revised eventually, but should hold for now. This implements
       the exact same syntax that was in place before, but keep
       "insert", "delete", and "replace" as keywords for now.
     o Fix to the "QNAME (:" parsed as function calls bug.

March 15, 2004 - Nicola

  - WSDL -- temporary solution for service export:
	the exported module is copied at the beginning of
	the server stub

  TODO: 
     - update the apache XQuery module in order to support the
       new stub structure
     - implement WSDL generation (this will need to wait until static
       typing is fully implemented)

March 15, 2004 - Nicola

  - WSDL: added a simplified version of the xquery2soap toplevel
    application

March 14, 2004 - Nicola

  - glx:soap-call(): uses now the default processing context because
    it must always send the request serialized as well-formed XML

  - HTTP tool:
     o changed the protocol from HTTP/1.1 to HTTP/1.0, but
       still using the "Host" header for compatibility with
       some servers

  - WSDL:
     o downgraded the client stub to a form similar to what it
	was last year (i.e. no typing, no imported modules)

March 11, 2004 - Mary
  - Added projects.html to website

March 10, 2004 - Jerome & Mary

  - Cleaned-up the sequence interface. Removed unused
    operations. Preparation to support cursors. Removed some dead-wood
    code.

  TODO:
  - We still need to fix fn:lang() when there is no context node, or
    no parameter.
  - We need to make a review of the DDO code. One problem we found, is
    that is sometimes uses the fs_distinct built-in name, and
    sometimes uses the wrong (fs_uri,"distinct") variant.
  - Change the implementation of the "has_element_content" method in
    the datamodel. Currently is implemented by looking at all of the
    children, which is very expensive. We should keep a boolean in the
    data model (main memory *AND* Jungle), and use that boolean
    instead of reconstructing it.
  - Review of the last call F&O draft.

March 9, 2004 - Jerome

  - Data Model:

     o Full clean-up of support for sequence. Sequences are now opaque
       types. This is to prepare the use of sequences as cursors.
     o The abstract DM and the Galax data model interfaces only use
       sequences now, never lists. The internal representation for the
       main-memory data model is still based on lists.
     o Re-implemented axis in the Galax data model to be using
       sequences properly.
     o ancestor and descendant are not implemented based on
       ancestor-or-self, and descendant-or-self.

March 03, 2004 - Mary
  - Precedence of normalization of Expr1//Expr2 was incorrect. 
    Previously, was:
    Expr1/(desc-or-self::node()/Expr2)
    It is now: 
    (Expr1/desc-or-self::node())/Expr2

February 24, 2004 - Mary

  - Fixed compilation error in c_api/galax_wrap.ml 

February 24, 2004 - Jerome

  - A few simple fixes for Nicola and Avinash.

     o Fixed pretty-printing of sequence types to comply to the new
       syntax.
     o Fixed bug in serialization of attribute nodes. Was using the
       default prefix in the case the attribute is in the default
       element namespace, instead of an explicit prefix.

February 15, 2004 - Nicola

  - WSDL

     o partial update of the server and client stub generators Still
       unfunctional, mainly because the AST printer also needs to be
       updated.

February 15, 2004 - Jerome

  - XQuery:

     o Imposed namespace discipline to variables. (Long-standing
       mismatch with the spec). Now variables are resolved as any
       other QName. See XQuery 1.0 document, Section [Section 3.1.2
       Variable References].
     o Implemented XQuery version declaration in the prolog. Galax
       supports XQuery version "1.0" only.

  - Documentation:

     o Updated the STATUS file. Identified the various kind of
       differences with the XQuery spec more clearly. (various XML
       markup indicate whether we provide an extension, whether we
       implement only a part of the spec, whether a functionality is
       missing, or whether we implement a wrong semantics).

February 12, 2004 - Mary

  - Command-line options

    * Split control of monitor on memory and CPU
      -monitor-mem on/off
      -monitor-time on/off
      -monitor on => -monitor-mem on -monitor-time on

  - C API

    * Added functions that control processing and monitor context.
      These are analogs of functions in Processing_context and Monitor_context: 

      galax_monitor_of_last_call
      galax_monitor_of_all_calls
      galax_set_monitor_mem 
      galax_set_monitor_time

      !!NB!! Following are not yet tested:
      galax_set_normalization_phase
      galax_set_typing_phase
      galax_set_cleaning_phase
      galax_set_evaluation_phase
      galax_set_sbdo_kind
      galax_set_typing_kind
      galax_set_serialization_kind
      galax_set_serialize_ns
      galax_set_projection_kind
      galax_set_xquery_whitespace
      galax_set_xml_whitespace
      galax_set_xml_pis_and_comments

   - All API Examples

     * Updated to conform to new syntax

February 12, 2004 - Jerome

  - XQuery:

     o Added support for external variables in the XQuery prolog. Now
       external variables should be declared in the modules and Galax
       checks that such variables are bound, as required by the spec.
     o Added support for type declarations for global variables.

February 12, 2004 - Mary

  - Monitor

     o Replaced full heap collection and compaction by minor
       collection (just promotes long-lived data to major heap).
       Should speed up monitoring considerably.

February 10, 2004 - Jerome

  - XML parsing:

     o Fixed bug introduced in dealing with lexical states during SAX
       parsing.

February 10, 2004 - Mary

  - Monitor

    * Previously, monitor was reporting *total size* of heap in bytes
      instead of *number* of bytes allocated from the heap during
      call.  Memory metrics now include:

      <!ELEMENT allocated_memory #PCDATA> 
        <!-- memory allocated during
             individual call, prolog, or statement, in KB, formatted as %.3fK -->
      <!ELEMENT live_memory #PCDATA>      <!-- live memory in heap at end of call KB, formatted as %.3fK -->

  - fn:root() corrected signature to fn:root(node?) as node?

February 9, 2004 - Jerome

  *** NOTE: This version will ONLY compile on OCaml 3.07 ***

  More work on parsing.

  - Made the XML and XQuery parsers re-entrant.

     o All lexers now take a lexing context as input. This lexing
       context contains the stack for the lexical states handling, the
       parsing kind and string buffers used during parsing.
     o All global variables for lexical states, buffers, etc. have
       been removed.

  - Lexing cleanup.

     o Generalized the use of the 'as' keyword in the lexers. This
       simplifies greatly the extraction of names from
       tokens. Simplified the Lexing_util module accordingly. This
       also forces the use of OCaml 3.07.

  - XQuery parsing bug fixes.

     o Added back support for computed constructors, which had
       completely disappeared (oops).
     o Removed wrong '=' in default namespace declarations.
     o Fixed bug in lexical states for sequence types.
     o Fixed bug when using optional variables in case clauses.

February 6, 2004 - Philippe

  - Fixed two bugs in the annot_ddo algorithm:
    
     o annot_ddo.ml: predicate expressions should continue making automaton
                     transitions from the current state (not from the init-state)
     o ddo_automaton.ml: fixed error in transition table
    (Bug found by Jerome)

February 5, 2004 - Mary

  - In Cleaning_rules and Procmod : 

     o Partitioned rewrite rules into non-overlapping sets of :
       Generic rules that are independent of other optimization & typing options 
       Rules that are applicable when either weak or strong static typing is enabled
       Rules that require strong static typing 
       Rule for OLD SBDO algorithm 
       Rule for removing ALL SBDO operations
       Rules for new Automata DDO algorithm 

     o Rules are combined in Procmod based on command-line options.

Feburary 1, 2004 - Jerome

 ** WARNING: Major syntax change to align with the November 2003, last
    call XQuery WDs. **

 ** Usecases all parse and evaluate properly. More fine-tuning and
    testing for lexing/parsing/pretty-printing is still required
    though. **

   - XQuery: Major upgrade to the XQuery parser, and alignment with
     the XQuery last call syntax.

      o Major reorganization of the XQuery parser and lexer(s). Split
        the lexers in smaller components based on lexical
        states. Aligned (with a few optimizations) to the lexical
        states as specified in the XQuery document, Appendix A.
      o Removed the need for reserved keywords, using a combination of
        long tokens and appropriate lexical states. Removed the syntax
        for escaped NCNAME (i.e., :item, or :person are no longer
        valid).
      o The following QNames are not allowed as function names:
        "comment", "text", "node", "processing-instruction", "if" and
        "typeswitch". See XQuery last call draft, Appendix [A.3
        Reserved Function Names].
      o Changed the syntax for declarations: now declarations are
        always terminated by the separator ';', and can appear in any
        order.
      o Removed the 'isnot' operator.
      o Fixed a bug in the parsing of external functions. Changed the
        syntax of signature to force the use of variables in the input
        parameters.
      o Fixed long-standing lexing bug: now function
        calls/declarations can have whitespace between the function
        name and the parenthesis. I.e.,: local:f () is supported.
      o Aligned existing sequence type to the new syntax (without
        extensions):
          empty --> empty()
          node  --> node()
          text  --> text()
          attribute --> attribute()
          element --> element()
          document -> document-node()
          element a --> element(a)
          element of type T --> element(*,T)
          attribute a --> attribute(a)
          attribute of type T --> attribute(*,T)
      o Removed obsolete 'sort by' syntax, which we had left around.
      o Updated the syntax for 'treat as', and 'cast as':
          treat as SequenceType (Expr) --> Expr treat as SequenceType
          cast as SequenceType (Expr) --> Expr cast as SequenceType

January 29, 2004 - Jerome

   - Main memory datamodel:

      o Renamed modules for consistency:
            Datamodel 	   --> Galax_dm
            Load      	   --> Galax_load
            Load_context   --> Galax_load_context
            Galax_copynode --> Galax_node_copy

   - XQuery:

      o Changed syntax of function declarations:
          define function --> declare function 
      o Switch to the new syntax for external functions (used in
        ./stdlib/pervasives.xqp only for now).
          declare function signature... --> declare function ... external

   - Namespaces:

      o Changed namespace environment behavior to keep track of
        default namespace rebindings.

   - F&O:

      o Implemented fn:deep-equal(), notably used in use case 'XMP'.

   - Demo:

      o Fixed the demo to work with the new Galax API, and processing
        model.

January 26, 2004 - Jerome

   - XML Schema:

      o Fixed bug in lookup of global element with substitution
        groups during validation.

   - XQuery:

      o Aligned global variable syntax to the new XQuery syntax:
          declare global --> define variable
      o Removed obsolete single line comment starting with '#'.

   - Namespaces:

      o Fixed bug in dealing with empty namespaces: now raises an
        error when binding an explicit prefix to the empty namespace,
        and reset the default namespace to the empty namespace. See
        Namespace REC [2. Declaring Namespaces]. [ Bug courtesy of
        Irini Fundulaki ]

   - F&O:

      o Implemented the fn:number built-in function. [Request from Jan
        Hidders.]

   - Tools:

      o Removed pxp_lexers module, not used.

January 23, 2004 - Philippe

    - Monitor:

        o Adjusted the monitor to use real time instead of user+system
          time.

January 20, 2004 - Jerome

   - Documentation:

      o Added Philippe and Rob in the list of contributors.

   - XQuery syntax:

      o Celebrated the start of the big syntax upgrade by switching
        from {-- --} to (: :) :)!

January 19, 2004 - Jerome

   - Jungle:

      o Fixed the compilation process to make sure Jungle can be
        compiled with the C and Java APIs, and provide additional
        flexibility when using Galax with optional Caml-C tools
        support.
      o Fixed bug in the generation of the prefix/uri namespace
        bindings in the persistant store.

   - Namespaces:

      o Separated various default set of namespaces for: XML, XML
        Schema and XQuery. This is a first shy step toward a more
        rational support for namespaces.
      o Put back Mary's fix for the initialization of the uri/prefix
        global table (for now).
      o Fixed insertion of uri/prefix binding to only register the
        first prefix for a given uri.

      ** Hope this is not going to create problems elswhere... **

   - XML Schema:

      o Hooked up Vladimir's import/normalization code to the
        "galax-parse" command-line tool. Validation can be called as
        follows:

        galax-parse document.xml -validate -xmlschema schema.xsd

      o Fixed problem in default namespaces for XML Schema import and
        normalization. Now uses the 'default_xmlschema_nsenv ()'
        function.
      o Replaced "galax-uri-for-anonymous-types" namespace by the
        built-in glx: namespace.

      ** Vladimir: please double check my changes there. **
      ** Still more debugging is required **
      ** Better error messages are required **

January 15, 2004 - Vladimir

   - XML Schema:

      o Beginnings of Schema normalization (from imported Schema
        syntax to XQuery named core syntax) in schema/schema_norm.

January 14, 2004 - Mary

   - Namespace: 
      o Added all built-in prefix/uri pairs to global prefix/uri table
        to fix bug in printing of normalized expressions.  

January 14, 2004 - Jerome

   - XML Schema:

       o Added back type with Bound+minOccurs+maxOccurs as part of the
         core type system/normalized schema. Propagated the change in
         the validation code.

January 13, 2004 - Jerome

   - XML Schema:

       o Fixed bug in computation of the root content model in the
         presence of substitution groups.
       o Added some support for the derived built-in atomic
         types. (e.g., xs:short, xs:token). There is no support for
         facets, just the structural and named part is supported.
       o Made all built-in type names visible in Datatype_util.
       o Added a module that contains the built-in XML Schema types.
         This module defines all primitive and derived atomic type,
         the Ur types, and the built-in attributes.

January 13 2004 - Avinash

   - Jungle:
       o Commited changes to smalldb_stubs.c which somehow
         did not get reflected in cvs repository.
       o Modified the loading of xml document in Jungle with
         the use of nodeid_context. 

January 12, 2003 - Jerome

   - Makefiles:

       o Fixed minor compilation problems with the new ./procctxt
         directory. (missing clean and depend rules).

   - XML Schema:

       o Fixed several bugs in validation of attributes, related to
         transitions in the attribute part of the content model.
       o Fixed bug in interpretation of list and union simple
         types. Now works when directly referencing built-in atomic
         types.
       o Fixed bug in lookup for substitution group (was doing it the
         wrong way around!).

   - SAX parsing:

       o Added an early check for duplicate attributes. This occurs
         when resolving the stream, as it is only then that you can
         know whether two attributes are really the same.

January 12, 2004 - Mary

  - API and API Examples

    o Fixed several bugs in Galax API

    o Aligned C and Java API with new Galax API 

    o Updated Caml, C, and Java examples in examples/*_api to be
      aligned with new Galax API that uses Galax_io input and output
      specs. 

January 11, 2004 - Mary

  - Monitor 
    o monitor_of_last_call yields document of element call
      monitor_of_all_calls yields document of element monitor

    o Use "-serialize wf" to print monitor output as well-formed document

  - Processing_context
    o Moved processing_context from base/ to new procctxt/ directory

    o Cache of loaded documents is maintained in processing_context 
      Global document cache _no longer_ exists
  
    o _All document-load functions (e.g., fn:doc(),
      Galax.load_document) use document cache in processing_context

January 10, 2004 - Jerome

   - XML Schema:

       o Fixed bug in computation of the content model transitions for
         sequence and interleaving.
       o Allows some of the non-well-formed error messages to filter
         up through validation (was catching exceptions a bit to
         aggressively before).

January 8, 2004 - Mary

  - Monitor 

    Re-structured the monitor module as follows: 

    o Removed monitor flags from Conf module

    o Added Monitor_context to base/ ; monitor_context is maintained
      in processing_context 

    o Added Monitor module interface to Galax user API
      (in Galax/lib/caml)

    o Monitor output corresponds to calls to Galax user API.
      See documentation in Monitor.{mli,ml} for details

January 8, 2004 - Jerome

   - XML Schema:

       o Improved the error message when element has xsi:nil set to
         true, the element declaration is nillable, but the element is
         not empty.
       o Applied minor changes to the Core Type AST, suggested by
         Vladimir.

January 7, 2004 - Jerome

   - XML Schema:

       o First version of XML Schema validation works! *** STILL
         TESTING! ***

January 6, 2004 - Jerome and Avinash

   - Node identity:
       o Added support for a nodeid_context, which can be used to
         parameterize the way node identifiers are created during
         loading.
       o Removed post order from the Galax main memory data model. It
         is not used, and we do not need to maintain it.
       o min and max pre/post order are now defined in the Nodeid
         module.

   - Updates:
       o Started folding back support for updates in the new Galax (at
         last!). This only concerns the main-memory data model for
         now. Insert and delete are (partially) supported again!
       o Dispatched updates methods at the right location in the Galax
         data model.

January 5, 2004 - Vladimir

   - Schema_import: 
       o added function import_embedded that takes a namespace
         environment as an additional argument; this required adding:
   - Namespace:
       o added function active_bindings: nsenv -> delta_bindings

January 5, 2004 - Avinash

   - Jungle :
       o Modified the nodeid to exclude post order. The nodeid is
         just the (docid, preorder) pair.
       o Increased the cache used by BerkeleyDB. 

January 3, 2004 - Nicola

   - Wsdl import and export:
       o added support for typing, needs changes in the schema import 
	 mechanisms to make it work for all cases

December 29, 2003 - Jerome

   - Node identity:
       o Fixed wrongly commited change to nodeid.ml.

December 29, 2003 - Jerome

   - Bug fixes:
       o Now works with local file names that contain ':'.
       o Built-in simple types now behaves corrrectly with namespace
         resolution (except in the old type system syntax still --
         shall be fixed when switching to the new type system).

   - XML Schema:
       o Added set of new modules for XML Schema validation. Still
         working on them. They are not tied to the rest of the system
         yet.

   - Node identity:
       o Added support for node identity based on a single Pre order
         of various kinds (integer, string, and float). This should be
         useful for both Jungle and the LDAP wrapper.

December 28, 2003 - Nicola

   - Wsdl:
       o Reverted Wsdl_load to Ocamlp4o syntax.
	 Replaced xml_streams with resolved_xml_streams and old namespace
	 environment management with standard namespace env. management based
	 on resolved streams.
	 Added a function in schema_import to be able to load a schema from 
	 an XML element.	 

December 26, 2003 - Avinash

   - Jungle:
       o Removed a bug in the stub, in reading through cursor

December 26, 2003 - Avinash

   - Jungle:
       o Added Jungle_encode module for encoding integer as string of
         characters, creating a fixed 4 byte, module 256 encoding.
       o Modified the IDL and stub to allow storing the char
         encoding of numbers. Currently few parts of the stub 
         are hand coded, i.e. to convert a c string to caml string.
       o IDL is also modified to compute the length of the string
         automatically within the stub instead of using strlen in C.
       o Modified the main record structure to use the above said 
         encoding thus creating fix length records. All other indexes
         also stores integer in the new encoding.

December 18, 2003 - Mary

   - Fixed typesafe_call_rewrite to be more efficient.

December 18, 2003 - Avinash

   - Jungle:
       o Fixed a bug in Jungle.sync method.
       o Implemented a new set of Indexes for Jungle. Now Jungle has
         a seperate index for pre_post, attributes and childrens. Post
         order is no longer stored in the main records. 
       o Children and attribute axis now uses these new indexes.

December 17, 2003 - Nicola

   - Wsdl:
       o Added support for the server interface of an exported module 
	 in ./wsdl/wsdl_apache.ml
	 It is not fully integrated with the Apache API.

December 15, 2003 - Jerome and Avinash

   - Validation:
       o Renamed Validation-sax to validation. Started implementation
         of SAX-based validation.

   - Jungle:
       o Fixed a couple of problems with built-in support in Jungle.
       o Moved implementation of descendant and ancestor axis from the
         generic DM to each specific implementations. Implementation
         is still based on recursive calls to child::, but now can
         easily be changed by each implementation.

December 15, 2003 - Mary

  - Error
    o Added generic Static_Error to cover all unspecific static 
      errors. 

  - Semantics of functions calls

    o Namespaces:
  
     (1) function declarations must have a prefix.
     (2) user function declarations must not be in one of the built-in
         namespaces, except the local namespace.
     (3) default default function namespace is fn
  
      [TODO: static name check for function decls in library modules,
      when modules are implemented]
    
    o Architecture:
    
     (5) fn_static moved back to ./stdlib
    
    o Compilation:
    
     (6) The AST for non-overloaded function *calls* in the core includes
         the signature for the function, in which each type parameter is
         optional.
  
     ***  NB: Use Norm_util.fmkCECall to construct _all_ function calls! ***
  
         Overloaded function calls unchanged as they are resolved to
         non-overloaded calls by applying type matching to arguments 
         dynamically.
    
     (6a) The AST for function bodies in the core includes the return
          type so result of function body can be type matched against
          return type. 
    
     (7) Normalization now uses function calls with in-line signatures
    
     (8) Normalization looks up the type signature for the function and
         inlines it at the location of the function call.
    
     (9) Type checking of non-overloaded function calls uses inlined
         function signature
    
         (The inlined signature is only applied to non-overloaded function
          calls; dispatching and typing of overloaded function calls still
          depend on the static context.)
    
     (10) Optimization of non-overloaded function call removes dynamic
          type matches on arguments
    
     (11) Dynamic semantics for non-overloaded function calls uses
          the in-line signature for type matching.

December 12, 2003 - Jerome

   - Extended fn:doc() to support access to back-end stores (e.g.,
     Jungle, PADS)

       o Added registration of new URI methods (e.g., "jungle:...",
         "pads:...") in Glx_http.
       o Renamed ./stdlib/store to ./stdlib/fn_doc
       o Moved most of the existing fn:doc code into ./stdlin/fn_doc
       o Implemented support to register access to a back-end from a
         given URI, and hooked this access to the fn:doc() function.
       o Added registration for Jungle in ./jungle/jungle_register.ml
       o Phased data model linking as: Galax native data model,
         followed by stdlib, followed by external data models (e.g.,
         Jungle, PADS).
    
     ** Note: registration for PADS should be added. **

   - Bug fixes:

       o Fixed long standing (about 1 year old!) bug that Galax would
         not run when installed in a directory containing a whitespace
         name on Win32, due to a failure when building the default
         base URI.

- 11 December 2003, Avinash

    + Added child index (btree) in Jungle to speed up
      navigation along child axis i.e computation of
      children .

December 11, 2003 - Avinash

   - Added child index (btree) in Jungle to speed up navigation along
     child axis i.e computation of children .

December 10, 2003 - Vladimir

  - Added ast/xquery_type_ast_core_named.mli
  - Minor clean-up: removed schema-related functions from top_util.ml, 
    in-lining their code in galax-mapschema.ml.

December 10, 2003 - Vladimir

   - Added ast/xquery_type_ast_core_named.mli

December 10, 2003 - Roel

   - Replaced the previous DDO automata (which stopped as soon as the
     intermediate result was not in document-order or there were duplicates)
     with a new automaton that will have intermediate results always in
     document order without duplicates (i.e. only use necessary DDO-operations).
     Using this approach we can remove more DDO-operations than with the old
     automata.

December 9, 2003 - Avinash Vyas

    + Modified Jungle to solve the namespace problem.
      Jungle now stores the uri-prefix mapping and builds
      the mapping inside Galax's in-memory Hashtables when
      a document from store is opened.

	+ Added a function to prefix_to_string in namespace module.


December 5, 2003 - Jerome

  - SBDO:

     o Fixed bug on the semantics of |, union, except, and intersect, now
       sorting by document order properly.
     o Added support for complete removal of SBDO operations.

  - Clean up of command-line parameters:

     o Turned the SBDO command-line parameters as one parameter with
       multiple options:

         -sbdo remove, preserve, adhoc, or automata (new)

         where preserve means : keep all SBDO operations
               remove means : remove all SBDO operations
               adhoc means : the old ad hoc algorithm
               automata means : the new automata-based algorithm

     o Removed the following global variables from ./base/conf:
         mapping, typing, optim, eval, sbdo, ddo, weak,
         serialize_kind, serialize_ns, serialize_xsitype,
         serialize_id, projection, optimized_projection,
         xquery_whitespace, xml_whitespace, xml_pis_and_comments,
         stdlib.

     o Removed usage of global variable 'typing' in the cleaning
       processing phase. This required access to the static_context
       notably in cleaning_judge.can_fail judgment.

     o Added an operation to get the processing context from the
       module context in the API.

     o Moved serialization behavior parameters in
       Processing_context. All functions in serialization are now
       parameterized by the Processing context. This propagates
       notably to the API, where serialization operations take the
       processing context as input.

     o Added support to set-up arguments from the command lines in the
       processing context in Top_config and Processing_context.

     o check_config now takes the processing context as input. Moved
       it to Processing_context.

     o Now always building the default processing context and passing
       it to process_args to set up the right context from the command
       line.

  - Processing model:

     o Fixed bug in serialization of the result, not printing the
       correct headers in verbose mode.

December 4, 2003 - Jerome and Avinash

  - Jungle:

      o Integrated and fixed the jungle-load utility written by
        Avinash. This allows to populate a Jungle store from the
        command line directly.
      o Made the Namespace.uri_prefix_table visible. Necessary to
        recover prefixes in Jungle.

  - Toplevel:

      o Split Toputils into Top_util and Top_config. Top_config should
        contain all the operations necessary to perform the
        configuration of Galax from the command line. Note that
        currently a lot of this is still in ./base/conf

December 1, 2003 - Jerome

  - Bug fixed:

     o Fixed XML parsing bug for documents using the '<!DOCTYPE doc
       PUBLIC ...>' kind of DTD declaration. [ Bug report courtesy of
       Dan Suciu ]

November 28, 2003 - Jerome

  - Support for configurable Input/Ouput.
      o Added a notion of I/O for Galax. Input currently supported
        are: string buffers, local files, URL's for both local files
        and http, and Caml input channels. Ouput currently supported
        are string buffers, local files, Caml output channels, or Caml
        formatters.
      o Created Galax_io module which contains the I/O types.
      o Cleaned up Glx_pxp: Moved all operations on character
        encodings from Glx_pxp to Encoding.
      o Renamed Glx_pxp to Io, contains now internal operations on
        Galax_io types.
      o Reimplemented toplevel parsing operations in terms of IO,
        avoiding duplicationg (_from_string, _from_file, functions are
        now covered in one _from_io function and also supports http
        now).
      o Reimplemented toplevel SAX parsing operations in terms of IO,
        avoiding duplicationg (_from_string, _from_file, and _from_uri
        functions are now covered in one _from_io function).
      o Reimplemented toplevel Galax API operations (Caml side) in
        terms of IO, avoiding duplicating (_from_string, _from_file,
        and _from_uri functions are now covered in one function).
      o Pulled calls to the 'escaping' function used for the web demo
        from the serialization call in Galapi, and left it at the demo
        level.

  - Processing model and parsing:
      o Added parsing as a processing phase. Parsing operations
        operate over Galax Inputs (see above).
      o Propagated the change to Galapi, which does not directly
        call low-level parsing functions anymore.
      o Separated parsing utilities (in Parse_util) from top-level
        parsing functions (in Parse_top). Removed Parseutil.
      o Moved 'parsing' of the pervasive module from toputil to
        Parse_top.
      o Removed dead wood in the SAX parser.

November 19, 2003 - Nicola

  - wsdl import/export
    o merged SOAP Envelope generation subroutines for WSDL
      import and export

November 18, 2003 - Mary
  - Updated galax/usecases directory with bug fixes to XMark

  - Optimization

    o cleaning_rules.ml:

      - Fixed bug that triggered infinite rewriting of some- and
        every- expressions.

      - Disabled the rewriting rules for eliminating functions that
      convert xdt:untypedAtomic values to a target type, because they
      depend on the soundness of static typing and validation, i.e.,
      on validation having correctly annotated data-model values with
      their types, but validation is not yet implemented completely.

      These rules should be re-enabled when named typing and
      validation are implemented.

    o rewrite_context.mli

      - Added 'a printer parameter so that 'a cexprs can be printed
        (for debugging) from within generic rewriter module.

  - Command-line options

    o -ddo on => generic rules + ddo 

November 18, 2003 - Mary

  - Changed command-line options to be mutually exclusive

    Default (i.e., no options) => -optim on -sbdo on 
    -optim on -sbdo off        => generic rules only
    -optim on -sbdo on         => generic rules + sbdo 
    -ddo on                    => ddo only (for now)
    
November 18, 2003 - Roel
  - Cleaning:
    o Fixed Not_Found exception: for now we consider namespace and attribute
      axes to behave in the same way as a child axis.

November 17, 2003 - Mary
  - Typing 
    o simplification.ml : 

    Change to elim_subtype_duplicates to remove duplicates of element
    references _before_ applying Subsumption.isSubtype, which does not
    terminate if comparing the same recursive type to itself.

    This problem arises when simplifying unions in which the same
    recursive type may appear multiple times (e.g., for desc-or-self).

    TODO: Jerome and Mary will find a permanent solution.

    o scheamutil.ml : 

    The definition of recfactor_aux was changed to yield more compact
    unions, which _dramatically_ speeds up type simplification.
    Please see long explanations in schemautil.ml
 
November 14, 2003 - Mary

  - Updated galax/usecases from XQueryUnit to align with new version

  - Moved Fn_static from stdlib/ to schema/, because normalization
    depends on info about functions with special typing rules.

November 14, 2003 - Mary
 
  - Changed typing.ml so that element constructors pass static typing.

    This is a temporary fix until element construction is implemented
    completely and correctly.  See comments with label **Element
    Construction HACK** in eval_expr.ml and typing.ml.

November 14, 2003 - Mary
  - Fixed problem in normalization of functions with special typing
    rules; should not use generic let-type assertions.

November 13, 2003 - Mary

  - Command-line switches

    o -ddo on  always implies -static on (for now) and -sbdo off
    o -sbdo on always implies -ddo off
    
  - Monitoring and Printing

    o Fixed bug in printing document-processing times (was printing
      absolute instead of relative);

    o Moved most of monitoring and printing of intermediate results
      into Procmod_phases

    o Monitoring and printing of Projection and Serialization remain
      in Galax API, but should be moved as well. 

  - Toplevel
    o Moved several common functions in Toputils

November 13, 2003 - Philippe
 - Cleaning
   o Added missing constructor in ddo-pattern matching
   o Added support for variable-cardinality optimization (annot_ddo.ml)
     Note: this optimization will work best with static typing turned on
     
November 13, 2003 - Philippe and Roel

  - Cleaning

    o Fixed bug in accept states of DDO-automaton


November 12, 2003 - Mary

  - Typeutil & Schemautil

    o Added max1_property to typeutil.mli

    o Moved following functions from schemautil to typeutil so that
      their result types could be simplified (Typeutil depends on
      Simplification module, which depends on Schemautil):

      apply_to_unit_types
      extract_content_model
      extract_attribute_model

      (We are not consistent about which type functions simplify
       before returning a result type -- we should revisit this issue
       when updating named typing)

  - Normalization

    o Fixed several bugs (using CECall to call overloaded functions)
    

November 12, 2003 - Jerome

  - Several fixes to the DDO support.

    o Fixed bug in the rewriting using glx:distinct-docorder instead
      of fs:distinct-docorder.
    o Removed dependency of ddo optim on static typing.
    o Added interface to the Ddo_automaton.
    o Fixed a few stylistic problem, and added the standard boiler
      plate at the beginning of modules.

November 10, 2003 - Mary

  - Printing annotations

    o Changed -print-type-annotations switch to -print-annotations 
      (default is false)
    o Changed Print_xquery_core.print_expr to conditionally print any annotation

November 9, 2003 - Mary

  More changes to align w/ Nov'03 Use Cases

  - Parser

    o Changed precedence of << and >> to be same as other infix
      (in)equality operators
      (Tripped bug in TREE use case)

  - Fixed bugs in serialize_float and _string_of_decimal

    Implemented serialization rules for floats & decimals in F&O
    Section 17.7.

November 9, 2003 - Nicola

  - WSDL

    o Updated wsdl stub and skeleton generation to respect the last
      changes in namespace declaration.
    o Fixed some errors in treating document style definitions.
      However, the namespace environment is still not implemented
      properly and predefined soap:faults are not integrated with
      the Apache module.
  
November 7, 2003 - Mary

  - Changed type semantics for fn:sum,... again.
    Default 2nd arg is now xsd:integer

  - examples/

    Fixed all local function def'ns and calls to be in local:
    namespace.

November 7, 2003 - Philippe and Roel
 
  - Cleaning

    o Added the distinct-docorder optimization algorithm (annot_ddo.ml*)
    o Added DDO-automaton implementation (ddo_automaton.ml)
    o Added the cleaning_rule for the DDO-optimization
    
  - Procmod

    o Procmod_phases.ml: added toplevel ruleset for application of
      ddo-optimization
      
  - General

    o Added entry for ddo_automaton in Makefile
 
 November 7, 2003 - Philippe and Roel

  - Normalization

    o Fixed the normalization rule for the slash expression, removed
      redundant insertion of distinct-docorder operation.


***** Galax Release 0.3.5 November 6, 2003 *****

 November 5, 2003 - Mary

  - Parser

    o Removed "end" as keyword -- no longer needed

  - Usecases

    o Aligned with Nov 14 WD 
      (Still working on this; Define deep-equal())
    o Regenerated files in usecases/

  - Datamodel

    o Fixed typed-value() of atttribute and text nodes to return
      untypedAtomic, not string.

  - Normalization

    o Fixed normalization of absolute path expressions (/Expr and
     //Expr) to be left associative. 

  - General

    o Added "local" namespace prefix

November 5, 2003 - Jerome

  - Fixes to serialization:

     o Fixed printing in XQuery syntax for atomic values. Now prints a
       valid XQuery expression with the appropriate type.

November 4, 2003 - Mary

  - F&O 

    o Added back fn:document() as synonym for fn:doc()

November 3, 2003 - Jerome

  - Fixes to pretty-printing:

     o Fixed bug on pretty-printing of enclosed expressions (was
       printing '{' '}' twice).
     o Fixed bug on pretty-printing of attribute content (was not
       dealing with multiple enclosed expressions properly. Now
       switching between enclosed expression and text properly).
     o Fixed bug in pretty-printing of path expressions ('/' is
       printed as left-associative, i.e., a/(b/c) != (a/b)/c = a/b/c).

November 2, 2003 - Mary

  - cleaning/cleaning_rules.ml
  
    o Fixed bug in which type-dependent rewrite rules were
      (incorrectly) applied after weak typing

November 1, 2003 - Jerome

  - Fixes to the Web site.

     o Upgraded to the new API.
     o Fixed bug in ./netstring tool under Ocaml 3.07 (HTML encoding
       did not work anymore due to changes to Str).
     o Added new entry for CVS-Web server.
     o Added Vladimir as a contributor.
     o Added install-front target to allow installation of the Web
       site, leaving the demo alone.

October 31, 2003 - Mary

  - datatypes/datatypes_lexer.mll

     o Fixed bug in which whitespace was not ignored when casting from
       untypedAtomic to any target atomic type.

  - Top-level rewrite_* and cleaning_*  functions now take two sets of
    rules: 
 
     o First set is applied only to the top-level AST node in the
       input expr/statement.

     o Second set is applied recursively to every AST node in the input
       expr/statement.

    See procmod_phases for initialization of rule sets. 

  - F&O Last Call Alignment 

    Removed:
      fn:context-item
      fn:node-kind
      fn:sequence-node-identical

    Renamed:
      fn:string-pad     => glx:string-pad 
      fn:distinct-nodes => fs:distinct
      fn:document       => fn:doc

    Added:   
      fn:abs 
      fn:reverse

    Changed: 
      fn:floor, fn:ceiling, fn:round to be polymorphic
      fn:sum to take optional 2nd argument

  - normalization/norm_util.ml

     o Removed conditional normalization of SBDO operations

       All SBDO operations as defined in FS are in normalized exprs.
       They may be removed by rewritings

  - cleaning/cleaning_rules.ml

     o Added type-based rewriting rules for following special
       functions:

       fs:untyped-to-{double,string,integer}
       fs:untyped-to-any
       fs:promote-to-numeric
       fs:convert-to-simple-operand

     o Added rule to convert overloaded function calls to
       non-overloaded calls

  - typing/typing_call.ml

     o Substantially simplified static typing of non-overloaded
       function calls.  Removed unnessary promotion rules

October 30, 2003 - Jerome

  - Bug fixes:

     o Fixed bug in API not serializing the namespace
       declarations. Added corresponding tests in the API examples.
     o XQuery element construction should not keep the xmlns:
       namespace declarations at the data model level. (Fix located in
       Xquery_ast_util.get_ns_attributes used during normalization).

October 29, 2003 - Jerome

** Warning, the Galax API has changed. People using it will have to
   upgrade their code to the new interfaces and library/header file
   names. **

The modules which are visible in the API are:

  Galax        	     (* Main interface *)
  Dm           	     (* Object-Oriented Datamodel *)
  Dm_functions 	     (* Short-cut functions on the data model *)
  Processing_context (* The basic processing context *)
  Namespace    	     (* For the namespace environment *)
  Error        	     (* For Galax errors *)
  Fn_error     	     (* For user-defined errors *)

  More work is probably needed to avoid having too much extra stuff
  available in Namespace,

  - Galapi: Serious consolidation of the API's

     o Removed duplicate glx module from the caml_api. Used part of it
       for the c_api stub.
     o Renamed all libraries to Galax (in Caml, C, and Java), instead
       of glx. Renamed all calls to use galax instead of glx in the C
       API.
     o Wrote a dm_functions module that provide functional operations
       on the data model, and is reflected in the C API.
     o Removed the load_document_validate, not used right now.
     o Separated datatypes definitions (in Datatypes) from operations
       on them (Datatypes_util).

 - Apache tool:

     o Fixed wrong directory in main Makefile.
     o Fixed library names (glx to galax) in conf/Makefile_apache.

October 28, 2003 - Nicola

  - Modifications in the wsdl stub construction

    o Wsdl_ast and Wsdl_load were adapted to recognize
      the whole SOAP binding

    o Wsdl_import changed to respect the semantics defined 
      in the paper draft

    o Fixed several bugs (e.g. generated variable names must
      be different from XQuery keywords) 

  - Added a new tool: apache

    o It implements an apache module for XQuery SOAP servers
      (see INSTALL file)

    o It is only experimental and it will probably work only
      in a limited number of cases.

    o Like for the client stubs, some implementations issues
      cannot be addressed without using strong typing and 
      schema import.

    o Modified the Makefile and config/Makefile


October 24, 2003 - Jerome

  - Test for the new CVS server.

October 23, 2003 - Jerome

  - Processing model:

    o More cleaning up of Galapi and Procmod. Notably all operations
      on expr and update have disappeared to use operations on
      statements only.

October 22, 2003 - Mary

  - Printing 'a cexpr:

    o Changed Print_xquery_core.print_cexpr to be polymorphic
      Now takes function that prints 'a annotation
  
    o In Print_top, added print functions for ucexpr and tcexpr

  - Bug fix: Added prototype value for anyURI values. 

October 22, 2003 - Nicola

  - Added declaration of glx_load_document_from_string()
    in galapi/glx.ml

October 21, 2003 - Mary

  - New normalization & static typing of path expressions

    o Added normalization & static typing for [1], [last()], and
      [NumericLiteral]

October 19, 2003 - Mary

  - New normalization of function calls:

    o Alignment of overloaded & non-overloaded function calls with
      formal semantics

    o Removed all typeswitches from normalization of arithmetic, node,
      general-comp, and value-comp operators.

    o CEApp replaced by CECall (for non-overloaded functions) and
      CEOverloadedCall (for overloaded functions).  

    o Implemented dynamic semantics of following FS functions, used in
      normalization of function calls:

      fs:convert-simple-operand 
      fs:untyped-to-{integer,double,string,any}
      fs:promote-to-numeric

    o Added namespace/funcs_and_ops -- contains mneumonic names for
      all functions & operators.  No more hard-coding of function or
      op names in code!!  Just say NO!

  - Aligned signatures of following F&O functions:

    o fn:resolve-uri(xsd:string) as xsd:string 
      fn:resolve-uri(xsd:string, xsd:string) as xsd:string
 
October 18, 2003 - Nicola

  - Bug fixes in the WSDL parser
      o If certain "basic" elements such as "part", "address", etc 
	contained subelements, the parser declared that the document
	was invalid. The correct behaviour is to ignore them, even 
	if they cannot be interpreted.
  - Corrected typing error in Galax.java, function 
	nativeLoadDocumentFromString()

October 17, 2003 - Jerome

  - Simple fixes for Nicola and Avinash.

      o Fixed bug with the built-in "empty:" prefix, necessary for
        Avinash and XML Schema import. I suspect it will also be
        useful for WSDL import...
      o Added load_document_from_string function in the CAML, C and
        Java API.

October 14, 2003 - Jerome

  - Fixes to the processing model:

      o Galapi now calls the right evaluation function for the query
        prolog, taking keys definitions into account (fixes bug to
        key/keyref support).
      o Made updates to return a value (empty sequence) as
        well. Propagated that change to Galapi, and the C API.

  - Fixed Galax to work with ocaml 3.07:

      o Upgraded ./tools/pxp-engine to use pxp-1.1.6, which use the
        standard ocaml lexer instead of the previous pxp-specific
        lexer (was not compiling on ocaml 3.07).

September 23, 2003 - Mary

  - Path expressions

    o Changed '/' and '//' to be left-associative
      !!NB!! Current optimization breaks associativity.
      Disable if query dependent on associativity of / or //.

    o Added following-sibling and preceding-sibling axes. 
    o Waiting on following and preceding axes until we have
      an efficient evaluation strategy

September 23, 2003 - Mary

  - Re-synchronization changes.  

    o Fixed exceptions in algebra/algebra_fn 
    o Fixed normalization of computed text nodes in
      normalization/norm_top.ml
    o Wrapped all errors in examples/c_api/c-test.c 

September 19, 2003 - Nicola
  - Wsdl:
     o Added Wsdl_util which provides generic functions for processing 
       WSDL ASTs.
     o Changed Wsdl_import and added Wsdl_export that generates 
       server stubs from WSDL ASTs.

September 19, 2003 - Jerome

  - Processing model:

     o Further clean-up of the processing model.
     o Renamed "rewriting" phase into "cleaning" phase.
     o Added normalization, evaluation as part of the processing
       model.
     o Separated the 'processing context' to be available right from
       the start in ./base. The 'processing context' is now added into
       the normalization context.

 - Optimization:

     o Duplicated the existing Galax core to serve as an algebra. (see
       ./ast/xquery_algebra_ast.mli)
     o Added a compilation phase from the XQuery core to the algebra.

 - Evaluation:

     o Changed the implementation of SBDO to first sort, then remove
       duplicates in the sorted list, which is much more
       efficient. [Change suggested by Philippe]


September 17, 2003 - Jerome

  - Major update to: normalization, static typing, and function
    calls. Still more to come.

*** WARNING: STILL VERY UNSTABLE! ALMOST NO TEST PERFORMED ***

September 15, 2003 - Avinash

  - Jungle:

     o Added support for sync function to flush the data from memory
       onto stable/permanent storage.

August 28, 2003 - Jerome

  - Misc bug fixes:

      o Fixed serialization of namespace declarations in top-level
        elements when needed. [Bug reported by Peter]
      o Fixed bug in serialization of namespaces declarations in the
        XQuery syntax. [Bug reported by Peter]
      o Fixed bug in new printing of expanded names. Was printing
        {uri}:ncname instead of {uri}ncname. [Bug reported by Avinash]
      o Added uri/prefix binding table hacks for Peter:
          - [clear_prefix_uri_table] clears the global uri/prefix
            table
          - [init_prefix_uri_table] re-initializes the global
            uri/prefix table with some given bindings.
      o Put xml:lang back, forgotten in the new data model. [Bug
        reported by Peter]
      o Fixed bug in stand-alone attribute / text constructors. Now
        use the Enclosed expression and generates the right whitespace
        between atomic values.
      o Fixed bug in construction of elements, allowing to contain
        document nodes, introduced when building the new data model.
      o Fixed bug in construction of elements, allowing to contain
        attribute nodes in the middle, introduced when building the
        new data model.
      o Fixed bug in fn:name, was not working for QNames in the
        default element namespace.
      o Fixed bug in website makefile, , should have used $(MAKE)
        instead of make.

August 26, 2003 - Avinash

  - Jungle now stores attribute values (reused the db for storing
	text) which was missing in earlier scheme. 
  - Added partial support for persistent jungle store. You could 
    open a existing jungle store, provided you closed it before
	exiting. (requires more work though).

August 25, 2003 - Jerome

  - Processing model. Deep clean-up of the processing model support:

     o Removed polymorphism at boundaries between processing
       places. Those were introduced when adding polymorphic type
       annotations to the core. Notably its removed from:
          The processing model modules
          Galapi
          Dynamic_context

  - AST.

     o Moved specific annotated AST's in the ./ast directory.

  - Rewriting. Reorganized the rewriting phase.

     o Separated the 'generic' rewriting code (into ./rewriting) from
       the specific cleaning rulse (into ./cleaning).
     o Rewriting supports two families of operations:
        -- rewritings which go from 'a cexpr to 'a cexpr. There must
       be the same type annotation as input and output so that fix
       point can be applied.
        -- annotations which go from 'a cexpr to 'b cexpr. This does
       not change the structure of the AST, but can change the
       annotation.

  - Typing.

     o Added Notyping module, which annotate the core with xs:anyType,
       and can be used when static typing is off.

August 22, 2003 - Avinash

  - Children and Parent evaluation working now in Jungle
  - Added support for getting attributes, element_name, 
	attribute_name, content.

August 22, 2003 - Jerome

  - Fixed Makefile bug in Jungle introduced yesterday... (sorry)

August 21, 2003 - Jerome, Avinash

  - Added support for children et al in Jungle.
  - Added functions to serialize/deserialize QNames for Jungle's
    benefit.
  - ** NOTE: Now all node ids start at '1' instead of '0'. '0' can
    then be used as an 'exception' value.

August 20, 2003 - Mary

  - Merged with Jerome's changes of August 19

  - AST
    o Removed TTuple from principle type AST -- only occurs in Core
      type AST

  - C-API
    o query_context is now a struct with module_context and
      statements. [Reported by Nicola]

  - Java API
    TODO : still need to split query_context into module_context and
    statements and also change ItemLists into Java Lists.

  - Constructors:
    o Element & attribute constructors w/ multiple enclosed
      expressions now work (finally!)

  - Streaming:
    o BUG ALERT: Nested typed streams with holes do not work
      Currently, no exprs yield nested streams w/holes but we should
      figure out why this doesn't work.
 
August 20, 2003 - Jerome

  - Printing:

     o Split the printing code in separate modules that parallel the
       distinct AST's.
     o Added support for printing core updates.
     o Removed some non-necessary top-level printing operations.

August 19, 2003 - Jerome

  - AST's:

     o Removed obsolete XML AST.
     o Split models into primary and core AST.
     o Split datatypes into primary and core AST.
     o Propagate that change throughout the code.

  - Name support: large cleanup of the names/symbol support.

     o All AST's use either unresolved names or resolved symbols.
     o Removed all uses of unresolved symbols and 'generic' symbols.

  - XML Schema import:

     o Added first version of Vladimir's XML Schema import. This is
       still not fully hooked up into the main system.
     o Removed old XML Schema import code written by Byron (not
       touched since 2001!). This served its time well. Many thanks to Byron!

  - Parsing:

     o Removed old XML/XML Schema parser (DOM-like) for good!

August 18, 2003 - Jerome

  - Datamodel:

     o Removed duplicate name accessors from the data model. Some of
       those accessors are still present as 'helper' function in
       Dm_util. None of those function rely on symbols externally.
     o Changed implementation of xs_QName to be a standard resolved
       qname, not a symbol.

August 15, 2003 - Mary

  - Documentation
    o Started 0.4 doc: README, STATUS, etc.

  - Makefiles 
    o Incorporated all suggested changes for MacOSX

  - Constructors/Streaming
    o Added CEEnclosed to core
    o Added TSAX_startEncl/endEncl to typed SAX events
    o Still debugging "erase_atomic_value" which takes new events into
      account 

  - Galapi 

    o Aligned with new data model 

    o Moved galapi/c_api/glx_wrap.{mli,ml} to
      galapi/caml_api/glx.{mli,ml}
  
      Glx module contains functional interface to Datamodel
      constructors. 

      Aligned Glx _exactly_ with C-interface in galapi/c_api/glx.h
      (i.e., glx.h functions call Glx module functions directly) 

    o Galapi/Glx interfaces still need work w.r.t. orthogonality 

    o Updated & checked examples in examples/caml_api,c_api,java_api 

August 13, 2003 - Jerome, Mary, Avinash

  - Jungle:

     o Draft version of the Datamodel OO API for Jungle!!!!

August 13, 2003 - Avinash

  - Jungle:

     o Fixed memory leak bug in the Jungle C stub.
     o Implemented support for a Galax-level Jungle store and loading
       from an XML stream to that store.

August 11, 2003 - Jerome and Mary

  - Datamodel:

     o Added nilled accessor for element nodes.
     o Added nilled flag in typed stream (start element events).
     o Added nilled parameter/implementation in the Galax datamodel.
     o Added abstract data model primitives for updates.

  - Updates

     o Recompiled higher-level update support on top of the new data
       model primitives.

August 11, 2003 - Mary

  - Abstract DM:

     o Implemented Base URI.


August 8, 2003 - Jerome

  - Datamodel:

     o Re-enabled support for node identity and proper handling of
       duplicate removal.
     o Added generic support for node identity in ./dm
     o Created Galax_nodeid module for specific handling of node
       identity for Galax nodes in ./datamodel.
     o Some tests requiring node identity seem to pass!! Yeepee :) We
       are almost back in business now.

  - Misc fixes:

     o Added Avinash's name as team member!
     o Fixed validate tools to perform URI resolution properly.

August 8, 2003 - Mary

  - Alignment changes to normalization module for attribute & text node
    constructors. 

August 4th, 2003 - Jerome, Mary, Vladimir

  - Separated the XQuery AST into:
       -- XQuery common AST
       -- XQuery type AST
       -- XQuery AST
       -- Core XQuery type AST
       -- Core XQuery AST
  - Created first version of an type system AST for named typing.
  - Added a built-in prefix 'empty' bound to the empty namespace.
  - Added support for 'Delta Namespace Bindings'. Added those to the
    SAX resolved and typed streams.

August 3rd, 2003 - Jerome

  - DEBUGGING of previous major reorganization:

     o Fixed normalization of attribute construction to use
       atomization.
     o Fixed infinite loop in the Stream operation computing the
       attributes at the beginning of a stream.
     o Added back support for parent. Parent pointers are built after
       constructing each element or document node. This is not inside
       the OO interface, but is done by the loader after each node
       creation.

  - Added new jungle modules from Avinash, integrated them in the
    compilation process.

July 31, 2003 - Jerome and Mary

  - Finished the major reorganization of the data model:

    Again, there are too many changes to be recorded in detail. Major
    changes occurred in : Eval_util, Eval_expr, Stream_types,
    Stream_ops, and Extended_values (replacing tuples). Among other
    things that occurred:

    o Hooked up streaming as the main implementation of constructors
      (for element, document, attribute, text, pi, comment
      constructors).

      Streaming now at the heart of query processing. This will
      eventually improve preformances and compactness of the code, but
      this is likely to result in instability in system for a few
      weeks.

    o A new form of stream has been introduced to represent 'resolved'
      streams I.e., SAX streams where namespace resolution occurred.

July 29, 2003 - Mary and Jerome

  - Major rewriting of the data model support. The data model now is
    object-oriented, with a 'virtual' layer in ./dm and a Galax
    implementation of that layer in ./datamodel.

    There are too many changes to try and list them here...

July 29, 2003 - Vladimir, Philippe, Mary, and Jerome

  - Added support for polymorphic type annotations on cexpr.
  - Added first cut at Philippe and Jan's algorithm for removing SBDO.
  - Major rewriting of the processing model.

*****    THE PROCESSING MODEL STILL NEED WORK!!!!!

July 28, 2003 - Avinash and Jerome

  - Jungle:

     o Added Jungle in 'tools', the low-level wrapper for BDB indexes.

  - Cosmetic:

     o Fixed all of the Str. regexp generating those 'wrong escape in
       string' warnings.
     o Changed compilation process to take changes in the ./tools
       directory into account.

July 14, 2003 - Mary

  Problems:
  - definitions of recursive atomic* that extend definition of atomicValue
    causes problems
  - All the functions that depend on clean nodes  in fn_dynamic are
    not implemented. See: 
    _base_uri_zero 
    _lang
    _sequence_deep_equal
    _glx_save_document
    _glx_copy

  - None of the update operations are implemented at the moment
  - API is not compiled

July 10, 2003 - Jerome

  - Datamodel:

     o Datamodel was allowing document node inside element nodes! This
       is now detected and raised as an error.

  - WSDL:

     o Integrated Nicola's WSDL mapping code. This is still only
       available as a 'pre-compilation' command line tool that can be
       used to generate an XQuery 'context' from a WSDL file. More
       tuning is necessary for a solid integration.
     o Fixed WSDL support for the new form of atomic values.
     o Uses List.sort instead of List.fast_sort so Galax still
       compiles on OCaml 3.04.

  - Compilation:

     o Remove additional line doing compilation of tools in the main
       Makefile. Was just breaking the compilation process.

     ** Does anybody know what was the intent of that line? **

July 9, 2003 - Mary 

  - Reviewed and updated signatures & implementations of F&O
    functions.

July 8, 2003 - Mary 

  - Re-implemented Datatypes module using objects.  Top of object
    hierarchy is Datatypes.atomicValue.

    o Moved casting operation into Datatypes.atomicValue objects.
    o Removed stdlib/cast_as*

  - Modified API examples to all produce same results
   
    (NB: Please keep all these tests in sync!)

July 8, 2003 - Nicola

  - Galapi. Corrected bugs in Java API:

     o AttributeList inherits from NodeList and not directly from
       ItemList.
     o xsUntyped inherits from Atomic and not directly from Item.

July 8, 2003 - Jerome

  - Processing Architecture

     o Separated rewriting from evaluation. Rewriting is a phase after
       typing and before evaluation.

  - HTTP/SOAP:

     o Fixed CURL compilation problems. Cleaned-up installation of
       ./tools modules.

July 8, 2003 - Nicola

  - HTTP/SOAP:

     o Changed Http and Glx_http to allow full options for soap calls
     o Changed glx:soap-call in Fn_dynamic to use these extra options

  - Galapi:

     o Corrected bug in Java API: ProcessingContext inherits from
       Context and not from Item

July 7, 2003 - Mary

  - Datamodel:

     o For each node kind, coalesced all node constructors into one
       constructor. 

July 7, 2003 - Jerome

  - HTTP/SOAP:

     o Integrated HTTP changes, and glx:soap-call function from
       Nicola, without Curl yet.
     o Fixed 'native' HTTP support to fail when receiving an http
       error.
     o Added curl directory in the ./tools section. Started setting up
       compilation to deal with curl in a smooth way.

  - Namespaces:

     o Fixed bug in normalization of element construction and
       namespaces. Normalization was not propagating the namespace
       declarations from within the element. Added corresponding tests
       in XQuery unit:

       <a xmlns:foo="xyz"><foo:b/></a>  succeeds.
       <foo:a xmlns:foo="xyz"/>         succeeds.
       etc.
     o Changed namespace support to treat xmlns:foo="" as *removing*
       the namespace binding for prefix foo! See the namespace rec.

June 30, 2003 - Jerome

  - Galapi:

     o Fixed bug in 'free' operation. Now free can operate several
       times on the save value safely. Useful when called from the
       Java API.

June 29, 2003 - Jerome

  - F&O:

     o Added Galax function glx:save-document(), which takes a file
       name, a *document* node, and output the document back into the
       given file. This is useful for application who want to perform
       updates on documents.

  - XML Updates:

     o Implemented replace value of properly. Was not updating the
       value as it should have.
     o Outlawed replace value of on a document not, which does not
       make any sense.
     o Fixed bug in parsing of replace within complex updates. Did not
       add them in the parser properly due to two separate production
       simple update and simple_update_with_empty in the case of
       complex updates.
     o Fixed bug in normalization of conditional update, was not
       computing the effective boolean value.

June 27, 2003 - Jerome

  - Printing:

     o Removed obsolete print QName functions.
     o Fixed bug in pretty-printing of function applications when one
       of the expression to which the function is applied is a
       sequence. (was missing parenthesis.)
     o Implemented pretty-printing for the query prolog and for the
       query module.

  - F&O:

     o Fixed bug in fn:substring. Index was off 1 because indexes
       start at 1 in XQuery, but at 0 in Caml.
     o Fixed fn:translate, which was completely wrong...

June 27, 2003 - Mary

  - Galapi:

    o Added support for document constructors to all APIs.
    o Added tests of document constructors to all examples/
    o Cleaned up galax/java_api/Makefile
    o Added java_api/galapi/Document.*

June 26, 2003 - Jerome

  - Serialization:

     o Fixed serialization in the canonical version to avoid an
       unnecessary initial new line.

  - Parsing:

     o Added some support for parsing of inlined general and parameter
       entities.
     o Added some support for expansion of general entities. Seems to
       work ok on a number of tests inside element content and attribute
       values.
     o Fixed bug in parsing of processing instructions in the document
       root. Was not parsing the first whitespace character after the
       target properly.

June 24, 2003 - Nicola

  - Galapi:

      o Removed debug messages from Item.java and Itemlist.java.


***** Galax Release 0.3.1 June 22, 2003 *****

June 22, 2003 - Jerome

  - A few fixes to the Web site: added pointers to the new
    distribution, added couple of breaking news entries for the new
    Web site, etc.
  - Fixed the release notes.
  - Fixed bug in pretty-printing for the document node in the XQuery
    serialization mode.
  - Changed default pretty-printing to be well-formed for the Galax
    API.

June 22, 2003 - Jerome

  - Tidying up serialization:

     o Fixed pretty-printing for the top-level sequence in the XQuery
       serialization form.
     o Factorized the serialization code dealing with generation of
       proper output character encodings.
     o Fixed printing of the document header in the Well-formed and
       canonical case. Avoid having a wrong extra new-line between the
       header and the root element.
     o Allowed pretty printing of the opening tag of elements with
       text content inside elements with element content. This enables
       pretty-printing for the corresponding attributes which is safe
       and somewhat nicer.
     o Fixed bug in the pretty-printer for attributes, was printing a
       wrong additional new line for the first attribute.

  - Fixed bug in static typing of FLWOR with order by + where
    clause. Changed normalization not to use the empty tuple but the
    empty sequence instead. Changed the data model to remove empty
    sequences in sequences of tuples.

June 21, 2003 - Jerome

  - Installed the new open source LICENSE.

  - Serialization overhaul:

     o Implemented serialization as a streaming operation. Passed the
       regression tests with this new version. Removed obsolete form
       of serialization within the data model.

     o Fixed many very serious problems with pretty-printing in the
       presence of whitespace. Pretty printing should at least be
       correct semantically now. This fixes problems in several
       usecases, notably XMark, STRING, NS, and SGML.

       Note: pretty-printing only occurs for outer-most elements as
       long as they are of element content. Elements with text/mixed
       content are never pretty-printed. The serialization does not
       make any attempt to pretty-print elements with element content
       if they are nested in elements that have text/mixed content.

     o Fixed pretty-printing of attributes.

     o Changed Galapi to use streaming serialization.

  - 'validate' command line tool:

     o Changed 'validate' to use streaming serialization. Now
       'validate' does not build any datamodel except when used with
       the -dm flag. That means 'validate' can be executed in a purely
       streamed fashion. This has been tested to work on the DBLP
       source document (about 160 Mb).
     o Added support for output encoding to the 'validate' tool.
     o Added clear error message to indicate that this version of
       'validate' does not support XML Schema validation.

June 19, 2003 - Nicola

  - Corrected bug when calling items_next in java_api/glx_jni_stub.c

June 17, 2003 - Mary

  - Removed datamodel/serialization_context from Makefile.
  - Fixed rewrite rules for "some" expr -- were not applied
    correctly. 
  - Added rewrite rules for "every" expr. 

June 16, 2003 - Nicola

  - Replaced the free() method in the Java API with the
    standard finalize() method. Consequentially, calls to
    destructors are made at garbage collecting time. 
  - Modified Java API to take this into account by removing
    explicit memory deallocations.

June 15, 2003 - Jerome and Nicola

  - Added generic SAX event-handler.
  - Added incomplete version of serialization working directly from an
    XML stream.

June 13, 2003 - Jerome

  - Tidying up for the release:

     o Fixed printing of error messages that had been wrongly
       disabled.
     o Fixed some small problem in the installation of the website.
     o Fixed some of the content on the new website. Added Mary's
       paper at ECOOP. Added latest version of the projection
       paper. Added the VLDB demo paper.
     o Changed compilation of the APIs not to use ocamlmklib which is
       not supported by the ocaml MinGW port.

June 12, 2003 - Mary

  - toplevel/
     o Fixed bug in which errors were not flushed to stderr

  - Errors

     o Added Undefined_Variable to error.ml, and error message for
       indicating that context item is not defined. 

  - Regress
     o xqueryunit-*.cfg files are generated automatically from
       xqueryunit-*.cfgp. 

  - Installation
     o Changed Makefile so that examples/ and usecases/ each does own
       installation steps and copies appropriate READMEs.

  - Documentation
     o Updated galapi/README to describe new API usage and how to run
       examples/. 

June 11, 2003 - Jerome

  - Galapi:

     o Added a few missing files in the repository.
     o Made sure the Solaris-packaged version recompiles cleanly under
       Linux and that the API examples run.

June 11, 2003 - Jerome & Mary

  - Galapi:

     o Added missing methods in the Java API.
     o Fixed Java examples to perform more complete testing.
     o Removed argv from the init calls in the C and Java APIs and
       fixed corresponding bug in the Java API.
     o Fixed the compilation process to work under Solaris.
     o Removed all dependancies to the Caml distribution in the C and
       Java API, which are now completely stand-alone.

June 10, 2003 - Jerome

  - Galapi:

     o Fixed the exception problems in the Java API.
     o Cleaned up the API to use GalapiException consistently.
     o Updated the examples to use try out exceptions as well.
     o Cleaned up the compilation for the java part of the
       Galapi. Dependencies seem to work more smoothly now...
     o Changed error printing not to add improper new lines at the
       end.

June 9, 2003 - Jerome

  - XQuery:

     o Implemented '/' !

  - Streaming:

     o Added attribute and atomic values in well-formed XML
       streams. This is not require for loading an actual XML files,
       but for closure when serializing back an arbitrary datamodel
       instance.
     o Added a typed-XML stream that can represent a typed XML
       instance, instead of an untyped one.
     o Added operation to build a typed XML stream from a data model
       instance.
     o Added a basic operation which turns a typed XML stream into a
       well-formed one. [This is in fact an implementation of
       erasure...].

  - Galapi:

     o Fixed all of the exception handling by wrapping Caml calls to
       raise a generic Caml error instead of a Galax error.
     o Added one C API function:
        'glx_eval_statement_from_file_with_context_item_from_xml'
     o Fixed definition of item list to use 'value *item' instead of
       'item item' which does not compile under C++.

June 4, 2003 - Mary

  - Documentation 
      
      o Remove galax/DEVELOPERS
      o Updated galapi/README to describe specifics of C & Java APIs,
        added CAVEATS, etc.

  - Rewriting rules:

      o Instead of removing glx:docorder, replace it by new function
        glx:node-sequence, which is no-op on node sequences, but
        raises a type error if any argument is not a node.
      o Added static typing optimization to remove glx:node-sequence
        when applied to Node*.

June 4, 2003 - Mary

  - Galapi:

      o Added error support in the C API.

June 3, 2003 - Nicola

  - Galapi:

      o Aligned the Java API with latest C API.
      o Made up a jar package for Galapi.
      o Added new tests in ./examples/java_api.
      o Builds javadoc for users' benefit.
      o Cleaned up installation process.
      o Fixed some bugs in the JNI stub (missing free of the allocated
        memory).
      o Implemented toString for items.
      o Context does *not* herit from item and is now in a separate
        class hierarchy.

May 30, 2003 - Mary 

  - Cleaned up Makefiles in examples/*_api directories

     o Added rules to "run" examples.

  - Updated README files

     o Installed galapi/README in $(GALAXHOME)/README_API.


May 29, 2003 - Mary

  - galapi:

     o Changed serialize_to_output_channel to serialize_to_formatter.

  - conf: 

     o Added default headers & footers values for printing output of
       phases.
     o Added function for escaping characters after pretty-printing.

  - website/demo

     o Changes to emit headers/footers and xml/expressions in correct
       format.

May 28, 2003 - Jerome

  - Galapi:

     o Fixed *killer* bug in the item_cons operation. Needed to
       protect the GC global root registration with CAMLxxx macros
       properly.
     o Fixed all mallocs to items to the size of values instead of
       item.
     o Removed unnecessary string copy in several stub functions.
     o Fixed bug in Java API: Attribute and Element construction was
       not setting the native item properly.
     o Fixed bug in Java API: Java_ItemList_nativeItemsFirst was
       wrongly named as Java_Item_nativeItemsFirst.
     o Fixed all bugs in Test2.java. All of the Java tests now run
       properly.

May 28, 2003 - Jerome

  - Parsing:

     o Fixed bug in parsing of attributes in XQuery element
       constructors (was not allowing proper whitespace before/after
       the equal sign). [Bug report thanks to Gargi, tripped in one
       replace update].
     o Fixed bug in parsing of entities in attribute content. [Bug
       report thanks to Peter Patel-Schneider].
     o Fixed bug in pretty-printing of processing-instructions. Should
       not escape characters like text.

  - Galapi:

     o Added Java API's code by Nicola and built corresponding
       examples.
     o Removed obsolete Java and C API files.
     o Created separate sub-directories for the Caml, C and Java APIs.
     o Cleaned the compilation process to use ocamlmklib whenever
       possible and remove unnecessary steps.
     o Made sure the compiler builds both static and shared C
       libraries whenever possible.
     o Fixed make install to put the right interfaces in libraries for
       the Caml and C APIs.
     o Merged with changes made by Mary (see below).

May 22, 2003 - Gargi

  - Added replace code to eval_update.ml.
  - Cleaned up the code for evaluating flwor updates.

May 21, 2003 - Mary

  - Changed literal integers from Caml int to BigInt

  - Changed all xsd:int arguments to xsd:integer arguments in
    pervasive.xqp + corresponding changes to function implementations
    in fn_dynamic.ml

  - Moved base_uri from Fn_dynamic to Datamodel
  - Added base_uri and node_name to Galapi
  - Added glx_base_uri and glx_node_name to C-API glx.h
  - Updated tests in examples/{ml,c}-api

  - (Jerome) fixed bug in glx_float_of_xs_double in glx_stub.c
 
May 21, 2003 - Jerome

  Complete reorganization/reimplementation of document
  projection. Clean up of document loading.

  - Document projection:

     o Cleaned up all of the document projection support. Renamed the
       analysis directory into the projection directory. Inside that
       directory, the code has been reorganized in the following way:

       	Path_struct[mli]   contains the types for path structures.
       	Path_structutil    contains basic operations on path
       	                   structures.
       	Optim_vars         contains operations on variables for
       	                   optimized projection.
       	Analysis_context   contains the context used during path
       	                   analysis.
       	Function_analysis  contains path analysis for built-in
       	                   functions.
       	Path_analysis      contains the main path analysis algorithm,
       	                   including optimized analysis.

     o Implemented projection as a pure XML stream operation. This
       separates projection from document loading, making loading a
       very simple operation and projection is now much more
       compact. Projection is implemented in the [Project_context] and
       [Stream_project] modules. An added benefit is that now the
       result of projection can be fed into any other stream
       operation.

     o Re-connected document projection to the top-level execution of
       the query. Modified the Galapi to call document projection at
       the appropriate time.

     o Fixed path analysis to work on this new version of Galax,
       notably:
         -- Added analysis for the glx:distinct-docorder function
       	 which is used in path expression. Analysis is such that it
       	 propagates returned and used paths through that function.
         -- Changed the top-level function from where to start the
       	 analysis from glx:document-sax to fn:document which now uses
       	 SAX parsing.

  - Data model loading:

     o Implemented cleaner extraction function for special attribute
       out of the SAX events.
     o Removed 'xmlns' declarations in the document from the data
       model.
     o Made sure the document is loaded in a "fresh" namespace
       environment! Otherwise, wrong namespace prefixes can be used
       in one XML document (e.g., <fn:a/> used to parse fine... it is
       now an error).

May 20, 2003 - Mary

  - Started alignment with May 2003 Working Drafts

  - XQuery 1.0 May 2003
    -------------------
    o Updated STATUS
    o Renamed 
      xs:AnyScalar => xdt:anyAtomicType
      xs:untyped => xdt:untypedAtomic
    o Added:
      xdt:yearMonthDuration
      xdt:dayTimeDuration

  - F&O May 2003
    ------------
    o Removed: 
      fn:copy
      glx:unique
      glx:optional
      fn:deep-equal(node, node)

    o Added: 
      fn:zero-or-one
      fn:one-or-more
      fn:exactly-one
      fn:trace 
        NB: glx:print-string, glx:print-item not removed
      fn:unordered
        NB: fn:unordered is implemented by CEUnordered expr.

    o Renamed:
      fn:insert => fn:insert-before      
      fn:sequence-deep-equal => fn:deep-equal(item*, item*)
      fn:sequence-node-equal => fn:sequence-node-identical

    o Changed:
      fn:index-of defined on atomic values, not nodes

May 12, 2003 - Jerome

  - Galapi:

     o Fixed bug in C stub: was using an un-initialized variable in
       glx_default_processing_context. Now the C example works for
       both the byte code and native code C libraries.

  - HTTP support:

     o Added function to create an XML stream parser from a string.
     o Integrated Trevor's 'mini http client' to Galax. (At
       last...).
     o Changed the fn:document() function to support 'file:' and
       'http:' URIs.

  - XML parser and serialization:

     o Fixed several bug in parsing of processing instructions.
         [ Was not removing leftmost whitespace from the PI content. ]
         [ Was not allowing PI target with leftmost whitespace. ]
     o Generates output encoding for all of the markup.
     o Added proper serialization of the document node as an XML
       declaration or as two bytes for auto-detection of UTF16
       encoding.

  - XQuery parser:

     o Fixed support for entity reference and character references in
       the XQuery parser. Only built-in entity references are
       supported in the XQuery syntax (i.e., &lt;, &gt;, &amp;,
       &quot;, and &apos;).

May 9, 2003 - Jerome

  - Whitespace in XML documents and serialization:

     o Fixed bug in pretty-printing of attributes.
     o Changed the serialization to use pretty-printing only for
       'element content' elements.

  - Parsing and streaming(!):

     o Removed obsolete code to load an XML AST to the data model.
     o Reorganized all of the parsing layers. Split the code into the
       following directories:
         ./ast    	   for the Abstract Syntax Trees and printing
         ./lexing 	   for the lexing layer
         ./parsing    for XQuery parsing (and the old XML parser)
         ./streaming  for XML streaming operations including SAX parsing
     o Added Stream_ops module for simple operations on XML
       streams. First operation is discarding a subtree and is used
       during projection.
     o Removed 'skip' operations from Load module. Removed
       construction of data model from XML AST in Load module.

May 9, 2003 - Gargi

  - Cleaned up update code
  - Fixed bug in insertasfirstinto in flwor update

May 8, 2003 - Jerome

  - XML Updates:
     o  Fixed bug in optimization of updates.

May 6, 2003 - Mary

  - Added "Quick Start" API functions to ML and C APIs
    eval_statement_with_context_item_from_document
    eval_statement_with_context_item_from_xml
    eval_statement_with_variables_from_xml
  - Added missing glx_build_external_context to C API
  - Updated API tests in examples/*-api

May 2, 2003 - Jerome

  - More work on XML update support:

    o Turned rewriting rules for sorting-by-docorder on again.
    o Applied rewriting pass on update expressions. This is necessary
      because of the effect of SBDO which makes the language unusable.
    o Added support for parsing, AST and normalization for the
      'replace' simple update.
    o Hooked up query rewriting to the update code. This way updates
      can be applied to real-size documents!

May 1, 2003 - Mary

  - Fixed bug in evaluation of order-by clause within multiple let-for
    clauses. 
  - Fixed bug in static typing of text{} expression.

May 1, 2003 - Jerome

  - Fixed bug in XML parsing of XML declaration for the
    document-validate function (still using the old parser for now...).

  - Cleaned up the evaluation directory:

     o Split the evaluation module Xaeval into several smaller
       components:
         Eval_util   contains some auxiliary evaluation functions.
         Eval_expr   contains the main eval loop for expressions.
         Eval_update contains the evaluation code for updates.
         Eval_top    contains the top-level evaluation code and the
                     evaluation code for the query prolog.
     o Moved Cast and Xerror to ./stdlib, since they are part of F&O.
     o Removed obsolete Xaeenv and Xaeval modules.
     o Renamed all Optim* modules to Rewrite_*, in prevision of the
       change to the new architecture.
     o Added small module description at the beginning of each
       module. Cleaned-up comments here and there.

  - Cleaned up stdlib directory:

     o Changed names for consistency:
         Xerror becomes Fn_error
         Cast   becomes Cast_as
         Built_in_static becomes Fn_static
         Built_in_dynamic becomes Fn_dynamic
     o Added small module description at the beginning of each
       module. Cleaned-up comments here and there.
     o Removed obsolete code for document() using the old parser.
 
April 30, 2003 - Mary

  - Variety of bug fixes.  See BUGS.
  - Tested ML & C APIs 
     o Added tests in examples/ml-api/ml-*.ml, examples/c-api/c-*.c
  - Completed command line options for loading documents and binding
    document values to context item or global variable:
    galax -context-item file-name ...
    galax -var var-name=file-file ...

April 29, 2003 - Jerome

  - More fixes on the XML updates feature:

    o Fixed bug in the replace_node function in the data model which
      is used to perform deep-copy of nodes related to element
      construction.
    o Trapped the datamodel error message for updates to be re-raised
      more nicely at the evalution level.
    o Fixed bug in normalization of conditional expression: was
      normalizing the first branch of the conditional into both final
      branches!

April 29, 2003 - Mary

  - Completed implementation of ML API in galapi/galapi.mli

April 29, 2003 - Jerome

  - Some minor additional fixes cleaning up after the big changes
    yesterday:

    o Added target to launch regression tests from the top-level
      directory.
    o Fixed XQuery-style serialization for the empty sequence. Is now
      () instead of nothing.
    o Fixed bug in the parser for CDATA sections, should not discard
      the text buffer when going out of the CDATA lexing state.
    o Aligned flags in the validate command-line tool with those in
      galax. Notably:
         -print is replaced by -serialize [wf/xquery]
         -whitespace is replaced by -xml-whitespace
         -xml-pic is added
    o Added proper error message when a non-declared general entity is
      encountered.

  - Fixes on the XML updates feature:

    o Fixed bug in datamodel delete: should delete the children before
      deleting the parent in the recursive deletion of the
      subtree. Otherwise, the access to the children fails because
      applied on an already deleted node!!
    o Fixed bug in datamodel insert: the inserted node must be
      'deep-copied' i.e., copied with node ids generated, in order to
      survive future access at the actual location.
    o Changed evaluation of statements in Galapi to return nothing
      in the case of update statements. This means the return type for
      the eval_statement is now an option.
    o Added one kind of error for updates. Replaced use of the
      Prototype error in the update code to the Update error. Made
      sure the update errors get some kind of file information to
      point the user to the location where the error occurred.

April 28, 2003 - Jerome

  - Minor fixes to the Web site scripts.
  - Minor fixes to the Makefiles in ./regress and ./usecase so that
    things do not depend on old regression test scrips.

April 28, 2003 - Jerome

  **XQueryUnit is now the official test suite for Galax.**
  **SAX is now the official XML parser for Galax.**

  - Yet more work on the SAX parser:
    o Removed dependency to PXP modules from the rest of the
      system. Only Glx_pxp depends on the PXP package now.
    o Fixed make depend to take tools into account correctly.
    o Removed dead-wood code from Encoding module.
    o Fixed bugs in parsing of the XML declaration <?xml ... ?>, and
      made the corresponding SAX code more compact.
    o DTD parsing put in place. Lexing of DTD tokens is now hooked up
      to the SAX parser, although discarded for now.
    o Added support for character entity references (&#???; now works
      in your documents!).
    o Added some minimal support for general entity references (&lt;,
      &gt;, &apos;, &amp; and &quot; are now working properly).
    o Added some preliminary parse context.

  - Regression tests:
    o Hooked up Galax regression tests to XQuery unit. There is now a
      Make target to call XQueryUnit directly from the galax source
      directory, setting up the right configuration, and gathering
      XQueryUnit results back into the ./regress directory.
    o Added Galax-specific XQueryUnit XML parsing unit in order to
      test the new SAX parser.
    o Added Galax-specific XQueryUnit XML update unit in order to test
      new support for XML updates.
    o Removed all of the old tests in ./regress.

April 21, 2003 - Jerome

  - Re-enabled support for updates. Updates now use the new form of
    context.
  - Added support for DTD *lexing*.

April 21, 2003 - Mary

  - Updated glx_wrap.ml,glx.h,glx_stub.c to call new galapi functions

April 20, 2003 - Jerome

  - Reorganized the content of the stdlib directory:
      * Split Bltin into Built_in_dynamic which contains evaluation
        code for all the built-in functions and operators, and
        Overloaded which contains support for overloaded functions.
      * Created Built_in_static which contains static typing code for
        built-in functions which have special static typing
        rules. This is extracted from Typing, which now uses this
        module.
  - Removed multiple bindings of function signature used for
    overloaded functions from the normalization context. Typing and
    Xaeval use Overloaded module instead. This fixes the bug for avg
    et. al mary pointed out.
  - Fixed wrong signature for max-int and min-int functions.
  - Moved support for configuration of character encodings from
    base/conf.ml to ./base/encoding.ml
  - Added -print-error-code option which prints an error code instead
    of a complete error message. This is for the benefit of unit
    testing.

April 18, 2003 - Mary

  In galapi/
  - Added import_main_module_from_string to load main module from string
    for website code.

  In website/ 
  - Updated to align with v0.3.1 code.
  - Fixed some link errors in .html files.

April 16, 2003 - Mary

  - *** Gargi's New update code modified to use contexts but does not compile ***
  - Changes to normalization/typing/evaluation to better align with
    galapi. 
  - Implemented much of galapi.mli using new context interfaces.
  - Updated toplevel/xmlquery to use galapi

April 11, 2003 - Mary
  - Removed normalization/{xadepend,xatenv,xaeenv}.{ml,mli}
  - Added normalization/{norm_context}.{ml,mli}
          typing/{static_context}.{ml,mli}
          evaluation/{dynamic_context}.{ml,mli}
  - Modified normalization/typing/evaluation to use above contexts
  - Started to write stubs in galapi/context.{ml,mli} to call new
    context code & new mapping functions

April 8, 2003 - Mary
  - Bug fixes for v0.3.1.  See BUGS file.
  - Added command-line option (-sbdo [on/off]) for disabling sort-by-doc-order in
  normalization of path and union expressions. 

April 7, 2003 - Mary
  - Aligned examples/c-api/c-example.c with examples/ml-api/ml-example.c
  - Fixed bug in galapi/glx_stub.c


April 4, 2003 - Gargi

  - Added new datamodel primitives for insert update operations
  - Added new evaluation functions for processing simple and FLOWR updates

April 3, 2003 - Mary

  - Addded support for aggregate functions defined on empty sequence.
  - Added some internal doc.

April 2, 2003 - Mary

  - New API documentation in galapi/README
  - More testing of ML and C APIs 
    (See TODO list on galax-dev mailing list)
  - Moved galapi/glx_params* to galapi/itemlist*

April 1, 2003 - Jerome

  - Removed pxp_lib.ml which is copied from two independant files
    depending on the kind of compilation and was making the CVS
    repository schizophrenic.

April 1,2003 - Krish

  - Added C-API functions for Evaluation context, Query evaluation.

March 30, 2003 - Jerome

  - Cleaned up configuration for various character encodings:
     o There is only one configuration variable for the internal
       encoding now.
     o Internal encoding can be setup from the command-line.
     o SAX parser now sets character encoding for the input document
       correctly, from the XML declaration.

  - Cleaned up serialization:  
     o Changed serialization to take the character encoding into
       account. By default, serialization outputs using the default
       character encoding. This can be overridden from the command
       line.
     o Removed obsolete code for xsi:type generation.
     o Galax now supports two forms of serialization: either
       well-formed, which always generates a well-formed XML fragment,
       as specified for XQueryUnit, or XQuery, which always generate a
       valid XQuery expression for the corresponding XML fragment.
     o Fixed bug in serialization when preserving whitespace was not
       using the proper configuration parameter.

  - Cleaned up command-line options:
     o Presented the command-line options in a more organized fashion,
       with a nicer usage message. Renamed some of the options to be
       more consistent.
     o Changed command-line options for serialization, requiring to
       give explicitly the serialization form.
     o Removed unused 'blocking' parameter.

  - Cleaned up the SAX parser:
     o Fixed all error messages to correctly report location and the
       kind of error. Error messages take transitions between tokens
       into account.
     o Fixed error messages when reaching the end of a file.
     o Allowed whitespace around the '=' sign in attributes.

  - Removed the MAKE variable from the makefile configurations, which
    screwed up propagation of the make executable used for
    compilation.

  - Fixed bug in error reporting when encountering EOF in the middle
    of parsing.

  - Cleaned up the validate command-line tool.
     o Removed obsolete wf-sax command line tool. Now validate should
       be used directly.
     o Changed the validate tool to use the SAX parser. Validation is
       actually turned on for now, until validation is re-implemented.

March 28, 2003 - Mary

  - More fixes to C-API (See TODO list on galax-dev mailing list)
  - Moved examples/c-api/glx_main.c to examples/c-api/c-example.c

March 27, 2003 - Jerome, Mary, Krish

  ** C API under construction
  - Updated galapi/galapi.{mli,ml} to align with Jerome's code. 
  - Updated galapi/glx* to implement C API.

February 25, 2003 - Jerome

  - Fixed compile query operation in Galapi.

February 19, 2003 - Jerome

  - Fixed problem in multiple binding of overloaded functions. **
    DIRTY FIX ** need to be cleaned up.

February 10, 2003 - Jerome

  - Fixed serialization to escape correctly markup characters such as
    '<', '&', etc.
  - Added support for XQuery 'castable as' expression, forgotten from
    Nov. 2002 working drafts.
  - Fixed bug in the ISO-8859-1 lexer, initializing with the wrong
    encoding.
  - Added two printing functions to stderr:

      glx:print-string-err(xsd:string)  as (empty)
      glx:print-item-err(item)          as (empty)

February 9, 2003 - Jerome

  - Support for all kinds of character encodings on the input. Hooked
    up Galax's XML lexers to PXP's resolvers. This takes care of
    reading a large numbers of characters encodings, including UTF-16,
    various ISO formats, etc. See PXP's documentation for more
    details.

  - Serious cleanup and improvement of the SAX parser:
     o Fixed bug in lexing of UTF-8 characters, now accessed properly
       from the lexing buffer.
     o Fixed SAX parser to keep track of the parsing location when an
       error occurs, and print errors according to the file currently
       being processed.

January 30, 2003 - Mary

  - All avg() & sum() functions were not defined on empty sequence.  
    avg( () ) => ()
    sum( () ) => zero

January 25, 2003 - Jerome and Amelie

  - Fixed projection to run again under 0.3. This is now working as
    before.


***** Galax Release 0.3.0 January 22, 2003 *****

January 22, 2003 - Jerome

  - Fixed the release number to 0.3.0

January 21, 2003 - Mary and Jerome

  - Fixed last minutes problems with use cases, tests and compilation.

January 19, 2003 - Mary
  
  - Added RELEASE_NOTES (documents changes in each new version)

  - Added RELEASE_NOTES and Bugs to web site.

  - Fixed evaluation and typing of FLWOR expressions with order-by
    clause and nested FLWOR expressions

    o Inner-for expression must contain sub-expressions (possibly
    if-then-else or let) that create ordered tuples *only*.

    o where expressions are now evaluated *before* order-by
    Evaluation of a where expression may yield the empty tuple. 

    o Changed _Concat_Extended_Value_Sequence to eliminate empty
    tuples during concatenation.

    o Changed typing of tuple expressions to permit the empty tuple
    type.
  
    o Changed CESort to take explicit sort variable instead of
    $glx:dot.  This is a necessary (and cleaner!) way to supported
    nested FLWOR with order-by clauses.

January 17, 2003 - Mary

  *** In the home stretch for v0.3!! ****

  - Updated website & usecases to include:
    NS, PARTS, REL, SEQ, SGML, STRING, TREE, XMP, XMARK, XML Schema
  - Deleted SEMANTICS from usecases
  - Removed evaluation/optimgen.{mli,ml}
  - Changed default return type of function to item*.
  - Fixed bug in FLWOR expressions that include both WHERE and
    ORDER-BY by normalizing:
    for $v in E1 where E2 order-by E3 return E4
    ==
    for $v in E1 order-by E3 return if (E2) then E4 else ()

    Inner-for expression must only create ordered tuples; any other
    complex expression raises error during evaluation.

    *Did not* touch corresponding UFLWR code. 

  - Replaced fn:concats by fn:string-join
  - Made fn:concat variadic.
  - Moved definition of typed-value from bltin.ml to into
    datamodel.ml.
  - Updated INSTALL/README
  - Fixed << and >> to be *strictly* less than/greater than
    comparators of document order.

January 9, 2003 - Jerome

  - Changed toplevel to apply optimization passes on the query
    prolog when the optim flag is on.
  - Turned the optim flag on by default!
  - Implemented first cut rough optimization rules to remove
    unnecessary sorting by document order. **THOSE RULES NEED TESTING**

January 9, 2003 - Mary

  - Added typed_value function to Datamodel.  Was named
    '_fn_data_node' in Bltin.

  - Added galapi/Datamodel_api.  Contains O'Caml object interface to
    Galax.  (Under construction)

January 8, 2003 - Mary

  - Moved built-in variables dot, position, last, sequence to
    Xquery_astutil so available to all modules.
  - Changed prefix of all built-in and generated variables to "glx" to
    avoid name conflicts.
  - Fixed bugs in parsing and printing of optional type declaration in
    'for' and 'let' expressions, and of printing optional variables in
    'case' clauses.
  - Changed signatures/implementations of all arithmetic operators to
    accept empty sequence as argument.  This change eliminates all the
    typeswitches that were polluting the normalized code. 
  - Implemented type-based optimizations for: 
    o fn:data
    o fn:boolean
    o dynamically dispatched functions converted to statically
      dispatched functions

January 5, 2003 - Mary

  - Updated Makefile and INSTALL, NEW, README, and STATUS
     documentation for v0.3 release.

January 5, 2003 - Mary

  - Following changes align arithmetic & comparison operators with
    *February 2003* working drafts:

    * Implemented general comparison operators 
      o untypedAtomic cast to double
    * Implemented new semantics for atomization/fn:data()
    * Implemented fn:boolean()
    * Implemented semantics of cast on (atomic?) only 
    * Changed EBV code to call to fn:boolean()
    * Implemented conversion rules for function arguments when
      required type is <: anySimpleType.

  - Fixed bug in node & order comparisons: empty sequence not handled.
  - Fixed type declarations to propogate declared type, not type of
    expression.

December 20, 2002 - Jerome

  - Fixed bug in parsing of function declaration when no type is given
    for the parameters / return clause. Now generates the proper item*
    default type.

December 18, 2002 - Mary

  - Following changes align arithmetic & comparison operators with 
    *February 2003* working drafts:

    * Implemented promotion semantics for arithmetic operators
      o untypedAtomic cast to double
    * Implemented "new" transitive "eq" operators 
      o untypedAtomic cast to string.
      o eq on () raises type error.

  - TODO: 
    - Change atomization to fn:data()
      Change EBV to fn:boolean()
      Implement fn:boolean()
    - Optimization rules for function dispatch & fn:data()
    - Clean up datatypes & datamodel : remove unnecessary operators

December 13, 2002 - Mary

  - Implemented dynamic function dispatch.
  - Implemented promotion of atomic values/types in dynamic & static
    semantics.  
  - Added monomorphic equality/inequality operators for all numeric
    types, string, QName, and anyURI.
  - Changed to implement above:
    stdlib/{bltin.ml,pervasive.xqp}
    toplevel/{toputils}*
    normalization/{mapping,maputil}*
    datatypes/{datatypes}*
    datamodel/{datamodel}*
    typing/{typing}*
  - Fixed string manipulation bugs.
    String.index and String.sub not guarded by 'try-with'.
    Changed split_string_on_char into split_left_on_char &
    split_right_on_char.
    Changed:
    base/gmisc.ml
    namespace/qname_lexer.src

  - TODO: 
    - New semantics for fn:data()
      New semantics for fn:boolean()
    - Optimization rules for function dispatch & fn:data()
    - Clean up datatypes & datamodel : remove unnecessary operators

December 9, 2002 - Jerome

  - Implemented two 'printing functions':

      glx:print-string   which prints a string on stdout
      glx:print-item     which prints an XML item on stdout

    to help users debug their XQuery programs [Requested by Bard Bloom
    Dec. 4th]

  - Fixed namespace handling to reject wrong rebindings of special
    namespace prefixes xml: and xmlns:.
  - Implemented support for fn:base-uri.
  - Implemented support for xml:lang:
       F&O function fn:lang
       Galax function glx:get-lang which merely computes
       the right language for any given node.

December 5, 2002 - Mary

  - Added 'Bugs' file for tracking bugs
  - Fixed bug in fn:data().  On content with type xs:anyType, returns
    string-value. 
  - TODO: align fn:data() with new definition.

December 5, 2002 - Jerome

  - Fixed bug in variable substitution for the position variable.
  - Fixed bug in galapi: should not put the default context in a
    global variable.

December 4, 2002 - Mary

  - Made glx:docorder and glx:distinct-docorder builtin functions.  
    Changed typing accordingly.
  - Implemented semantics of path expressions, including ordering by
    document order and removing duplicates. [Section 3.2, FS Section
    5.2]
  - Changed core grammar to distinguish between CEForwardAxis and
    CEReverseAxis. 
  - Added warning print functions to base/error
  - Fixed bug in sort : stable & non-stable were swapped
  - Changed errors to warnings in schemamapping for unsupported
    features: annotation and any group.

December 4, 2002 - Jerome

  - Attributes now allow the '....' syntax as they should!
  - Turned xf into fn everywhere. [Nov. 2002 WD]
  - Changed fn: and op: URI's to:

          http://www.w3.org/2002/11/xquery-functions
     and  http://www.w3.org/2002/11/xquery-operators

    respectively.
  - Implemented normalization for FLWOR updates. This makes use of the
    tuple features.

December 3, 2002 - Jerome

  - Changed the syntax of function declarations to use the new syntax
    with 'as SequenceType'.
  - Changed the syntax of typeswitch to use the new 'as' keyword
    before sequence types.
  - Changed () to empty in the sequence type [ Old change missing ]

December 2, 2002 - Jerome

  - Fixed last dirty bug in the semantics of sort by due to the fact
    that variable substitution was not working properly for the '.'
    variable.

November 27, 2002 - Jerome

  - Fixed numbers of bugs in handling of legacy E1 sort by E2
    expression.
  - Fixed problem in whitespace handling in the pretty printer (should
    not print whitespace-only text in element content to the output).
  - Fixed static typing for distinct-node, missing.
  - Fixed parsing bug in XML documents, was not allowing whitespace
    around equal sign in attributes.
  - Added support for tuples in the data model and type system.
  - Added new Core operations dealing with tuples.
  - Added support for "order by" in FLWR expressions using the new
    tuple operations.
  - Fixed variable substitution to work when the substituted variable
    is $fs:dot! Note that in that case, the expression AxisNodeTest
    must be treated separately. The way variable substitution is done
    is by adding the following let just around it:

    Axis NodeTest [ Expr1 / $fs:dot ]
    ------------------------------------
    let $fs:dot = Expr1 in Axis NodeTest

November 25, 2002 - Mary

  - Hacked Typing module to ignore empty text node expressions, which
    are never constructed, and therefore never contribute to a
    type. Empy text nodes exist solely to delimit enclosed expressions,
    and therefore cannot be removed during normalization.  Instead, we
    ignore them during static typing.  This code must be revisited
    after we implement validating element constructors.

November 20, 2002 - Jerome

  - Added some support for simple updates and conditional updates.
  - Fixed parsing problem for document type declarations.

November 18, 2002 - Jerome

  - Added support for xml:space directive.

November 18, 2002 - Mary

  - Changed config/Makefile.gen to use OCAMLBIN variable, if it is
    set.
  - Fixed dynamic semantics of attribute and element
    constructors and of whitespace in constructors
  - Fixed static typing of attribute and element constructors -- is
    now conservative: raises error if content type is mix of values and
    nodes. 
  - Changed element node constructors in datamodel to collapse
    adjacent text nodes and eliminate empty text nodes. 
  - Added tests to XQueryUnit (see xqueryunit/Changes)

  - *** Noto Bene: the document loader is still in flux, therefore 
    loaded documents do not "round-trip" correctly when serialized ***

November 17, 2002 - Jerome

  - Added support for 'at $i' in for expressions.

November 16, 2002 - Jerome

  - Fixed some of the compilation phase for the toplevel query
    prolog/context.
  - Fixed bug in galapi. Made sure the externally bound variables are
    added to the context *before* compiling the XQuery prolog.
  - Added support for multiple kinds of statements in the toplevel
    expressions. This is mostly transparent now, but is required to
    add support for updates.
  - Added support for keys and keyref operations. Syntax to define a
    key is:

    define key "KeyName" { Expr1 } { Expr2 }

    "KeyName" is a string literal used to identify the key used. Expr1
    is used to define the set of nodes on which the keys are
    defined. Expr2 is an expression which has implicitly each node
    from Expr1 bound as the context node '.' and describe the
    component used to define the key. What ever Expr2 returns, the
    string value is used for to define the key.

    glx:keyref("KeyName", Expr3) is a special function which can be
    used to access the XML node bound to the result of Expr3 in the
    key whose name is "KeyName". The first parameter of that function
    *must* be a string literal.

November 14, 2002 - Jerome

  - Added some new functions in Galapi.

November 13, 2002 - Jerome

  - Fixed bug in lexer having to do with string buffers not properly
    reset during lexical state changes. This fixes problems in
    attributes content escaped as XQuery expressions.
  - Added separate module to handle calls to the error()
    function. Xerror, which can now contain a proper data model value
    as a message.

November 12, 2002 - Mary
  - Testing CVS server

November 5, 2002 - Jerome

  - Made fixes to projection now working for: child::, attribute::,
    descendant, self::, descendant-or-self::. Improved projection for
    attributes.

November 4, 2002 - Jerome

  - Fixed XMark query 14.

November 2, 2002 - Jerome

  - Fixed XMark query 19.
  - Fixed wf-sax to use the monitor.
  - Fixed a bug in the makfile which was building a symbolic link to
    the wrong executable for wf-sax and validate.
  - Fixed tabulation in the monitor.
  - Added a mini update parser. Not connected to the rest of the
    system yet.

October 29, 2002 - Jerome

  - Cleaned up the top of the XQuery AST. Removed the notion of module
    list. Fixed the parser and toplevel loop accordingly. Now a single
    file contains only one module.

    The ';' syntax is now only used to separate several statements
    (currently only single expressions) at the end of the module.  The
    only difference visible externally is that one cannot write prolog
    components after a given expression anymore.

October 28, 2002 - Jerome

  - Added statistics on the number of data model nodes created.
  - Fixed bug in monitoring for DOM-like parsing and loading of
    document.

October 27, 2002 - Jerome

  ** Welcome to the big fall cleanup! **

  ** This is a large reorganization which will require people to
  ** migrate. Recommended is to do a full checkout again, except for
  ** people who made actual changes. More changes in the terminology
  ** inside the Datamodel and Datatypes modules are still expected.

  - Fairly deep cleanup of the directory structure, file names, and
    module organization. Here is the new organization, in compilation
    order:

        o ./base	 --  Basic modules, not really XML specific

            XML related modules have been moved later in the
            compilation chain. Remains:

              Gmisc -- Additions to Caml's stdlib
              Id -- Identifiers generation
              Pool -- String pools
              Finfo -- Handles locations in file
              Error -- Galax error handling
              Pervasive -- Wrapping of the pervasive.xq file
              Conf -- Global variables and configuration parameters
	      Args -- Dealing with function arguments
	      Whitespace -- Operations on whitespace in strings
	      Occurrence -- Operations on occurrence indicators

        o ./fsa		 -- Generic package for finite state automatas

            This package has been made generic. Most modules only
            contains functors, parameterized by an alphabet of
            letters, which will make the transitions for the
            automatas.

              Fsautil -- Functions for the rest of the FSA package
	      Nfa -- Non deterministic FSAs
	      Dfa -- Deterministic FSAs
	      Regexp -- Regular expressions and Glushkov construction
	        from regular expressions to NFA and DFA. This module
	        relies on the fact that states in the automata should
	        be positions in the regular expression, as specified
	        in the Glushkov construction. This also contains
	        intersection and reachability (emptiness check) for
	        the automata(s)

        o ./namespace	--  Operations on XML names, and namespaces

            All basic support for XML names is located here.

	      Qname_lexer_{encoding} -- Lexer for qualified names for
  	        a given character encoding
              Qname_lexer -- toplevel access to the QName lexer(s)
	      Namespace -- Support for namespace names, local names,
  	        namespace prefixes and URIs. It also implements
  	        namespace environments
              Sym -- Support for 'compiled' names, i.e., names stored
  	        somewhere in a string pool
              Resolve -- Support for namespace resolution. Given a
  	        qname and a namespace environment, it can return you a
  	        resolved name

        o ./datatypes	--  Support for XML Schema datatypes

            Support for XML Schema datatypes should now all be located
            in this directory.

              Decimal -- xsd:int, xsd:integer, xsd:decimal
	      DateTime -- xsd:dateTime, xsd:date, xsd:time
	      AnyURI -- xsd:anyURI
	      Datatypes_lexer -- Lexer for the various datatypes
	      Datatypes -- Top-level support for datatypes. Contains
	        implementation for atomic values in the XML Query data
	        model

        o ./parsing	--  Abstract Syntax Trees, XML&XQuery parsers

            The parsing 

              Xml_ast -- XML Abstract Syntax Tree (if materialized)
	      Xquery_ast -- XQuery Abstract Syntax Tree
	      Print -- Printing functions for the AST's
 	      Xml_astutil -- Construction and misc. operations on the
	        XML AST
 	      Xquery_astutil -- Construction and misc. operations on
	        the XQuery AST
              Lexutil -- Lexing utilities, including support for
  	        multiple lexical states
              Xquery_parser -- XQuery parser
              Document_lexer_{encoding} -- XML Document prolog lexer
	        for a given character encoding
              Opening_tag_lexer_{encoding} -- Opening tag lexer for a
	        given character encoding
              Closing_tag_lexer_{encoding} -- Closing tag lexer for a
	        given character encoding
              Text_lexer_{encoding} -- Text content lexer for a given
	        character encoding
              Lexers_{encoding} -- Toplevel access to the lexer(s) for
                a given encoding
              Xmldecl_lexer -- XML declaration lexer
	      Type_lexer -- XQuery type system lexer
	      Xquery_lexer -- Main XQuery lexer

        o ./monitor	--  Monitoring of Galax activity

              Monitor -- Used to monitor CPU and memory consumption

        o ./schema      --  XML Schema and XQuery type system

            One step further in typing, now support for XML Schema and
            the XQuery type system. A major rewrite of this package is
            expected [Support for named typing].

              Glushkov -- Wrapper on top of the FSA
                package. Implements low-level inclusion and
                intersection over content models
              Schemamapping -- Mapping from XML Schema to the XQuery
                type system
              Schemautil -- Various operations on types. Pretty messy
                in there...
              Theta -- Subsumption mappings between type names, used
                during the subtyping algorithm
              Subsumption -- Main subtyping module
              Simplification -- simplification over computed types
              Typeutil -- More operations on types... used in the rest
                of the type checking

        o ./normalization -- Query normalization

            This module corresponds to normalization as specified in
            the XQuery 1.0 and XPath 2.0 Formal Semantics.              

  - Clean up of terminology in several modules, notably datatypes, and
    data model. Here are the detailed changes:
      --> Bound module		becomes Occurrence module
          Bound.bound type	becomes Occurrence.occurs type
      --> Simple module		becomes Datatypes module
      --> Data in Simple	becomes Atomic in Datatype
      --> _text in Simple	becomes _untyped in Datatype
      --> oforest 		becomes Sequence
      --> InstanceNode		becomes Node
      --> Instance		becomes Item
      --> InstanceForest	becomes Value

  - Node names are aligned with names in the XQuery data model
    document.
      --> ElemNode		becomes ElementNode
      --> AttrNode		becomes AttributeNode
      --> DocNode		becomes DocumentNode
      --> NSNode		becomes NamespaceNode
      --> PINode		becomes ProcessingInstructionNode

  - Fixed bug in toplevel Galax processing files in reverse input
    order when given on the command line! :^)

October 26, 2002 - Jerome

  - Made a number of changes to try to understand the reason the data
    model is so big in memory. Problems are in part due to the way
    memory measure is done, which makes difficult to measure the
    actual size of a memory object.

      o Added some printing of the size of the string
        pools.
      o Simplified the way some of the loading with projection is done
        to avoid registering names which are discarded by projection.
      o Changed the SAX events to use QNames instead of symbols from
        the pools. This is better since this way events do not get
        recorded in the string pools. SAX event are discarded anyway.

October 25, 2002 - Jerome

  - Fixed bug in xf:index-of which was not using node identity has it
    should have. This bug was triggered by projection, since it was
    removing many other nodes and making it more likely that other
    nodes would have the same structure among siblings.
  - Fixed monitor to print the phase name for each line in the
    summary.
  - Fixed the document function to make sure it calls the monitor.

October 21, 2002 - Mary

  - Implemented element constructor semantics as described in Section
    3.7.1.  An item sequence containing both non-attribute nodes and
    atomic values is converted to a node sequence before applying
    element constructor.  Effects:
      datamodel/datamodel.mli   
      evaluation/xaeval.ml

  - Implemented typing for element constructors: an item-sequence type
    containing both non-attribute node types and atomic value types is
    converted to a node-sequence type before typing element
    constructor.  Effects:
      typing/typing.ml

  - NB: The above changes will be effected by validating
    constructors. 
  
October 17, 2002 - Amelie

  - Changed projection print function to output to formatter
  - Fixed some projection bugs

October 16, 2002 - Jerome

  - Added fairly extensive support for monitoring memory and CPU
    activity in Galax. Monitoring can be turned on with [-monitor on]
    along with the now standard [-output-monitor filename].

    This gives detailed memory and CPU consumption for each processing
    phase. It also allows to wrap monitors around sub-activities, for
    instance this is used to monitor document loading, which is
    typically an expensive process.

    There is a new directory and interface for this in ./monitor.


***** Galax Release 0.2.0 October 15, 2002 *****

October 14, 2002 - Jerome, Vasilis, Mogan

  - Hooked up minimal support for the xs:dateTime datatype. Not
    enabled in the system yet.

October 13, 2002 - Jerome

  - Cleaned up the Caml API.
  - Added possibility to add a global XQuery variable from the Caml
    side.

October 12, 2002 - Jerome and Amelie

  - Implemented projection for a complete query.
  - Added some regression tests for projection in the
    ./regress/projection directory. Those trip some bugs in loading
    with projection when // is involved. Those tests are not done run
    by default yet. use make project in the ./regress directory
    instead.
  - Added options in the Galax interpretor related to projection.
        [-print-projection on] prints the path used for projection.
        [-print-optimized-projection on] prints the path used for
         optimized projection.
        [-output-projection filename] prints back the input document
        after projection.

October 11, 2002 - Jerome

  - Updated the STATUS file, notably the section on F&O.
  - Fixed bug for usecase tests under windows.
  - Fixed bug in makefile for ISO-8859-1 lexer.
  - Put all of the non standard functions in the glx: namespace. This
    impacts the following functions:
       glx:docorder
       glx:distinct-docorder
       glx:unique
       glx:optional
       glx:union-values
       glx:intersect-values
       glx:except-values
       glx:document-sax
       glx:document-validate
  - Remove xf:distinct from pervasive, obsolete.
  - Fixed some flag problem generated during printing of core
    expressions.
  - Added optimization as an option in the demo.
  - Aligned parameters in the demo with processing phases/printing
    options from the command line.
  - Moved the 'english text' in the web demo to be a comment at the
    beginning of the query expression instead of part of the form.
  - Fixed the web demo form field names.
  - Fixed some remaining bugs in printing of expressions in the Web demo.
  - Added value-based union, intersection and difference.
  - Added NEW file to record major changes between releases.
  - Changed release number to 0.2.0.
  - Renamed ./capi into ./galapi.
  - Created automatic packaging for the source and for the
    pre-installed binary versions.
  - Fixed symbolic links to the executables to be relative instead of
    absolute for better portability (e.g., moving the installed binary
    to another directory).

October 11, 2002 - Mary

  - Added support for arithmetic operators on xsd:integer
  - Added rules for promoting between xsd:integer and other numeric
    types. 

October 7, 2002 - Amelie and Jerome

  - Added support for document projection at loading time. Made
    preliminary tests.

  ** BREAKING NEWS Tuesday October 7th, 2002 6:30pm EDT: The first
     Galax query to run on 1Gb document is XMark usecase 1. It ran on
     a Linux 256Mb memory box, and took about 1 minute to give a
     (correct) result!

October 5, 2002 - Jerome

  - All the use cases and regression tests now work with -optim
    on. Included a third pass on the usecases which tests them with
    optimization on.
  - Fixed bug in recursive factor function used to compute the type of
    descendant. Was forgetting to keep document node types. This fixes
    a bug in the sgml use case.
  - Changed toplevel to turn -print-xml off automatically when using
    -dynamic off.
  - Complete fix to the treatment of quantified expressions.
     o Added support for multiple bindings in quantified expressions
       in the AST and the parser [Forgotten August 16th change].
     o Fixed pretty-printer to be pretty for quantified expressions
       and support multiple bindings.
     o Added quantified expressions in the core. Now some and every
       are preserved pretty much 'as is'. Multiple bindings are turned
       into nested unary some/every. This enables a nicer normalized
       form for many expressions and optimization. See below.
     o Changed normalization to preserve quantified expression in the
       core. Changed normalization of 'map_existential' to use
       existential quantification directly.
     o Added typing rules and evaluation code for quantified
       expressions. It seems the evaluation should be more efficient
       like this than the previous one using not(empty(for...)).
     o Added optimization rules for quantified expressions. Those are
       quite similar to for. Ones, this enables actual optimization of
       predicates and operators in XPath in many standard cases.
     o Added a few regression tests for quantified expressions.

October 4, 2002 - Mary

  - Added examples/ and examples/ml-api/ directories.
  - Minor changes to README and INSTALL.

October 3, 2002 - Jerome

  - Added JAVAINCLUDE parameter in the configuration makefile. This
    will allow users of varions JDK's to point to the correct include
    directory where jni.h is located. [Thanks to Lars Nilsson for
    pointing out the problem and the fix!].
  - Fixed README and INSTALL files.
  - Fixed serialization for floating point numbers, doubles and
    decimals, implementing XML Schema canonical value properly.

October 3, 2002 - Jerome & Mary 

  - Added missing test results in regress/
  - Added tests regress/typeswitch_empty*.xq    
  - Fixed numerous bugs in typing/typing.ml related to factored types 
    (for CEFor, CESort, CETypeswitch, distinct-values, CEUnordered)
  - Fixed typing bug in which some core expressions were not typed
    o Caused optimization to fail

October 2, 2002 - Mary

  - Fixed normalization & dynamic semantics of function application

    o Added dynamic check that actual argument is instance of
      corresponding formal type and that return value is instance of
      return type.  

    o Added check at function application site (not in function body)
      so that rule can be applied consistently to user-defined and
      built-in functions.  (Could be moved into function body for
      user-defined functions.)

      (NB: Dynamic type check is only necessary if static analysis is
       off.  We should generate dynamic checks conditionally, but will
       wait to do this until context issue is resolved.)

    o TODO : SequenceType matching not fully implemented (DTTypeRef)
      in Xaeval.of_type_by_name and Commontype.small_type_unit.

  - Updated stdlib/pervasive.xqp to use data types instead of type
    names in function signatures
  - Fixed signature of item-at() & corresponding tests
  - Added Galapi.eval_xquery_fun 

October 2, 2002 - Jerome

  - Fixed compilation chain to work smoothly under Windows. Tested
    with MinWG.

    ** Warning: for some obscure reason, some wrong tests have been
       committed. No record was found. **

September 30, 2002 - Jerome

  - Fixed the document() function to keep track of previously opened
    documents. Multiple access to the same document result in the same
    data model instance.
  - Removed dependency to physical pervasive.xq file in the Galax
    directory. The pervasive file is now compiled directly within the
    Galax engine. This allows installation of the system anywhere
    without requiring knowledge of the installation directory at
    compile time. This will be good for binary packages.

September 27, 2002 - Jerome

  - Implemented F&O Section 15. Context Functions:
        xf:context-item
  - Implemented F&O Section 13. Functions and Operators on Nodes:
        xf:name
        xf:namespace-uri
        op:node-equal
        xf:deep-equal
        op:node-before
        op:node-after
        xf:copy
        xf:root
  - Finished implementing F&O Section 14. Functions on Sequences:
        xf:empty
        xf:exists
        xf:distinct-nodes
        xf:distinct-values
        xf:insert
        xf:remove
        xf:subsequence
	xf:sequence-deep-equal
	xf:sequence-node-equal
	op:union
	op:intersect
	op:except
        xf:count
	xf:avg-{int,decimal,float,double}  // Monomorphic versions...
  - Removed xf:length obsolete.
  - Added a glx: prefix and URI for Galax additions to F&O, types,
    etc. URI for the Galax namespace is http://db.bell-labs.com/galax

September 27, 2002 - Mary

  - Tested CAST expression. 
    o See test matrix in regress/evaluation/cast_test_matrix.txt
  - Implemented type promotion rules for 
    o Arithmetic operators
    o Value comparison operators
    o Updated regression tests and usecases effected by this change

September 25, 2002 - Mary

  - Changed version to 0.2
  - Revised STATUS document
  - Changed default serialization to emit item sequences as XQuery
    sequences (i.e., ',' delimited)
    o Updated regression tests and usecases effected by this change
    o Sec 3.7.1: Element Constructor
      Added instance_forest_to_node_forest to datamodel.ml to 
      implement conversion of item sequence to node sequence.
  - Added xq: prefix and namespace for serializing <xq:result/>
  - Implemented 
        xf:string()
        xf:string(item?)
        xf:error(item?) - now reports user-defined error

  - Fixed expression printing bug in website/demo

September 24, 2002 - Jerome

  - Fixed bug about the default namespace not being remembered.
  - Fixed bug about default namespace for attributes, in printing and
    path expressions.

September 22, 2002 - Jerome

  - Implemented F&O Section 9. Functions on QNames:
	xf:expanded-QName
	xf:get-local-name-from-QName
	xf:get-namespace-from-QName
  - Revised F&O Section 7. Functions and Operators on Booleans:
        xf:true
        xf:false
        xf:not
        op:and      (* Missing in F&O *)
        op:or       (* Missing in F&O *)
  - Implemented F&O Section 14. Functions on Sequences:
        op:to
        xf:boolean
        op:concatenate
	xf:item-at
  - Implemented range expressions, backed up by the op:to operator.

September 21, 2002 - Jerome

  - Fixed function normalization to lookup function *after* namespace
    resolution instead of based on the namespace prefix. This fixes
    errors on type constructors (xs:string, etc.) and special
    functions (xf:position, xf:error, etc.).
  - Removed "ERROR" function obsolete.
  - Fixed built-in simple type table so that both xs: and xsd:
    prefixes would work in casting and constructor operations.
  - Made sure all errors are reported correctly at the toplevel.
  - Made names of errors and error messages more consistent.

September 20, 2002 - Jerome

  - Turned xf: into op: for operators.
  - Merged 3 source branches from Amelie, Mary and Jerome.
  - Implemented URI operations:
      xf:base-uri    (partially -- See XQuery issue)
      xf:resolve-uri

September 20, Jerome and Amelie

  - Added first prototype of path analysis.
  - Created new ./analysis directory.
  - Changed compilation order to have normalization before
    datamodel. The ./analysis directory is compiled between
    ./normalization and ./datamodel.

September 20, 2002 - Mary

  - Implemented -serialize-wf switch

September 19, 2002 - Mary

  ** MAJOR CHANGES TO COMMAND LINE INTERFACE **
  - See documentation in galax/README for new options
      o Changed toplevel/xmlquery.ml capi/toputils.ml
      o Updated regress/runtest.perlp, usescases/runtest.perlp,
        & website/demo to use new options

  - Implemented xsd:TYPE constructor functions

September 19, 2002 - Jerome

  ** CHANGE TO COMMAND LINE EXECUTABLE NAME **
  - Changed the name of the command line tools:
      xmlquery --> galax      (main command-line executable)
      galax    --> galaxtop   (Caml interpretor with Galax embedded)
  - Integrated netstring version 0.10.1 directly into Galax. This
    removes dependencies to both netstring and findlib, and is the
    base for URL support in Galax. It adds dependency to both Unix and
    Str modules in Caml, but this is probably inevitable (e.g., for
    date and time support).
  - Moved the character sets to ./tools/charsets. All third-party code
    is now located in the ./tools directory.

September 18, 2002 - Jerome

  - Fixed the expected output for the usecases which do not use static
    typing. (XMARK, PARTS, and NS).
  - Started implementing F&O, version August 16th. Added:
      xf:string
      xf:local-name
      xf:node-name
      xf:node-kind

September 18, 2002 - Mary

  - Implemented new normalization rules for Atomization and Effective
    Boolean Value (EBV).  These rules enforce stricter static typing
    than previous rules.
  - Some regression & usecase tests that previously raised dynamic
    errors now raise static errors. 
      o Disabled static typing for NS, PARTS, and XMARK usecase tests
  - Modified demo
      o Disabled static typing for NS, PARTS, and XMARK usecases
        See usecases/Usecases file
  - Added Static_Type_Error to base/error.ml

September 17, 2002 - Mary & Jerome

  - Changed static typing of typeswitch to allow branches to fail. The
    overall typeswitch fails if all the branches fail.
  - Changed the static typing to make sure all sub-expressions are
    annotated with a type. All ill-typed expressions are annotated
    with the 'none' type. This should make the optimizer happy.

September 15, 2002 - Jerome

  - Fixed the Web demo to work with the new use case organization.
  - Slightly reorganized the use cases, for the needs of the automatic
    testing:
      o All XML and XSD files are now in ./usecases/docs
      o All usecases have now a _usecase.xq prefix
  - Fixed bug in the parser which should parser a sequence of toplevel
    expressions as a single expressions sequence instead of a list of
    expressions.
  - Finished fixing the pretty printer for XQuery
    expressions. Precedence should work fine now.
  - Tested the pretty printer through boot-strapping, using the
    following steps:
      o parsed use cases.
      o pretty-printed back all of the expressions.
      o checked that the pretty-printed expression parse back and run
        the use cases as the original query.
  - Added test script in the use case directory. This test runs all of
    the working use cases, and also performs test of the
    pretty-printer through boot-strapping (as described above).
  - Hooked up testing of the use cases to the toplevel 'make tests'.

    ** From now on, each commit must be preceded from a make tests'
       directly at the toplevel. **

September 12, 2002 - Jerome

  - Removed binary operation 'to' obsolete, already had been replaced
    by RangeExpr.
  - Removed attribute list from computed element constructor's AST,
    unused.

September 11, 2002 - Jerome

  - Fixed many problems in the pretty printer for expressions. Started
    implementing support for precedence in the pretty printer.

September 8, 2002 - Jerome

  - Fixed error message for unallowed castings to xs:boolean, xs:integer
    and xs:int.
  - Bug on sorting with empty greatest was already fixed.
  - Fixed all of the namespace support for functions. Implemented
    support for default function and element namespaces.
  - The default namespace for functions is now the one from F&O. xf:
    and op: are now recognized by default.
  - Reorganized static typing for functions. Built-in functions with
    specific typing rules are now handled separatly, through a table
    lookup.
  - Fixed various problems with namespace handling. Added xmlns: as a
    built-in namespace prefix. Added xsi: as an XML Schema
    namespace prefix, will be used as soon as support for XML Schema
    named typing is added. Fixed all built-in namespaces
    (xml,xmlns,xs,xsd,xsi,xf,op) to the correct URIs.

    ** TODO: Better support of nested namespaces inside element
    constructors is still needed. Some bugs are lurking there. **

September 6, 2002 - Gargi and Jerome

  - Installed latest version of the optimization rules.
  - Fixed all of the optimization regression tests so all errors are
    recorded as bugs.
  - Fixed the regression test program so that -all is working.

    ******************************************************************
    ** All committed changes should now use be preceded to a run of **
    ** the regression tests with the -all option                    **
    ******************************************************************

  - Moved the GALAXSRC variable from ./config/Makefile to

    ./regress/Makefile.conf

    This is now setup by default to .. so the standard user should not
    have to do anything about it, except if he wants to move the
    regression tests to a different location.

  - Fixed the main Makefile so that it will still work if there is no
    API to compile.
  - Fixed output formatting for the regression tests, so it will print
    a simple report on stdout.
  - Added a "make tests" target in the main makefile for Galax users
    will be able to run them when they compile the system. Document it
    in the INSTALL file.

September 4, 2002 - Jerome and Amelie

  - Fixed the SAX parser wrt. new XML lexers. Notably, whitespace in
    document root and opening tags was not handled properly.
  - Implemented loading from well-formed documents to the XML Query
    data model from SAX streams.
  - Added a document-sax(...) function in the built-in library which
    uses SAX parsing instead of DOM parsing.

August 26, 2002 - Jerome

  - Added support for multiple character encodings.

    This required pretty deep changes in the compiling
    process. Notably, separate Makefiles are used for various
    character encodings. The configuration makefile can be used to
    select a subset of the available encodings. The compilation
    process then gets parameterized by the selected encodings. The
    -linkall flag *must* be passed to the Caml compilers.

August 23, 2002 - Jerome

  - Added back support for escaped names (e.g., :item, :for) in the
    XQuery parser.
  - Simplified the compilation process to two simple steps: "make",
    followed by "make install".
  - Implemented UTF-8 and ISO-8859-1 support at the parsing level.
  - Implemented appropriate handling of XML character sets at the
    parsing level.

    Note:
      This change required some pretty serious rewrite of each lexer,
      which is not partially generated from auxiliary files 'a la'
      PXP.

      Character sets and pre-compilation tools are courtesy of Gerd
      Stolpmann, and Claudio Sacerdoti Coen.

  - Fixed all bugs related to handling of characters in the parsing
    tests.

August 21, 2002 - Jerome

  - Factorized the XML-DOM and XQuery parsers. Remove the original XML
    parser and lexer. Added document prolog lexer.

August 20, 2002 - Jerome

  - Fixed the syntax for the query prolog.
     Works: declare namespace
            default element namespace
            import schema namespace prefix = uri
            import schema default element namespace = uri

August 19, 2002 - Mary

  - Fixed bug in normalization of "<<" and ">>"
  - Fixed bugs in regress/evaluation/node_comp*
  - usecases/
    - Pre-August usecases compile & run, but are not up-to-date
      w.r.t. 18 August Usecase document
  - website/
    - Recompiled and tested
    - Previously deleted functions:
      Gmisc.load_file added to demo/demo_conf.ml 
      Toputils.print_{core_}xquery_text_from_string added to
       demo/querycgi.ml

August 15, 2002 - Jerome, Mary, Gargi

  - Traded a bug for another: Fixed typing of axis so it would not
    fail for complex content. Currently atomic types are turned into
    text node types, but the adjacent 'text node types' are not
    merged.
  - Fixed parser to deal with the new grammar (August 16th) for
    expressions. Notably:
        o for/let/quantifiers now have an optional type assertion.
        o Removed type assertions (assert as).
        o Added unordered expression which does nothing for now.
        o Changed the syntax of typeswitch.
        o Changed 'sortby' to 'sort by'
  - Added optimization which removes type assertions when the static
    inference shows it is a no-op.
  - Changed makefile so that xmlquery et al will be links to the
    native code version if available.

August 14, 2002 - Mary and Gargi

  - TODO: dynamic type checks in For/Let is not implemented

August 13, 2002 - Jerome

  - Split the XQuery lexer into smaller parts. This will allow further
    work on the parser and enable factorization with the XML parser.
      [ Files created:
          o opening_tag_lexer.mll for opening tags
       	  o closing_tag_lexer.mll for closing tags
       	  o text_lexer.mll for XML text content
       	  o type_lexer.mll for type declarations ]

August 13, 2002 - Gargi

  - Implemented generic optimizer which takes a list optimization
    rules as a parameters. This makes the optimizer easily extensible
    and more modular.
       [ Files created:
           ./evaluation/optimgen.ml
           ./evaluation/optimrules.ml
           ./evaluation/optimjudge.ml ]
       [ Files deleted:
           ./evaluation/optimutil.ml ]
  - Improved let optimization: now works for step expressions.
  - Added empty() and length() as "safe" functions [in the can_fail
    judgement].
  - Added optimization tests.
  - Indicated all BUGS properly in the regression tests.

August 12, 2002 - Jerome

  - Fixed malloc bug during parameter construction in the C API. Bug
    fix courtesy of Yana Kadiyska.
  - Fixed bug in parsing of single-quoted strings.
      [ Fixes:
          parsing/string_05_pos.xq
          parsing/string_06_pos.xq
          parsing/string_11_pos.xq ]
  - Fixed bug in parsing of doubles. Was forcing a digit after the
    decimal point.
      [ Fixes:
          parsing/double_10_pos.xq and
          parsing/double_11_pos.xq ]
  - Fixed bug in parsing of attribute content. Was allowing the
    forbidden character '<'.
  - Fixed bug in parsing of element content. Was allowing the
    forbidden sequence of characters ']]>'.
  - Fixed bug in parsing of element opening tags. Was allowing
    duplicate attributes!

August 6, 2002 - Jerome

  - Fixed bug in pretty printer for 'every', 'some', and 'if'.

August 1st, 2002 - Mary

  - Regression tests in regress/evaluation/:
     o re-ran each test by hand; 
     o populated expect/ directory with expected results
     o annotated first line of each test query that triggers bug 
     o regress/results reports 24 files with bugs
  - Fixed bug in pervasive.xqp -- xs:AnyDocument not contained in xs:AnyNode
  - Fixed bug in load.ml -- document node was not assigned node
    identifier during loading
  - Fixed bug in typing.ml -- type check of data() incorrectly raised
    error when applied to document {}
  - Fixed bug in datamodel.ml -- elemNodeWithValue incorrectly created
    empty text node
  - Fixed bug in xmlquery.ml -- print errors to stderr

August 1st, 2002 - Jerome

  - Now prints input XQuery using pretty-printing. Removed
    Gmisc.load_file function which was not working under
    Windows. Fixes Windows bug raised by Phil [ xmlquery was not
    working properly on windows file with new lines. ]

July 31, 2002 - Jerome

  - Implemented the string-value() Data Model accessor.

July 27, 2002 - Jerome

  - Removed '=>' operator, removed from XQuery.

July 26, 2002 - Jerome

  - Fixed bug in parsing of SequenceType for occurrence indicator
    '*'. Was getting confused with multiplication.

July 25, 2002 - Jerome

  - Added support for text node constructor.
    Syntax : text { Expr }
    Semantics : the expression should evaluate to a single string. It
    creates a new text node with the corresponding string as text content.
    Limitations : it does not deal with string concatenation and still
    allows empty strings.
  - Added text representation for element with a simple type. This
    allows to use the XPath axis on e.g., <a> { 1 } </a>.
  - Fixed bugs in typing for XPath axis in the case of simple type
    elements and attributes.

July 24, 2002 - Jerome

  - Added a 'concats' function which concatenates a list of string.
  - Fixed bug in parsing of numeric types.
  - Fixed bug in factoring for sortby. Cleaned up factoring functions.

July 22, 2002 - Mary

  - Changed -noverbose option to -verbose option. (Terse output should
    always be the default)
  - Changed serialization of query output to always produce a document,
    contained in <xq:result>.  This was the default, but was changed
    at some point.
  - Updated regress/runtest.perl to look for expected output in
    expect/ subdirectory.
  - Updated regress/evaluation/expect
    TODO: expected output for regress/typing 
  - Re-ran regress/evaluation to check for errors. 

  *** NOTE: All tests in evaluation/ should be run and
      regress/results file should be examined before committing changes
      to CVS repository ***

  - Fixed bugs in: 
    - datamodel/datamodel.ml in isInstanceValue
    - evaluation/xaeval.ml 
       
      in of_type_by_name :
        did not match datatype Node with PIs or comments; 
        incorrectly matched datatype node with atomic values
      in evaluation of CEAxis: $dot may be bound to atomic value
  - Changed: 
    - parsing/print.ml: added printf_datatype_kind 
    - typing/typing.ml: in typing of CETypeswitch case clause, when
      intersection of expression's type and datatype is None, we type
      the case-clause body "conservatively" by using the data type of
      the case clause.
    - schema/schemautil.ml:
      Added function factor_type_in_typeswitch, which factors the type
      of the argument expression.  For the empty type TEmpty,
      factor_with_units (used to type for and sort expressions)
      returns (TNone, 0, 0), which is not applicable to typeswitch.
      In this case, the factored type is converted back to TEmpty.
    - TODO : normalization rule for 'treat as'. 
 
July 19, 2002 - Zewei

  - Put in a temporary hack to protect against the backslash bug of
    ocamldep for the benefit of Yu Xu

July 18, 2002 - Gargi
  
  - Completed typeswitch simplification by adding intersects function and
    and fixing code for remove-empty-clauses
  - Extended let simplification by adding a new rule
  - Implemented two rules for "for" simplification
  - Changes made to typeswitch static type evaluation


July 15, 2002 - Zewei

  - Separated the installation destinations of html documents and cgi
    programs.  This is necessitated by multi-user systems that have
    restrictions on where a user can place cgi programs.
  - Updated documentation to reflect the changes
    Website/README content is replaced with a reference to ../INSTALL
    Website/INSTALL content is replaced with a reference to ../INSTALL
    INSTALL is updated to distinguish the configuration of web server
    from configuration of website/demo wrt to a pre-configured server.
  - All code changes are confined to website/demo
  - Added a local module, Demo_conf, to provide run-time parameterization
    (based on a config file) of:
        DEMOSITE (html document directory)
	DEMOCGIBIN (cgi program directory)
	CGIBIN_PREFIX (how the webserver refers to the per user cgi-bin)
  - Changed files: 
      Makefile (generates the configuration file, among other changes)
      Makfile.config (only changes the user should be aware of)
      demo_conf.ml[i] (new, defines demo_init: string -> unit)
      viewxml.ml
      querycgi.ml
      usecase.ml
      template.html
  - TODO: demo_init reads in parameter definitions from a config file.  
    It has to be performed at each toplevel entry to the demo program 
    suite, namely: querycgi.ml, viewxml.ml, usecase.ml.  The module 
    base/conf.ml hardcodes a number of parameters at compile time.  
    To use the configuration file approach here would improve it.
    But there is currently no convenient entry point like an init
    function.  It can be added.  Should it?
  - Also, does it make sense to install LICENSE INSTALL STATUS in
    the $(WEBSITE)?

July 12, 2002 - Zewei

  - Separated the installation destinations of html documents and cgi
    programs.  This is necessitated by multi-user systems that have
    restrictions on where a user can place cgi programs.
  - Updated documentation to reflect the changes Website/README
    content is replaced with a reference to ../INSTALL Website/INSTALL
    content is replaced with a reference to ../INSTALL INSTALL is
    updated to distinguish the configuration of web server from
    configuration of website/demo wrt to a pre-configured server.
  - All code changes are confined to website/demo
  - Added a local module, Demo_conf, to provide run-time
    parameterization (based on a config file) of:
        DEMOSITE (html document directory)
	DEMOCGIBIN (cgi program directory)
	CGIBIN_PREFIX (how the webserver refers to the per user cgi-bin)
  - Changed files: 
      Makefile (generates the configuration file, among other changes)
      Makfile.config (only changes the user should be aware of)
      demo_conf.ml[i] (new, defines demo_init: string -> unit)
      viewxml.ml
      querycgi.ml
      usecase.ml
      template.html
  - TODO: demo_init reads in parameter definitions from a config file.
    It has to be performed at each toplevel entry to the demo program
    suite, namely: querycgi.ml, viewxml.ml, usecase.ml.  The module
    base/conf.ml hardcodes a number of parameters at compile time.  To
    use the configuration file approach here would improve it.  But
    there is currently no convenient entry point like an init
    function.  It can be added.  Should it?
  - Also, does it make sense to install LICENSE INSTALL STATUS in
    the $(WEBSITE)?

July 12, 2002 - Gargi, Jerome and Mary

  - Fixed bug in static typing of typeswitch: was sometimes not doing
    typing of certain branches of the typeswitch.

July 10, 2002 - Jerome

  - Fixed bug in subtyping for document nodes.

July 9, 2002 - Jerome

  - Added '-version' option on the command line tool.
  - Made '-verbose' the default. Verbose printing can be turned off
    with '-noverbose'.
  - Hooked-up the optimization module to the system. Added two options
    in the command-line interface: '-optim' turns optimization on, and
    '-print-optim-core' prints out the optimized core expression.
  - Removed the temporary 'test_opt' command line tool. The same
    behavior can be recovered through: xmlquery -notype -noeval -optim
    -print-optim-core xqueryfile
  - Added PI's and Comments in the XQuery core. Fixed normalization
    rules for PI's and Comments.
  - Added support for document node constructors (document { ... }).
  - Added normalization for 'instance of'.
  - Removed instance of only obsolete.

July 8, 2002 - Jerome

  - Removed many obsolete files and directories from the CVS
    repository.

July 8, 2002 - Gargi

  - Implemented the can_fail judgement for the let expression. Other
    than the cast-as, treat-as and error expressions, all user-defined
    functions and some of the built-in functions have been defined to
    fail.  Note: Most of the built-in functions can fail if the static
    typing is turned off.

July 3, 2002 - Jerome

  - Added preliminary support for updates in Galax:
     o Added low-level update operations at the data model level.
     o Added marker in the data model for deleted nodes.
     o Made sure that all accessors on nodes fail for deleted nodes.

June 28, 2002 - Jerome

  - Added user-level API for Galax in ML. Name of the module is
    Galapi.
  - Fixed a bug in function calls normalization: was not raising an
    appropriate error for function calls having a mismatch with the
    function signature.

June 27, 2002 - Gargi, Mary and Jerome

  - Added mini optimization module. This supports recursive
    simplification of unused let bindings. Added toplevel test_opt to
    test optimization.
  - Reorganized the ./algebra directory into three directories:
      o normalization
      o typing
      o evaluation
    Everything about optimization should go into the ./evaluation
    directory.
  - Removed the type handle from the 'expr' type (XQuery AST).
  - Cleaned up normalization to initialize the core expressions type
    handles to None.

June 26, 2002 - Jerome

  - A number of fixes to the demo.
  - Fixed bugs in the XML parser: added support for <!DOCTYPE
    ... >. Added support for comments and pis in the document root.

May 29, 2002 - Jerome

  - Implemented docorder() function which sorts by document order.
    Note: implementation maps that function directly to a core sort
    expression using '.'.

May 28, 2002 - Jerome

  - Fixed the dynamic and static semantics of sortby.
  - Added support for 'empty greatest' and 'empty least'.
  - Added support for stable and unstable sort.
  - Fixed a bug in the data() function.
       was : data(<a/>) = "" 
       now : data(<a/>) = ()
  - Cleaned up the organization of the evaluation module.

May 26, 2002 - Jerome

  - Implemented ancestor and ancestor-or-self axis.
  - Fixed typing for parent, ancestor, descendant, ancestor-or-self.

May 25, 2002 - Jerome

  - Turned == into is
          !== into isnot
  - Removed before and after unused.
  - Implemented follows and precedes.

May 24, 2002 - Mary and Jerome

  - Fixed the new Web site.

May 14, 2002 - Jerome

  - Fixed bug in parsing of comments in XML documents and XQuery
    file. Was allowing '--'. Was closing comments with ' -->' instead
    of '-->'.

May 5, 2002 - Jerome

  - Fixed error messages to print file and line number where the error
    occured. NOTE: Some of the typing errors are behaving strange:
    raising error() dynamically instead of actual type error. TODO:
    remove ./parsing/line.mll from the repository. Not needed anymore.
  - Added check for matching open/close tags in SAX parser.


***** Galax Release 0.1-alpha April 25, 2002 *****

April 25, 2002 - Mary

  - Added Galax website in website/.  Includes demo subdirectory.
  - TODO: remove demo/ subdirectory, add ptrs to other XQuery
    implementations, track downloads, add intro slides to 'about galax'.

April 24th, 2002 - Jerome

  - Fixed bug in deep copy, was wrongly returning a new internal Node
    instead of updating the dirty one.
  - Implemented support for parent.

April 23rd, 2002 - Mary and Jerome

  - Implemented support for node identity in the data model, loading
    and validation.
  - Implemented lazy deep copy in node construction. Copy is done only
    when an actual data model accessor requires it.
  - Fixed attribute content to be text instead of a node.

April 1st, 2002 - Jerome

  - Implemented a prototype SAX parser.

March 11, 2002 - Jerome and Mary

  - Added XMark use case. 19 out of 20 queries are working. Tested on
    up to 10Mb document on Jerome's Linux laptop with 128Mb memory.
  - Fixed pointers to use cases in the demo.
  - Implemented cast of UnknownSimpleType argument to type of second 
    operand in value comparisons
  - Fixed comparison operators on decimal, integer, int
  - Fixed attribute constructors: required type is sequence of atomic
    value; constructor expression automatically converted to sequence
    of atomic values. 

March 6, 2002 - Jerome and Mary

  - Implemented arithmetic on decimals, floats, and doubles
  - BUG: Automatic type promotion is not implemented.  User must use
    explicit casts. 
  - Fixed places where minOccurs maxOccurs could still build anything
    else than +, *, ?. Fixes bug in function call 3.
  - Fixed bug in typing for data(), children() and attribute(). Now
    working on union of atomic types.
  - Fixed confusion use case.
  - Replaced xs:AnyTree by xs:AnyContentModel, less missleading.
  - Fixed bug in validation for mixed content. Was not walking into
    text node content models properly.

March 5, 2002 - Jerome

  - Rationalized the use of subtyping in typing and mapping modules.
      Created a new module schema/typeutil.ml which contains subtyping
      wrapper functions. No modules in ./algebra should now rely on
      anything else but Schema, Simplify, and Typeutil (notably, they
      should not call Subsumption directly).
  - Fixed name of schema environment component in Xatenv.
    Changed tyenv operation into schemaenv
  - Changed printing of the core expression to use namespace prefixes.

March 1, 2002 - Mary 

  - Implemented formal semantics of:
       2.2.4 Function Calls

  - BUG: builtin-functions data(), nodes(), children(), etc.  are
    defined on singleton nodes, not on unions of singleton nodes.

  - BUG: xsd:int (minoccurs 3) (maxoccurs 3) not subtype of xsd:int+
    See regress/evaluation/function_arg_3_bug.xq

February 28, 2002 - Mary

  - Implemented formal semantics of: 

    2.5 Arithmetic Expressions
    2.6.1 Value Comparisons
    2.6.2 General Comparisons
    2.7 Logical Expressions
    2.11 Conditional Expressions
    2.12 Quantified Expressions

  - Fixed printing bugs in print.ml & error.ml
  - Added 'text', 'atomic', and 'numeric' as datatype keywords to
    lexer & parser.
  - Added AnyNumeric to builtin types in sym.ml
  - Replaced (+), etc. by numeric-add, etc. in pervasive.xqp

February 27, 2002 - Mary

  - BUG: In bltin.ml, built-in functions are denoted by unresolved QNames
      (prefix, localname), but should be denoted by expanded QNames
      (uri, localname). 

  - In maputil.ml, implemented formal semantics of : 
      optional_atomic_value
      optional_node
      effective_boolean_value
    which are used in mapping of arithmetic, (in)equality, logical,
    and conditional expressions.

  - Usecases
    Non-deterministic type bug in confusion 

February 27, 2002 - Jerome

  - Added support for "cast as". Works for:

    casting to xsd:string, xsd:int, xsd:integer, xsd:decimal,
               xsd:float, xsd:double, xsd:boolean, xsd:qname,
               xsd:anyuri.
  - Fixed bug introduced in the parser for e/text().
  - Fixed bug in typing for element constructors containing text.
  - Fixed bug in parsing of simple types, impacting both validation
    and casting.

February 26, 2002 - Jerome

  - Removed some 'long tokens' from the lexer, moving towards a
    simpler XQuery parser.
  - Fixed bug in schema loading/normalization introduced by the text
    node.
  - Fixed bug in the parser related to whitespace between keywords.

February 25, 2002 - Jerome

  - Added support for escaped QNames. Syntax is:
       ':'NCNAME
  - Added support for mixed content in validation. Mixed content is a
    content model with text nodes but no simple value.

February 25, 2002 - Jerome and Mary

  - Added support for datatype 'text' and for type model 'TText' and
    for typing of text nodes as 'text'. 
  - Changed definition of data() on nodes: 
 
      data(element { anySimpleType }) : anySimpleType
      data(element { anyTree }) : UnknownSimpleType*
      data(text) : UnknownSimpleType*

  - Added UnknownSimple value to data model

  [ Mapping of binary operators and functions still does not align
    with FS doc. ]

February 23, 2002 - Jerome

  - Fixed bug in validation for optional attributes screwing up
    content model processing (the system was thinking some content
    model were not deterministics passing over optional attributes).

February 20, 2002 - Jerome

  - Fixed bug in the empty namespace handling due to empty xml: prefix
    in the configuration.

February 12, 2002 - Jerome

  - Changed printing to print the default namespace even if default
    namespace is the empty namespace.
  - Fixed validation to preserve type names (required for named typing
    and legodb).

January 27, 2002 - Mary
 
  - Implemented "predicate truth value" semantics
    (step-qualifiers) for XPath expressions.
    (effects: algebra/{mapping,maputil}.ml)
  - Added () (DTEmpty) to Datatype production -- necessary to
    implement step qualifiers.
    (effects: algebra/{typing,xaeveal}.ml
    parsing/{xquery_ast,xquery_parser}.ml schema/schemautil.ml)
  - Fixed subtag bug in base/sym.ml
  - Removed "value" from stdlib/bltin.ml
  - Added NS usecase
  - Added "xml" namespace as builtin namespace in
    stdlib/pervasive.xqp; defined in config/Makefile.  Necessary for
    NS usecase.
    (effects: serialization in datamodel/datamodel.ml)

January 26, 2002 - Jerome

  - Cleaned XML Schema validation.

January 23, 2002 - Mary

  - Fixed some more use cases.

January 22, 2002 - Jerome

  - Added support for single quoted strings in the lexer. now 'hello =
    true' is a valid string.
  - Fixed bug in load_file basic operation. Function 'input' is not
    reading the complete file, using really_input instead.

January 21, 2002 - Mary and Jerome

  - Fixed bug in reachability module for FSA (was blowing-up). This
    fixes one known bug for the subtyping non terminating.
  - Fixed bugs in simplification module for union types.
  - Fixed type syntax from 'group' to 'type'
  - Fixed type syntax for document nodes to M. Kay's proposal. Fixed
    AST, mapping and typeswitch accordingly.
  - Fixed use cases XMP, Rel, Tree, Semantics, Confusion to the new
    syntax and mappings.
  - Fixed bugs in typing of name() and nodes() built-in functions.

January 16, 2002 - Lori

  - In the regress/parsing directory, 
    Renamed all the xmltest files to fit into our naming scheme:
    xxx_pos.xq, xxx_neg.xq, xxx_neg_expect.xq.
    Removed the badly named files from the repository, 
    and added the correctly named files to the repository.

January 14, 2002 - Lori

  - Updated config/Makefile. Has new PERL and GALAXSRC variables.
  - Updated top-level Makefile. New regress/runtest.perl target,
    generated from runtest.perlp and PERL variable.

January 14, 2002 - Mary
  - Fixed bug in pretty printing of errors

January 10, 2002 - Lori

  - Added regress/runtest.perlp. When the new config/Makefile is updated,
    and you run "make depend", it will generate runtest.perl.

  - Added regress/Makefile. This allows 
    make all, make parsing, make typing, make evaluation, make clean.

  - Added regress/parsing/xmltest_valid_*.xq and - valid xml expressions
    regress/parsing/xmltest_notwf_*.xq           - not well-formed xml exprs

  - Still need to commit top-level Makefile and config/Makefile. Must resolve
    conflicts first.

January 10, 2002 - Mary 

  - Added regress/evaluation/typeswitch* - to test new typeswitch 
          regress/mapping/               - start of XPath tests

  - Updated algebra/maputil.ml to support 'fortran'-style operators eq,lt, etc.

January 10, 2002 - Mary and Jerome

  - Fixed typing for typeswitch.  

    - Added datatype production and modified typeswitch (typing
      & evaluation) to operate over datatypes.

    - Added small_type(t1,t2) in lieu of intersect_type(t1,t2)
      as semantics of typeswitch.

    - In (1) type subsumption and (2) testing that a value is in the
      domain of a type (see: Xaeval.of_type_by_name), only symbols are
      compared. 

      For elements and attributes, it *only* compares for equality the
      element/attribute value's symbol/name with the type's
      symbol/name.  It does not recursively check the node's content.

      For simple values, it *only* checks that the type of the value
      and datatype are equal.  It should use type subsumption on
      simple types.

      Group references and 'element of type' not yet supported in
      typeswitch. 

  - Added a document unit symbol to sym.ml
  - In AST, added support for element of type (group and basetype)

January 6, 2002 - Phil and Jerome

  - Updated the pretty printer for types to deal with precedence.

January 4, 2002 - Jerome

  - Updated the mapping to deal with value comparisons: lt gt eq, etc.
      *** Mary: please double check: see the note in ./algebra/maputil.ml

January 3, 2002 - Jerome and Mary

  - Cleaned command line option for Lori's test suite.
      * -notype : turns types off
      * -noeval : turns eval off
      * -parse	: only do parsing
      * -algebra is removed
      * -print-core : prints XQuery Core. False by default.
  - Removed junk from ./regress directory
  - Added directories and README for test suite in ./regress:
      * ./regress/parsing     for parsing tests
      * ./regress/mapping     for XQuery to Core mapping tests
      * ./regress/xmlschema   for XML Schema to types mapping tests
      * ./regress/namespaces  for Namespaces tests
      * ./regress/typing      for typing tests on the Core
      * ./regress/eval	      for evaluation tests on the Core
      * ./regress/README      documents the test suite.
  - Updated the XQuery AST, lexer and parser to handle the new
    datatype production.

January 2, 2002 - Mary

  - Added command line options for xmlquery: 
    -verbose     Default behavior : print everything
    -quiet       Only print result XML document and type
    -o <prefix>  Prints XML document to prefix.xml and type to prefix.xsd
    
  - Changes to stdlib/pervasive/bltin, etc. to align with 20 Dec F&O document
  - Fixed tests & usecases to align with new grammar & 20 Dec F&O document
  - Updated STATUS document to reflect alignment with 20 Dec WDs.

January 2, 2002 - Jerome

  - Improved pretty printing of serialized documents.
  - Fixed bug in simple type lexer for decimal and doubles.
  - Fixed Fsa and Dfa module interfaces to compile under Ocaml-3.04.
  - Updated attribute constructor's syntax to reflect the new WD. New
    syntax is:
	   <a a="{1 + 2}"/> instead of <a a={1 + 2}/>
  - Removed obsolete ZERO token in the parser.
  - Removed minOccurs maxOccurs from the type syntax.
  - Fixed lexing of XML names (should be consistent with XML 1.0 with
    ISO-88591 encoding.
  - Cleaned whitespace after keywords in XQuery no longer needed.

December 20, 2001 - Mary

  - Updated examples in test directory.
    See tests/Status for examples that do not terminate
  - Fixed usecases/{xmp,tree}* to conform to new syntax

December 13, 2001 - Jerome

  - Added Use Case "Tree" to the demo.
  - Fixed the formatting of the Use Case "Tree" for the demo.
  - Fixed bug in typing rule for 'for'. Was not working for () and
    NONE types.
  - Added a configuration parameters to whether to print xsi:type
    information in the results or not. Updated the demo to take this
    parameter into account.
  - Added dependency in the Makefile for the toplevel galax
    interpretor.
  - Fixed simple type parser for doubles (was not accepting values
    without exponent).
  - Fixed some of the XML schema mapping for datatypes. NEEDS MORE
    CLEANING.
  - Cleaned up text, pi and comment parsing.
     o Fixed bug in child:: axis of document nodes.
     o Fixed printing of Comments and PI's expressions.
     o Removed CDATA from expressions. CDATA are not treated as text
       and folded into the current text node.
     o Fixed parsing of text in XQuery to deal with whitespace
       consistently.
  - Cleaned up parsing of XML document declaration.
  - Added an UnknownSimpleType in the type system for untyped text
    content from XML document.
  - Added text creation in the core.
     o Added CEText in cexpr in the XQuery Core AST.
     o Added typing rule and evaluation rule for CEText.
     o Changed mapping to the core to create CEText instead of
       CEScalar for text.
  - Fixed document root node bug in the data model loading and
    validation. Both now create root node appropriately.
     o This introduces a bug in the type system!! What is the type of
       rooted documents?
  - Turned all keywords in lowercase, following the new spec.
  - Turned "instanceof" into "instance of".
  - Fixed usecases to use only lowercase.
  - Added root type. Syntax is:
      o document { ... } added in type production
      o document { ... } added in datatype production
      o document() and document-validate() both return document { ... }
        types
  - Fixed usecases to use document functions and fixed several related
    bugs in path expressions.
  - Fixed whitespace handling bug. \013 was not recognized as
    whitespace, breaking the demo.
  - Changed typing rule for axis::text() to return UnknownSimpleType.
  - Fixed bug in mapping rule for [i]: index_of renamed to index-of.
  - Fixed AnyScalar in pervasive.xq.
  - Fixed printing bug for core expressions : let in instead of let
    return.
  - Fixed the max-int, min-int, max-float, min-float to return () when
    input is (). Fixed the signature in pervasive.xq to allow () as
    output.
  - Fixed sorting for xsd:integer and xsd:int datatypes.
  - Fixed integer arithmetics (+, -, *, div, mod) to allow null
    values.
  - Added a large part of use case "R".
  - Fixed printing for text("...") expressions in the core.
  - Fixed bug in printing of attribute names.

December 1, 2001 - Jerome

  - Added a Caml toplevel with built-in Galax libraries (galax).
  - Some fixes in the XQuery parser:
     o removed obsolete 'true' and 'false' keywords.
     o fixed numeric literals to support:
         --> XQuery numeric literals: double & decimal
         --> int (not part of the XQuery spec)
     o fixed string literals to avoid escaping a la Caml.
  - Some fixed in simple type parsers:
     o fixed numeric literals to support:
         --> xs:double  (IEEE floating point number, see XML schema
			 xsd:double, implemented as Caml float)
	 --> xs:decimal (Arbitrary precision decimal values,
                         implemented as Caml Num.num)
         --> xs:integer (Arbitrary integers, implemented as Caml
                         Big_int.big_int)
         --> xs:int (Implemented as Caml int)
     o fixed boolean parser to support 1 and 0 for true and false.
     o fixed double parser to support NaN, INF, and -INF
  - Fixed printing of simple types: decimal was printed as integer.
  - Fixed bug in subtyping of simple types.
  - Fixed several bugs in validation:
     o Validation is now working on empty content models.
     o Validation is now working on simple content models with
       attributes.
  - Fixed several bugs in recfactor. not terminating for recursive
    elements. wrong bounds.

Nov 30, 2001 - Mary
  - base/error.ml -- errors not printed to stderr
  - schema/schemamapping.ml - made some functions public

Nov 29, 2001 - Mary

  - added TREE use case
     o Validation error when trying to validate:
       define element image { 
         attribute source { xsd:string }
       }
     o Stack overflow error in //

November 28, 2001 - Mary and Jerome

  - Fixed several bug in validation. Validation now takes a root
    global *element* instead of a global *group*.
  - Fixed several bug in compilation process and Makefiles.
  - Separated API compilation from the mail Galax libraries
    compilation.
  - Moved usecases in a root directory.
  - Removed obsolete examples from the demo.
  - Fixed names of F&O functions and operators from '_' to '-'.
  - Fixed aggregate functions to be type specific (this is in
    contradiction with the F&O document). This allows more precise
    typing. Functions are:
       min-int	      min-float
       max-int	      max-float
       sum-int	      sum-float
       avg-int	      avg-float
  - Fixed the XMP use cases. Notably the data() function only works on
    a single node, which requires additional for loops in some cases.
  - Heavy fix of the Simple module: aligned all built-in primitive
    simple types with XML Schema spec. Propagated that change
    throughout the code.
  - Moved builtin type resolution from the parser to the Simple
    module.
  - Added serialization of xsi:type information.
  - Updated INSTALL file.
  - Fixed bug in validation of optional elements. (was only occuring,
    when an element had several optional attributes).

November 21, 2001 - Jerome

  - Fixed Makefile to create parent directories during installation as
    well.
  - Fixed bugs in the demo. The demo now uses well-formed XML
    documents directly using the new parser.

November 20, 2001 - Jerome

  - AST update:
	o separated XML documents and XQuery AST in new modules:
		    Xml_ast
		    Xquery_ast
	o moved AST utilities in corresponding modules:
		    Astutil -> Xquery_astutil
                    (new)      Xml_astutil
    The new XML ast is much smaller and should reduce memory
    consumption when doing parsing.
  - Created separate XML lexer and XML parser using Xml_ast.
  - Migrated the XML parsing and loading functions to use Xml_ast and
    the new parser. This impacts:
        Parseutil
	Schemamapping
	Load
	Bltin
	Toputil
  - Implemented partial Schema validation. This is in the Validation
    module. The implementation uses the XQuery type system and
    performs tree-walking on the AST for XML well-formed document,
    parses text to check for values according to the schema and
    creates an instance of the XML Query Data Model with values.
  - Added document-validate function that performs both document
    parsing and schema validation.
  - Did some minimal debugging and testing of validation. See:
    ./tests/xq21.xq.

November 20, 2001 - Mary

  - Updated INSTALL, README
  - Added STATUS
  - Removed TODO
  - Updated Makefile to install executables in bin/
  - config/Makefile
     o added GALAXHOME as install/release directory 
  - stdlib/pervasive.xqp
    stdlib/bltin.ml
     o fixed bugs in function signatures
     o alignment with F&O doc
  - Changed several Internal_Error exceptions to Prototype exceptions
  - base/config.mlp 
     o changed environment variable GLXLIB to GALAXLIB

November 13, 2001 - Byron

  - Added some functions from F and O document
  - the change is very localized. It's in the end of the bltin.ml and
    the pervasive.xqp
  - We may consider to put all those F and O in a separate file
  - The test script file is in tests/fando.xq
  - More complete test is needed.

November 12, 2001 - Mary

  - Added builtin functions true(), false()
  - Added integer arithmetic operators '*', 'div', 'mod'
     o see stdlib/pervasive.xpq for ugly lexer hack
  - XPath 2.0 Semantics: 
     o Added support for implicit conversion of nodes to their
       simple-content values in conditional, binary-op, and unary-op
       expressions and in the key expression of the sortby.
     o NOT DONE: implicit conversion of node arguments to functions
       expecting simple values.
     o NOT DONE: 3-valued logic for all equality/comparison operators
  - Fixed bugs: (see CVS comments)
     o in schemamapping (element values now EText)
     o in datamodel/instance_eq : element's children or simple valued
       content are compared for equality; previously, only children
       were compared.
     o typewitch variable now available in default clause
     o typeswitch may now refer to 'unit' group types, e.g.,
       xs:AnyElement
  - Added error messages during type checking for undefined variables,
    functions
  - In tests and tests/usecases, removed all unnecessary applications
    of data().
     o Changed confusion use case to use text() instead of data()
       because it is applied to an 'untyped' document.
  - Documentation:
     o split README split into README, INSTALL, and DEVELOPERS.
       INSTALL contains installation instructions.  DEVELOPERS
       contains internal documentation for implementors.

November 11, 2001 - Jerome

  - Worked on the C, C++ and Java API. Some bugs remain.

November 9, 2001 - Arnaud and Jerome

  - Added Java JNI interface to Galax! (not usable yet)
  - Started cleaning the C and C++ interfaces.

November 8, 2001 - Jerome

  - Renamed Xasyn into Xquery_ast.
  - Fixed bug in the XQuery lexer when parsing whitespace in a closing
    tag.

November 4, 2001 - Jerome

  - Removed statix code out of Galax. StatiX is now an independant
    piece of software.

October 30, 2001 - Jerome

  - Added a specific statix command-line tool for the statistics.
  - Fixed many bugs in the statistics generation.

October 26, 2001 - Jerome

  - Hooked the statistics generation to the validate command
    tool. Improved validate tool interface.
  - Moved normalize_model from Subsumption to Schemautil.
  - Added resolve_rgroup_stat, used for statistics generation.

October 25, 2001 - Jerome

  - Added -wf option in the validate command line tool to only check
    for well-formedness.
  - Added basic lexer module for XML Schema simple types. Most of the
    simple type values are string place-holders for now.
  - Added conversion operations from text to simple values in Simple.
  - Support for simple type values (Simple) can now shared between the
    AST and the data model. This simplifies a lot of the evaluation
    and loading operations. Modified Xasyn accordingly to use
    Simple. Propagated that change to the AST accross the system.

October 24, 2001 - Jerome

  - Moved simple values from Datamodel to Simple module. Move the
    simple types module into the ./base directory.
  - Renamed ./misc directory into ./base.
  - Added a ./validate command-line tool that only performs schema
    validation and reports errors.

October 21, 2001 - Jerome

  - Major update of the Data Model:
      o Implementation of text nodes
      o De-nodification of values. Values are not nodes anymore, do
        not have an identity of a parent.
      o Implementation of instances as (possibly heterogeneous)
        sequences of values or nodes.
      o Major cleaning of the data model interface itself. Reorganized
        function according to types.
      o Propagated data model update to the rest of the system. Subtle
        changes in the behavior of the run time can occur due to the
        difference in handling text and values.
  - Modified serialization to distinguish printing of values and text.

October 19, 2001 - Jerome

  - Removed obsolete InfoItem and Reference nodes from the data model.

October 2, 2001 - Jerome

  - Added an XML Schema use case in the demo. This contains XML Schema
    import, static checking of input/output schemas and some
    namespace related examples.
  - Fixed bug in uri <--> prefix table.
  - Fixed bug in default namespace and handling.
  - Fixed pretty printing for models and sort declarations.
  - Fixed pretty printing of core expressions.
  - Improved error messages for failed type assertions. ASSERT AS.
  - Fixed namespace support for the document() function. The document
    function takes the namespace's environment into account.
  - Allowed the default prefix in the XML Schema mapping.
  - Added mapschema command line tools. This takes an XML Schema file
    and prints the corresponding type system representation of that
    schema.

August 26, 2001 - Jerome

  - All the namespace examples in ./tests seem to work (This needs to
    be checked by a namespace guru!!!)
  - Fixed bug in printing for the default namespace.
  - Fixed bug in mapping of element constructors: it should use
    locally defined namespaces.
  - Fixed bug in ASSERT mapping for global element and attributes.
  - Added support for Floats.
  - Moved validation module from ./schema to ./datamodel.
  - Implemented the new syntax for global variables:
      DEFINE GLOBAL $a { expr }
    Updated the usecases accordingly.
  - Fixed bug in prevasive.xqp: signature of the document() function
    should be xs:AnyType.
  - Fixed error message in case of parsing or lexing error.
  - Fixed bug in Parser: XML fragment should return text instead of string.
  - Fixed bug in Lexstate: string buffer was not initialized properly.
  - Fixed bug in Parseutil: opened files where not closed properly.
  - Fixed bug in mapping of global variables: model wasn't mapped.
  - Fixed bug in subsumption: now raises an exception for
    non-deterministic content models instead of returning false.
  - Added TREAT in the core. Hooked its evaluation to validation (not
    implemented yet).
  - Renamed Simplify module into Simplification.
  - Recompiled the C-API and the demo.

August 25, 2001 - Jerome

  - Fixed many bugs due to the cleaning process, notably in
    subsumption.
  - Changed subsumption to avoid copying schema if not necessary.
  - Fixed schema compilation (previously called renaming) to complain
    about unknown global elements/attributes/groups.
  - Simple types now use the "xsd" prefix. Changed examples to reflect
    that.
  - Changed printing of models to use namespaces for infered types.

August 24, 2001 - Byron and Jerome

  - **Important** Schema replaces sort_decl pretty much everywhere to
    avoid the previous type information duplication (there was no
    clear rule at when to use one or the other). Related changes:
     o Schemautil replaces Typeutil but now works on schema
       structures. Typeutil has disappeared.
     o All the modules using type information are now using schemas,
       notably: subsumption, intersection, simplify, mapping, xatenv,
       typing and eval.
  - We are getting close! Everything but the C-API and the demo
    compile. Several namespace bugs have been resolved in Namespace
    and Sym modules. Printing is fixed somehow. Several typing bugs
    are still there.

August 23, 2001 - Jerome, Phil and Byron

  - We keep pushing hard on the cleaning. Recompiled: ./datamodel and
    ./stdlib modules.
  - Further updates on Sym, Namespace and Resolve modules.
  - Pushed cleaning up to typing. It is now time to clean the
    Schema/Sort Decl mess.

August 22, 2001 - Jerome and Byron

  - Major cleaning of the Symbol support, of the QName support, of the
    Namespace support, of the Schema support, of the XML Schema
    support and of the Mapping support. Here is a list of changes:

    In Misc
    -------

    o New ./misc/pool.ml module implement generic support for name
      pools
    o New ./misc/namespace.ml module contains support for qnames
      (resolved and unresolved) and namespaces environment.
    o ./misc/sym.ml:
        * Removed namespaces from there
	* Created separate symbol pools for the 6 following kinds of
          names:
            --> uelem_symbol  / unresolved element names
	    --> relem_symbol  / resolved element names
	    --> uattr_symbol  / unresolved attribute names
	    --> rattr_symbol  / resolved attribute names
	    --> ugroup_symbol / unresolved group names
	    --> rgroup_symbol / resolved group names
	* The interface has separate functions for the 6 above kinds
	  of names plus a few additional generic functions on certain
	  kinds of names.
    o New ./misc/resolve contains the basic resolution functions for
      namespaces.

    In Parsing
    ----------

    o Removed the algebra parser and the algebra parse utililities
    o Removed the warning module
    o ./parsing/xasyn.mli:
	* Removed support for nameclasses and named types (until we
	  know what to do with them).
	* Added schema names (used in the schema module).
	* Put qname back in Path expressions (instead of symbol)
	* Added types for XML AST and XML Schema Ast
    o ./parsing/parseutil: Added function for XML Schema.

    In FSA
    ------

    o Created a new fsa directory with all the support for FSA.

    In Schema
    ---------

    o moved ./parsing/xmlschema.ml to ./schema/schemamapping.ml
    o moved all fsa related code to ./fsa
    o Folded all namespace resolution and conflict detection in the
      ./schema module to type declarations (i.e., when compiling the
      schema).
    o Finished recompiling everything with the new symbol support.

  - merged with Mary's new Typeutil (renamed typeutils to typeutil)

August 17, 2001 - Mary

  - Added '//'
    - fixed bug in typeutils.ml

August 16, 2001 - Mary and Byron

  - Added support for importing Schemas into type system: 
      SCHEMA prefix = "uri" 
    Regression tests in regress/schema-test.xq
  - Added built-in functions : 
    length, contains, ends-with
  - Added support for XPath '[]' predicates, 
    including last(), position() and [IntegerLiteral]
    tests in tests/xq01b.xq
  
August 10, 2001 - Byron

  - Implemented the below in mapping.ml.
    Jerome wrote:
      Having static static type assertion only for global elements. In
      any case, a static type assertion checks the type of all its
      descendants so we do not have to put them everywhere.

August 8, 2001 - Everyone

  - Merged support for namespaces with for-factoring and nodeswitch
  - Added syntax of function signatures: 
	DEFINE FUNCTION SIGNATURE fname (Datatype, ... , Datatype)
        RETURNS QName | '(' Datatype ')'
  - Changed pervasive.xq to use XQuery syntax
  - Debugged namespaces

August 7, 2001 - Mary and Byron

  - Added symbol spaces for namespace name and group.
      symbol_val_c ::= Datasym .. | Elemsym .. | Attrsym .. 
                    | Groupsym qname | NSsym (prefix opt, uri, targetnamespace)
    Added accessors and constructors for the symbols
    (See Also: misc/sym.ml)
  - Element, Attribute, Group no longer carry qname or tname along.
    They carry symbol now. (See Also: parsing/xasyn.mli: expr_desc,
    cexpr_desc, model_desc, and nameclass)
  - All the occurence of Element, Attribute, Group (core) expressions
    need to be modified. The change is minor. But, it's everywhere.
  - Slight modification on print module, as expected.
  - The symbols are generated and added into the correct symbol space
    during parsing
  - mapping.ml does two things:
    1. Map the Full AST to Core AST
    2. Resolve the qnames all over the place (Element, Attribute,
    Group, Path, etc.) 
  
    Note: all are done in 1 walk of the AST
    Warning: mapping.ml becomes heavy-weight and less trivial.
             maputil.ml are not doing much work.
  - Modify Namespace node in datamodel (See also datamodel.ml: NSNode )
      o the change we made is small. Essentially modifying the node,
        its accessor and constructor And the change on fserialize
        support serialization of the namespace-awared datamodel
        forest.
      o load.ml is modifyed accordingly.
  - toplevel: use pervasiveX.xq (in Full XQuery synatx) instead of
    pervasive.xq (in Core syntax)

  - Added tests/result directory: It contains the redirected output of
    the tests, located in tests/
    Usage:
    >  xmlquery xqXX.xq > xqXX.res
    >  diff xqXX.res result/xqXX.res
    If your xmlquery executable is correct, diff will tell you that
    there is no difference between your output and the one in result
    dir.

August 6, 2001 - Mary

  - xquery_parser: limitedmodel. TNone and TEmpty are added add adhoc
    rules to parse function signatures

August 3, 2001 - Jerome

  - Added descendant and descendant-or-self accessors in the data
    model. Added corresponding calls in the evaluation module. Typing
    for descendant is not working yet. Sorting in doc-order and
    duplicate removal is not working.

August 2, 2001 - Jerome and Phil

  - Turned back to for-factoring.
      o Added axis in the core.
      o Added direct evaluation rules for axis.
      o Modified mapping modules not to generate typeswitch and
        preserve axis.
      o Checked examples with the new system.
      o Put back factoring in the for type inference rule.
      o Added direct typing rules for axis.
  - Added primitive support for warning when an expression is not
    empty but has type empty.

August 1, 2001 - Jerome

  - Fixed bug in typing for children.
  - Added Jonathan Robie's example on well-formed vs. invalid
    documents.

July 30, 2001 - Jerome

  - Fixed bug in type intersection due to bug in or-patterns in
    Caml(sic!).
  - Fixed formatting and presentation bugs reported by Phil in the
    demo.
  - Added support for Attribute and Element references, except in XML
    schema mapping.
      tests are in ./tests/subtypes2.xq

July 26, 2001 - Jerome

  - Added wadler as a CVS user.
  - Put all the demo file in the Galax ./demo directory.
  - Added current usecases in ./tests/usecases.
  - Added automatic generation of JavaScript and HTML files for the
    demo from the current use cases.
  - Added [-context file] option in xmlquery command-line so one
    context file can be shared among many queries.

July 24, 2001 - Jerome

  - Added support for tracing of accesses to the data model children
    operation.
     o Extended algebra parser to support a (&1) [ "john" ] where 1 is
       the ID of the a element node.
     o Added tracing global variable in Config. This variable is set
       to true if tracing is on.
     o Added -trace file option in the command line xmlquery. The file
       stores the node id of the elements accessed during the query in
       the order where the query accesses these elements.

July 23, 2001 - Mary

  - Added maping for error()/ERROR() functions

July 23, 2001 - Byron

  - parsing/xmlschema: map XML syntax to the sort_decl
               - partially implemented, had some simple tests
  - parsing/MAP2SCHEMA: the mapping rules from XML Schema to the galax
    sort declaration and models

July 21, 2001 - Jerome

  - Computed names can now be of type string or qname.
  - Fixed bug in the name function (did not work for attributes).
  - Added type AnyNode = AnyElement | AnyAttribute in pervasive.xq.

July 20, 2001 - Jerome

  - Fixed bug in the typing rule for the children function. children
    now takes any node as input and returns empty sequence for values.
  - Added support for computed elements and attributes in the parser.
  - Added mapping rule for computed elements and attributes.

July 17, 2001 - Jerome

  - XMP use case to use XML element constructors. Added Q7 and Q10.
  - Fixed XQuery parser to support expression lists within enclosed
    expressions. (i.e., <a>{ 1,2 }</a> instead of <a> { (1,2) } </a>.
  - Added support for XML element constructor and EText in the mapping.
  - Modified XQuery parser to generate EText instead of string values
    for XML text.
  - Added EText expressions for text nodes in the AST.
  - Finished rewritting all the demo examples. All the examples from
    the XML DevCon demo are now running.
  - Cosmetic changes in xmlquery and in the demo scripts.

July 16, 2001 - Byron

  - translated xq18.xq, xq19.xq, xq20.xq to Full XQuery syntax.

July 16, 2001 - Byron

  Jerome wrote: 
  - Fixed bug in DFA to NFA automata mapping.
     o Why is there such a conversion ??

  val includes : NFA -> NFA -> bool    (old)
  val includes : NFA -> DFA -> bool    (new)

     o When we construct a DFA from a nondeterministic regular
       expression, it will raise exception in runtime.

     o We can pass NFA of a bigger type to includes
       functions. (statically checked)

  val intersects: NFA -> NFA -> NFA        (needed in includes)

     o This code is tested and working before DFA is written. (the
       code is rather non-trivial.)
     o write the conversion of DFA to NFA to reuse intersects.
     o the conversion is not needed after intersects is modified to
       intersects: NFA -> DFA -> NFA

July 16, 2001 - Jerome

  - Implemented existential quantification in predicates.
  - Updated the demo with XQuery examples.
  - Fixed bug in Schema.weakunion_types and Simplify.
  - XML loading function to remove leading and trailing whitespace.
  - XML loading function to remove empty text nodes.
  - Fixed Mary's subtype operator:
     o resolved grammar ambiguity.
       new syntax is : ASSERT AS (m1 <: m2)
     o implemented mapping rule for <:
  - BUG: subtyping should not return false when the bigger type is non
    deterministic.

July 16, 2001 - Mary

  - commented debugging statements
  - added more tests to subtypes.xq 

July 15, 2001 - Mary

  - Added subtype operator model <: model for testing.
    See tests in tests/subtypes.xq
  - Added check_type to algebra/typing.ml to check type
    of non-empty expressions.

July 15, 2001 - Jerome

  - Recompiled the demo.
  - Cleaned the toplevel modules and the CGI modules.
  - Added support for global variables in the parser. Syntax is:
       let $v as m := (e)
  - Added mapping rule for global variables.
  - Cleaned mapping module.
  - Fixed e/@a bug located in mapping module.
     --> mapping for attributes was not creating labelattr symbol.

July 13, 2001 - Mary

  - Implemented mapping for following XQuery expressions: EApp, EAttr,
     EAssert, EBinaryOp, EElem, EFLWR, EIf, EList, EPath (partially
     implemented) EScalar, ESelf, ESome, EEvery, ESort, ETypeswitch,
     EUnaryOp, EVar
  - Implemented index_of using value equality (needs node equality but
    DM doesn't have identity fully implemented)
  - Updated tests/Status; most working upto xq18.xq.
  - LEXER BUG : in function definition, function name must be
    immediately followed by lparen.

July 13, 2001 - Jerome
 
  - more updates on the XQuery parser for types.
      o typeswitch supports full models
      o function declarations take a limited forme of
        model/element/attribute references
      o TREAT/CAST/ASSERT take only a limited form of
        model/element/attribute references
      o made more tests and examples in regress. everything seems much
        better now.
  - updated the XQuery parser to support precisely the current
    consensus grammar.
      o removed TYPE declarations
      o simplified wildcards
      o minOccurs maxOccurs are not optional

July 12, 2001 - Jerome

  - added many subsumption examples, most work. still a few bugs:
      o RNone is not supported properly in the automata
      o schema renaming must differ from bigger and smaller types.
  - Updated syntax for types
      o added ( model ), model?, model+
  - Fixed bug in DFA to NFA automata mapping.
      o Why is there such a conversion ??
  - Added new implementation of the subsumption module.
      o still supports subsumption mappings
      o works for smaller type non deterministic
      o hooked to Byron's automata for inclusion of regular
        expressions

July 12, 2001 - Mary

  - Translated all 'algebra' examples in tests/xa*.xq into XQuery
    syntax in files tests/xq*.xq.  Not all examples are complete.
  - Added static type assertion expression: ASSERT AS model (expr)
  - Extended mapping.ml to handle type declarations, function
    definitions, and simple path expressions, e.g., $v/a
  - Minor changes to algebra/xmlquery_parser.mly to parse TYPE
    declarations (BUG ALERT: See ??MF about parsing ATTRIBUTE/ELEMENT
    tokens )

July 12, 2001 - Byron

  - Added an automata module during the past few days.
    o fsa.ml(i): a fsa functor, plus some another structures for 
      creating an automata for type
    o glushkov.ml(i): the implementation of the algorithm in Regular
      Expressions into Finate Automata [Anne Bruggemann-Klein]

      (Definitions of follow, nullable, first,..., at the front
       Definition 2.3: Definition of Glushkov automaton 
       Definition 3.1, 3.2: Star normal form)

      Normalization of Type AST is added. The normalization of TBound
      (t, m,n) is follow:
      Modified from Jerome's code
                 (I)      t{0,*} == t*
                 (II)     t{1,*} == t+
                 (III)    t{m,*} == t, t{m-1, *}

                 (IV)     t{0,1} == ()|t
                 (IV')    t{0,n} == ()|t,t{0,n-1}
                 (V)      if m > 1 then t{m,n} == t,t{m-1,n-1}

    o intersectfsa.ml(i): given two TypeName_NFA.nfa, M1 and M2,
      compute the intersection of M1 and M2.
 
      The algorithm is mainly running special Dijkstra's on two
      automata in parallel.

      NB. Print function is provided to print out the automata. For
          debugging reason.

    o reachablity.ml(i): provide a function to decide the reachabilty
      problem: Given an automaton M, a (start) state s, and a set of
      (final) states F, is there a simple path from s to any memeber
      of F.

      The algorithm is adapted from Dijkstra's.

      Limitation: M can only be the TypeName_NFA_Pair.nfa.  It should
      be more generic. Need modification.

July 3rd, 2001 - Jerome

  - Added printing function for full XQuery expressions.
  - Recompiled the C and C++ API.
  - Fixed algebra parser. Sort rule only allows one sort
    criteria. Fixed precedence problem.
  - Fixed xquery parser. Fixed lexer that kept 'element ' in the
    ELEMENT token instead of only keeping the qname.

July 3rd, 2001 - Byron

  - sortby is working on simple types. 
    o the parser is updated (ie., expr1 sortby expr2 sortkind)
    o the typing rule is added
    o the evaluation rule is added
    o example query are
	query bib0/book/author sortby data() descending, data() ascending
	query bib0/book sortby year/data() descending sortby year/data() ascending
    o To Do
      . need to propagate the typing error message properly. It
      screams "Fatal error" when a Query Prototype exception is
      raised. Quite scary.
      . need to factor out the dummy sort variable, "_dot". It is now
      hardcoded in xaeval.ml, typing.ml and xaparser.ml. Which file
      should it go to?
      . need to pretty print the core sort expression.
      . map the Full XQuery Sort to the Core Sort. Need to deal with
      Path expression. Not sure how to do.

July 2nd, 2001 - Byron

  - the simple test queries for the rules are located at
    galax/tests/xquery/
  - Added mapping from Full XQuery to the Core.

    o mappping.ml(i) : contains code for transversing the tree
    o maputil.ml(i) : definition of intermediate structures for the
    mapping and applying the mapping rule

    Below are the rules that has been implemented and passed simple
    tests.

    o Binary operators
    o FLWR expressions
    o Conditional expressions
    o Typeswitch
    o Unary operators (*NOT TESTED*)
    o Variables
    o Scalar values
    o Function application
    o Sequences
    o standalone element constructor
    o standalone attribute constructor

  - tests (only the evaluation part) the algebra test cases this is a
    brief summary of the test:

                evaluation/translation	
	xa01.xq          correct
	xa02.xq          correct
	xa03.xq          correct
	xa4.xq		 syntax error "for b <- ..."	
	xa05.xq          correct
	xa06.xq          correct
	xa07.xq          correct
	xa08.xq          correct
	xa09.xq		 'unique' not supported
	xa10.xq 	 'index' not supported
	xa11.xq 	 'sort' not working
	xa12.xq		 'group' not supported
	xa13.xq		 correct
	xa14.xq		 correct
	xa15.xq	         syntax error in type declaration
	xa16.xq		 syntax error "<-", case
	xa17.xq	         correct
	xa18.xq		 syntax error "<-", case
	xa19.xq		 syntax error "<-", case
	xa20.xq		 syntax error case

  - removed all the debugging print statements.

June 29th, 2001 - Mary, Byron and Jerome

  - Compiled the whole thing except the C api with new AST
  - Found lots of bugs in typing and subtyping.
  - Added ELEMENT A { } ATTRIBUTE A { } in the xquery parser. Did not
    debug that part yet.

June 28th, 2001 - Byron

  - Updated the parser so that both element and attribute are
    correctly parsed and constructed.
  - Done more tests systematically. Reveals some bugs. Not fixed yet.
    The scripts are located at ~/galax/regress
  - Bug on subsumption and simplify.  The new test set revealed the
    problem with simplify.ml It has problem in simplifying Unions.
    Fact:
     o subsumption.ml's bug: None <: Empty (true) Empty <: None
       (true!!! it should be false)
     o None is a subtype of everything (from XQuery 1.0 Formal
       Semantic section 4.2)

    In simplification, it deunions everything, simplifies, REMOVE
    types that is a subtype of something else in the new union.
 
    If we simplify () | 0, it will remove () because of the
    subsumption bug. It will give us 0. But, the correct result is ().
    The code checks whehter () <: 0. The bug tells wrong answer. So
    that () is removed.

June 27th, 2001 - Byron

  - wrote a simple regression test. 4 remarks are made, mainly about
    the xquery parser

      o This doesn't parses: 
	parseStringFXQuery "default namespace = \"pennDBSpace\"";;
        This does: 
	parseStringFXQuery "default namespace mySpace = \"pennDBSpace\"";;

        We are expecting the opposite.

        ToDo: EDefaultNamespaceDecl is (string * string). It should be
        string. EDefaultNamespaceDecl ENamespaceDecl are not used
        anywhere. => I can't tell what the pair of string means and I
        didn't modify.

      o model -> ... | ELEMENT nameset LCURLY typedmodel RCURLY | ...
        This parses: "element * { String }" This doesn't: "element
        myElem * { String }"

        qname is missing. => The element model is not properly created

      o model -> ... | ATTRIBUTE nameset LCURLY typedmodel RCURLY | ...
        This parses: { attribute myAttr* { String }
        This doesn't: { attribute myAttr * { String }

        ATTRIBUTE -> "attribute" whitespaceplus qname
        Cause: qname followed by nameset
        ToDo: put whitespaceplus between these two 'terminals'
  
      o Subsumption fails in this trivial case:
       	myMNone is a model of None
     	myMEmpty is a model of Empty
	Both of the below return true.
		isSubtype sortDeclList myMNone myMEmpty;;
		isSubtype sortDeclList myMEmpty myMNone;;

June 26th, 2001 - Byron

  - Updated Schema and Subsumption to compile with the new AST (not
    test yet).

  - subsumption.ml(i): the internal structure, sub_theta is updated
    to:

	type sub_theta = (tname * tname) list

    this affects operations on subsumptions mappings and subsumption
    functions.

  - Many functions in schema.ml match on the model_desc
      o Exceptions are raised if it matches TElemRef, TAttrRef.  I am
        not sure what's the right thing to do with them.
      o TAmp is not supported yet.

June 25th, 2001 - Mary, Byron, and Jerome

  - Updated Simplify, Intersection to compile with the new AST.
  - Updated all interface under schema directory to compile with the
    new AST.
  - Updated the first half of Schema.
  - Updated Typeutils to compile with the new AST.

June 24th, 2001 - Jerome

  - Checked algebra parser and printing functions on ./test
    examples. Updated all examples with the new syntax.

June 23th, 2001 - Jerome

  - Updated Parseutil to compile with the new AST. Cleaned detection
    of multiply defined functions and variables.
  - Added new printing functions from the core to the XQuery syntax.
  - Added back printing functions for old algebra type syntax.
  - Added printing functions for the new type systax.
  - Updated the Print module to compile with the new AST. Removed
    obsolete printing functions.

June 22th, 2001 - Jerome

  - Moved printing for minOccurs/maxOccurs values from the Print to
    the Bound module.
  - Updated the XQuery parser to compile with the new AST.
  - Added additional AST creation functions in Astutil that take finfo
    and other parameteres into account.
  - Added precedence for SORT in Algebra parser.
  - Modified syntax for match in Algebra:

    match e                        match e as v
      case v1 : t1 do e1     -->     case t1 do e1
      ...                            ...
      else en                        else en

  - Updated the Algebra parser to compile with the new AST. The
    algebra syntax for types handled by the parser is still the same,
    but it now creates the new AST using groups instead of
    types. Notably:

    TYPE X = t    -->    GROUP X = t
    a [ X ]       -->    ELEMENT a { GROUP X }

  - Updated the AST to handle the new new syntax for types.

June 21th, 2001 - Byron and Jerome

  - Created Astutil containing AST creation functions.
  - Put locinfo() in the Finfo module.
  - Updated Xasyn to the new AST for XQuery and the core.
  - Fixed Makefile in ./capi to avoid trying the installation is the
    CAPI is not compiled.
  - Fixed printing function for Sort.

June 10th, 2001 - Jerome

  - Removed support for unordered collections (cf. new version of the
    data model).
  - Added support for process instructions, CDATA sections and
    comments in the parser.

June 6th, 2001 - Jerome

  - Made CVS repository available to the rest of the Galax team.

June 4th, 2001

  - Mary Fernandez and Byron Choi join the Galax team!
  - Moved bltin.ml module from algebra/ to in stdlib/
  - Preliminary support for the new syntax for types in XQuery (Phil
    Wadler and Jonathan Robie's proposal).

May 25th, 2001

  - Created toplevel/toputils.ml file and updated toplevel tools
    accordingly.
  - Removed the serialization module. Transfered the corresponding
    function in the data model module.
  - Made _Node and _ValueNode abstract in the datamodel.
  - Did some cleaning on the datamodel interface.
  - Made _oforest and _uforest abstract in the datamodel.
  - Created lexer and parser for XQuery. Added corresponding files
    parsing/xquery_lexer.mll parsing/xquery_parser.mly. Removed old
    attempts xqlexer.mll and xqparser.mly. Updated parseutil.ml and
    xasyn.ml accordingly. Added all_usecases.xq in the test directory.
  - Created stdlib directory. This directory will contain the XPath
    2.0 operators, pervasive.xq, etc.

