Introduction
************

All collections published by a lazr.restful web service work pretty
much the same way. This document illustrates the general features of
collections, using the cookbook service's collections of cookbooks and
authors as examples.

    >>> from lazr.restful.testing.webservice import WebServiceCaller
    >>> webservice = WebServiceCaller(domain='cookbooks.dev')

==========================
Collections and pagination
==========================

A collection responds to GET by serving one page of the objects in the
collection.

    >>> cookbooks_collection = webservice.get("/cookbooks").jsonBody()
    >>> cookbooks_collection['resource_type_link']
    u'http://...#cookbooks'
    >>> cookbooks_collection['total_size']
    7
    >>> cookbooks_collection['next_collection_link']
    u'http://.../cookbooks?ws.start=5&ws.size=5'
    >>> cookbooks_collection.get('prev_collection_link') is None
    True

    >>> from operator import itemgetter
    >>> cookbooks_entries = sorted(
    ...     cookbooks_collection['entries'], key=itemgetter('name'))
    >>> len(cookbooks_entries)
    5
    >>> cookbooks_entries[0]['name']
    u'Everyday Greens'
    >>> cookbooks_entries[0]['self_link']
    u'http://.../cookbooks/Everyday%20Greens'
    >>> cookbooks_entries[-1]['name']
    u'The Joy of Cooking'

There are no XHTML representations available for collections.

    >>> print webservice.get('/cookbooks', 'application/xhtml+xml')
    HTTP/1.1 200 Ok
    ...
    Content-Type: application/json
    ...

You can get other pages of the collection by following links:

    >>> result = webservice.get("/cookbooks?ws.start=5&ws.size=5")
    >>> second_batch = result.jsonBody()
    >>> 'next_collection_link' in second_batch
    False

    >>> cookbooks_entries = sorted(
    ...     second_batch['entries'], key=itemgetter('name'))
    >>> cookbooks_entries[0]['name']
    u'Construsions un repas'

You can also get a larger or smaller batch than the default:

    >>> bigger_batch = webservice.get("/cookbooks?ws.size=20").jsonBody()
    >>> len(bigger_batch['entries'])
    7
    >>> 'next_collection_link' in bigger_batch
    False

    >>> smaller_batch = webservice.get("/cookbooks?ws.size=2").jsonBody()
    >>> len(smaller_batch['entries'])
    2
    >>> smaller_batch['next_collection_link']
    u'http://.../cookbooks?ws.start=2&ws.size=2'


But requesting a batch size higher than the maximum configured value
results in a 400 error.

      >>> print webservice.get("/cookbooks?ws.start=0&ws.size=1000")
      HTTP/1.1 400 Bad Request
      ...
      Content-Type: text/plain...
      <BLANKLINE>
      Maximum for "ws.size" parameter is ...

A collection may be empty.

    >>> from urllib import quote
    >>> url = quote("/cookbooks/Cooking Without Recipes/recipes")
    >>> result = webservice.get(url)
    >>> list(result.jsonBody()['entries'])
    []

==========
Visibility
==========

There are two recipes in "James Beard's American Cookery", but one of
them has been marked private. The private one is hidden from view in
collections.

    >>> from urllib import quote
    >>> url = quote("/cookbooks/James Beard's American Cookery/recipes")
    >>> output = webservice.get(url).jsonBody()
    >>> output['total_size']
    2
    >>> len(output['entries'])
    1

Why does total_size differ from the number of entries? The actual bugs
are filtered against the security policy at a fairly high level, but
the number of visible bugs comes from lower-level code that just looks
at the underlying list.

This is not an ideal solution--the numbers are off, and a batch may
contain fewer than 'ws.size' entries--but it keeps unauthorized
clients from seeing private data.

==============
Element lookup
==============

The elements of a collection can be looked up by unique identifier:

    >>> from lazr.restful.testing.webservice import pprint_entry
    >>> url = quote("/cookbooks/The Joy of Cooking")
    >>> cookbook = webservice.get(url).jsonBody()
    >>> pprint_entry(cookbook)
    confirmed: u'tag:launchpad.net:2008:redacted'
    copyright_date: u'1995-01-01'
    cover_link: u'http://.../cookbooks/The%20Joy%20of%20Cooking/cover'
    cuisine: u'General'
    description: u''
    last_printing: None
    name: u'The Joy of Cooking'
    price: 20
    recipes_collection_link: u'http://.../cookbooks/The%20Joy%20of%20Cooking/recipes'
    resource_type_link: u'http://...#cookbook'
    revision_number: 0
    self_link: u'http://.../cookbooks/The%20Joy%20of%20Cooking'

A collection may be scoped to an element:

    >>> url = quote("/dishes/Roast chicken/recipes")
    >>> result = webservice.get(url).jsonBody()
    >>> print result['resource_type_link']
    http://...#recipe-page-resource
    >>> cookbooks_with_recipe = sorted(
    ...     [r['cookbook_link'] for r in result['entries']])
    >>> len(cookbooks_with_recipe)
    3
    >>> print cookbooks_with_recipe[0]
    http://.../cookbooks/James%20Beard%27s%20American%20Cookery
    >>> print cookbooks_with_recipe[-1]
    http://.../cookbooks/The%20Joy%20of%20Cooking

================
Named operations
================

A collection may expose custom named operations in response to GET
requests. A named operation may do anything consistent with the nature
of a GET request, but it's usually used to serve search results. The
custom operation to be invoked is named in the query string's
'ws.op' argument. Here's a custom operation on the collection of
cookbooks, called 'find_recipes'.

    >>> import simplejson
    >>> def search_recipes(text, vegetarian=False, start=0, size=2):
    ...     args = ("&search=%s&vegetarian=%s&ws.start=%s&ws.size=%s" %
    ...             (quote(text), simplejson.dumps(vegetarian), start, size))
    ...     return webservice.get(
    ...         "/cookbooks?ws.op=find_recipes&%s" % args).jsonBody()

    >>> s_recipes = search_recipes("chicken")
    >>> s_recipes['total_size']
    3
    >>> sorted(r['instructions'] for r in s_recipes['entries'])
    [u'Draw, singe, stuff, and truss...', u'You can always judge...']

    >>> veg_recipes = search_recipes("chicken", True)
    >>> veg_recipes['total_size']
    0

A custom operation that returns a list of objects is paginated, just
like a collection.

    >>> s_recipes['next_collection_link']
    u'http://.../cookbooks?search=chicken&vegetarian=false&ws.op=find_recipes&ws.start=2&ws.size=2'

    >>> s_recipes_batch_2 = search_recipes("chicken", start=2)
    >>> sorted(r['instructions'] for r in s_recipes_batch_2['entries'])
    [u'A perfectly roasted chicken is...']

Just as a collection may be empty, a custom operation may return an
empty list of results:

    >>> empty_collection = search_recipes("nosuchrecipe")
    >>> empty_collection['total_size']
    0
    >>> [r['instructions'] for r in empty_collection['entries']]
    []

Custom operations may have error handling. In this case, the error
handling is in the validate() method of the 'search' field.

    >>> print webservice.get("/cookbooks?ws.op=find_recipes")
    HTTP/1.1 400 Bad Request
    ...
    search: Required input is missing.

If a named operation takes an argument that's a value for a vocabulary
(such as Cuisine in the example web service), the client can specify
the name of the value, just as they would when changing the value with
a PUT or PATCH request.

    >>> general_cookbooks = webservice.get(
    ...     "/cookbooks?ws.op=find_for_cuisine&cuisine=General")
    >>> general_cookbooks.jsonBody()['total_size']
    3

POST operations
===============

A collection may also expose named operations in response to POST
requests. These operations are usually factories. Here's a helper
method that creates a new cookbook by invoking a factory operation on
the collection of cookbooks.

    >>> def create_cookbook(name, cuisine, copyright_date, price=12.34):
    ...     date = copyright_date.isoformat()
    ...     return webservice.named_post(
    ...         "/cookbooks", "create", {},
    ...         name=name, cuisine=cuisine,
    ...         copyright_date=date, last_printing=date, price=price)

    >>> print webservice.get(quote('/cookbooks/The Cake Bible'))
    HTTP/1.1 404 Not Found
    ...

    >>> from datetime import date
    >>> print create_cookbook("The Cake Bible", "Dessert", date(1988, 1, 1))
    HTTP/1.1 201 Created
    ...
    Location: http://.../cookbooks/The%20Cake%20Bible
    ...

    >>> print webservice.get("/cookbooks/The%20Cake%20Bible")
    HTTP/1.1 200 Ok
    ...

POST operations can have custom validation. For instance, you can't
create a cookbook with a name that's already in use. This exception is
raised by the create() method itself.

    >>> print create_cookbook("The Cake Bible", "Dessert", date(1988, 1, 1))
    HTTP/1.1 409 Conflict
    ...
    A cookbook called "The Cake Bible" already exists.

A POST request has no meaning unless it specifies a custom operation.

    >>> print webservice.post("/cookbooks", 'text/plain', '')
    HTTP/1.1 400 Bad Request
    ...
    No operation name given.

You can't invoke a nonexistent operation:

    >>> print webservice.named_post("/cookbooks", "nosuchop", {})
    HTTP/1.1 400 Bad Request
    ...
    No such operation: nosuchop
