documentation + tests + debugging for formparsers

2 files changed, 74 insertions, 2 deletions

diff --git a/djangorestframework/tests/__init__.py b/djangorestframework/tests/__init__.py
index 55e386b5..5d5b652a 100644
--- a/djangorestframework/tests/__init__.py
+++ b/djangorestframework/tests/__init__.py
@@ -10,5 +10,5 @@ __test__ = dict()
 for module in modules:
     exec("from djangorestframework.tests.%s import __doc__ as module_doc" % module)
     exec("from djangorestframework.tests.%s import *" % module)
-    __test__['%s' % module] = module_doc or ""
-    
+    __test__[module] = module_doc or ""
+
diff --git a/djangorestframework/tests/parsers.py b/djangorestframework/tests/parsers.py
new file mode 100644
index 00000000..a2831b63
--- /dev/null
+++ b/djangorestframework/tests/parsers.py
@@ -0,0 +1,72 @@
+"""
+..
+    >>> from djangorestframework.parsers import FormParser
+    >>> from djangorestframework.resource import Resource
+    >>> from djangorestframework.compat import RequestFactory
+    >>> from urllib import urlencode
+    >>> req = RequestFactory().get('/')
+    >>> some_resource = Resource()
+    >>> trash = some_resource.dispatch(req)# Some variables are set only when calling dispatch
+
+Data flatening
+----------------
+
+Here is some example data, which would eventually be sent along with a post request :
+
+    >>> inpt = urlencode([
+    ...     ('key1', 'bla1'),
+    ...     ('key2', 'blo1'), ('key2', 'blo2'),
+    ... ])
+
+Default behaviour for :class:`parsers.FormParser`, is to return a single value for each parameter :
+
+    >>> FormParser(some_resource).parse(inpt) == {'key1': 'bla1', 'key2': 'blo1'}
+    True
+
+However, you can customize this behaviour by subclassing :class:`parsers.FormParser`, and overriding :meth:`parsers.FormParser.is_a_list` :
+
+    >>> class MyFormParser(FormParser):
+    ... 
+    ...     def is_a_list(self, key, val_list):
+    ...         return len(val_list) > 1
+
+This new parser only flattens the lists of parameters that contain a single value.
+
+    >>> MyFormParser(some_resource).parse(inpt) == {'key1': 'bla1', 'key2': ['blo1', 'blo2']}
+    True
+
+Submitting an empty list
+--------------------------
+
+When submitting an empty select multiple, like this one ::
+
+    <select multiple="multiple" name="key2"></select>
+
+The browsers usually strip the parameter completely. A hack to avoid this, and therefore being able to submit an empty select multiple, is to submit a value that tells the server that the list is empty ::
+
+    <select multiple="multiple" name="key2"><option value="_empty"></select>
+
+:class:`parsers.FormParser` provides the server-side implementation for this hack. Considering the following posted data :
+
+    >>> inpt = urlencode([
+    ...     ('key1', 'blo1'), ('key1', '_empty'),
+    ...     ('key2', '_empty'),
+    ... ])
+
+:class:`parsers.FormParser` strips the values ``_empty`` from all the lists.
+
+    >>> MyFormParser(some_resource).parse(inpt) == {'key1': 'blo1'}
+    True
+
+Oh ... but wait a second, the parameter ``key2`` isn't even supposed to be a list, so the parser just stripped it.
+
+    >>> class MyFormParser(FormParser):
+    ... 
+    ...     def is_a_list(self, key, val_list):
+    ...         return key == 'key2'
+    ... 
+    >>> MyFormParser(some_resource).parse(inpt) == {'key1': 'blo1', 'key2': []}
+    True
+
+Better like that. Note also that you can configure something else than ``_empty`` for the empty value by setting :class:`parsers.FormParser.EMPTY_VALUE`.
+"""