Handle decimal serialization internally

With this change, user applications does not need to depend on simplejson.

The precision of the decimal is lost during serialization. There are
several reasons for this choice:
- first, an exact level of precision is unnecessary;
- second, modern versions of python json are more protective, and the
  2 hacks found in
  http://stackoverflow.com/questions/1960516/python-json-serialize-a-decimal-object
  does not work anymore.

Additionnal info can be found at:
- http://bugs.python.org/issue16535;
- http://bugs.python.org/issue18264.

dev_requirements.txt

@@ -4,5 +4,4 @@ coverage
 WebTest
 psycopg2
 pyramid_handlers
-simplejson
 flake8

docs/geojson_renderer.rst

@@ -55,10 +55,6 @@ implement the Python Geo Interface.
 
 Notes:
 
-* The GeoJSON renderer requires simplejson 2.1 or higher. Indeed, to be able to
-  deal with ``decimal.Decimal`` values, which are common when using SQLAlchemy,
-  we set ``use_decimal`` to ``True`` when calling the ``dumps`` function, and
-  only simplejson 2.1 and higher support that argument.
 * The GeoJSON renderer supports `JSONP <http://en.wikipedia.org/wiki/JSONP>`_.
   The renderer indeed checks if there's a ``callback`` parameter in the query
   string, and if there's one it wraps the response in a JavaScript call and

docs/index.rst

@@ -17,6 +17,9 @@ Protocol defines a HTTP interface for creating, reading, updating, and deleting
 Papyrus includes lower-level objects, like the GeoJSON renderer, that may come
 in handy for your apps even if you don't need or want MapFish views.
 
+Papyrus GeoJSON encoder converts decimal types to float numbers, which
+may incur an infinitesimal loss of precision.
+
 Installing
 ----------
 
@@ -26,11 +29,6 @@ Papyrus can be installed with ``easy_install`` or ``pip``::
 
 (Installing Papyrus in an isolated ``virtualenv`` is recommended.)
 
-You may want to install ``simplejson`` as well. This dependency is not required
-but it provides more functionnalities than the ``json`` module from the
-standard library like the possibility to parse float numbers as Decimal.
-``papyrus`` can use such options if ``simplejson`` is present.
-
 Most of the time you'll want to make Papyrus a dependency of your Pyramid
 application. For that add ``papyrus`` to the ``install_requires`` list defined
 in the Pyramid application ``setup.py``. Example::
@@ -66,7 +64,7 @@ Papyrus includes unit tests. Most of the time patches should include new tests.
 
 To run the Papyrus tests, in addition to Papyrus and its dependencies the
 following packages need to be installed: ``nose``, ``mock``, ``psycopg2``,
-``simplejson``, ``pyramid_handlers``, ``coverage``, and ``WebTest``.
+``pyramid_handlers``, ``coverage``, and ``WebTest``.
 
 For these packages to install correctly, you have to have header files for
 ``PostgreSQL``, ``Python``, and ``GEOS``. On Debian-based systems install the

papyrus/geojsonencoder.py

@@ -1,4 +1,5 @@
 import datetime
+import decimal
 import functools
 
 from sqlalchemy.ext.associationproxy import _AssociationList
@@ -9,7 +10,7 @@
 
 class GeoJSONEncoder(PyGFPEncoder):
     # SQLAlchemy's Reflecting Tables mechanism uses decimal.Decimal
-    # for numeric columns and datetime.date for dates. simplejson
+    # for numeric columns and datetime.date for dates. Python json
     # doesn't deal with these types. This class provides a simple
     # encoder to deal with objects of these types.
 
@@ -18,11 +19,14 @@ def default(self, obj):
             return obj.isoformat()
         if isinstance(obj, _AssociationList):
             return list(obj)
+        if isinstance(obj, decimal.Decimal):
+            # The decimal is converted to a lossy float
+            return float(obj)
         return PyGFPEncoder.default(self, obj)
 
 
-dumps = functools.partial(_dumps, cls=GeoJSONEncoder, use_decimal=True)
+dumps = functools.partial(_dumps, cls=GeoJSONEncoder)
 """
 A partial function for ``geojson.dumps`` that sets ``cls`` to
-:class:`GeoJSONEncoder` and ``use_decimal`` to ``True``.
+:class:`GeoJSONEncoder`.
 """