If the CirclesDatasource class was not available from the python_plugin_tests module then the 'load_good_maps' test
would fail. The class would not be available if, for example, the python plugin had not been built or if shapely was not
available on the test system. (Pull request #1414 removes the dependency on Shapely.)
The error reporting should be tidied up for this case but for the moment, move the Python plugin's test maps into their
own directory since they're not guaranteed to be 'good maps' in all cases at the moment.
Hopefully addresses issue #1407
This plugin allows you to write data sources in the Python programming language.
This is useful if you want to rapidly prototype a plugin, perform some custom
manipulation on data or if you want to bind mapnik to a datasource which is most
conveniently accessed through Python.
The plugin may be used from the existing mapnik Python bindings or it can embed
the Python interpreter directly allowing it to be used from C++, XML or even
JavaScript.
Mapnik already has excellent Python bindings but they only directly support
calling *into* mapnik *from* Python. This forces mapnik and its input plugins to
be the lowest layer of the stack. The role of this plugin is to allow mapnik to
call *into* Python itself. This allows mapnik to sit as rendering middleware
between a custom Python frontend and a custom Python datasource. This increases
the utility of mapnik as a component in a larger system.
There already exists MemoryDatasource which can be used to dynamically create
geometry in Python. It suffers from the problem that it does not allow
generating only the geometry which is seen by a particular query. Similarly the
entire geometry must exist in memory before rendering can progress. By using a
custom iterator object or by using generator expressions this plugin allows
geometry to be created on demand and to be destroyed after use. This can have a
great impact on memory efficiency. Since geometry is generated on-demand as
rendering progresses there can be arbitrarily complex 'cleverness' optimising
the geometry generated for a particular query. Obvious examples of this would
be generating only geometry within the query bounding box and generating
geometry with an appropriate level of detail for the output resolution.
