Latest from mapnik development team
Text placement re-factoring
One of the goals for “text-placement” was to re-factor rather messy text placement code into more manageable chunks, and hopefully to make it easier to follow and implement new features. So, now we have a better framework for text labeling - let’s use it. A classic example is labeling multipolygons that make up one geographical entity; country, county or state etc. Here are some maps to demonstrate how this would work for the US states :
1) Labeling regression (actually it never worked properly but changes to how multi-geometry is processed in Mapnik exposed this):
2) Try and label every polygon:
3) Order polygons by “size” and place label in first largest polygon:
 is now current default behaviour in master, giving a much better cartographic result. We need to expose other possible options in XML/Python/Node etc.
Feature concept changes
New ‘context’ based feature implementation landed in Mapnik. The interface was preserved as much as possible to make it easier to update existing code: the following Python code shows the differences with the new approach:
>>> import mapnik >>> ctx = mapnik.Context() # create Feature context >>> ctx.push("NAME") # push some names, defining a schema >>> f = mapnik.Feature(ctx, 1) # create feature by passing context and ID >>> f["NAME"] = "test"
The main gain here is a memory footprint and faster feature construction - no overhead of maintaining std::map per feature. Looking to the future, it will be interesting to see how we can leverage this for more robust/compact interchange formats.Posted by Artem Pavlenko on 10 February 2012.