Carson Farmer

Last week I attended the 2009 QGIS Developers Meeting in Vienna, Austria. We all had a really good time, met many new people, and actually got a lot done in the process. There have been updates about the meeting (hackfest) on the QGIS blog, and Tim Sutton has written a few words about our progress as well. I’m not going to repeat what others have said, but I would like to give a quick update on the work that I was doing at the meeting, and show off the new geoprocessing features now available to all QGIS developers (Python and C++).

My main goal for the meeting was to start/continue work on the new ‘Analysis Library’ for QGIS. Basically, this was intended to be a port (to C++ from Python) of the fTools suite of tools already available in QGIS. These tools currently provide geoprocessing, geometry, and various other analysis functionality natively within QGIS. However, the Python implementation is simply a Python plugin, and so does not provide these functions to other developers hoping use/add geoprocessing capabilities to their own plugins or tools. In addition, in some cases the Python fTools functions can be quite slow, and would benefit greatly from the speed-ups afforded by a compiled language like C++.

Porting these functions to C++ is still underway, and so far we have implementations for buffering, dissolving, centroids, convex hulls, layer extents in the QgsGeometryAnalyzer class, and intersections in the QgsOverlayAnalyzer class. We have set it up so that it should be relatively easy to add functions and classes to the library in the future, though I expect there will be many changes to the library along the way. We have also created Python bindings for the library, so these functions will also be available to Python developers. By doing so, it is now extremely easy to perform geoprocessing functions directly from the QGIS Python console, or from QGIS Python plugins, with only a few lines of code.

The old way
In the past, Python developers had to operate directly on the geometries of the layers in order to do any sort of geoprocessing. For example, to perform a (very) simple buffer, one had to use the following code in the QGIS Python console:

from qgis.core import iface
mc = iface.mapCanvas() # get a reference to the map canvas
layer = mc.layer(0) # get a reference to the first layer in the layer list
provider = layer.dataProvider() # data provider for the layer
provider.select(layer.pendingAllAttributesList(), QgsRectangle(), True, True) # select features to buffer
in_feat = QgsFeature() # empty input feature
out_feat = QgsFeature() # empty output feature
writer = QgsVectorFileWriter( "output_path.shp", provider.encoding(), provider.fields(),
QGis.WKBPolygon, provider.crs() ) # use this to write results to disk (as shapefile)
while(provider.nextFeature(feat)): # for each feature that we selected...
    geometry = feat.geometry() # grab it's geometry
    buffer = geometry.buffer(100,10) # buffer the geometry
    out_feat.setAttributeMap(in_feat.attributeMap()) # set the attributes for the output feature
    out_feat.setGeometry(buffer) # set the bufer as the output geometry
    writer.addFeature(out_feat) # write the feature to file
del writer # delete/close the writer to save to disk

The new way
With the new QGIS Analysis Library, things are much simpler, and the same (or more complex) buffering example can be completed with only 5 lines of code in the QGIS Python Console:

from qgis.core import iface # import iface (interface)
from qgis.analysis import QgsGeometryAnalyzer # import (par of) the analysis library
mc = iface.mapCanvas() # get a reference to the map canvas
layer = mc.layer(0) # get a reference to the first layer in the layer list
QgsGeometryAnalyzer().buffer(layer, "output_path.shp", 100, False, False, -1) # perform the buffer

Note: only the first two parameters in the buffer function are really required, and the additional parameters control the buffer distance, whether to use a subset (selected features) of the layer, whether to dissolve the output buffer regions, and whether to use a field (field ID) from the input layer’s attribute table as the buffer distance.

Pretty nice eh? For now, the QGIS Analysis Library is only available in Trunk, however, before long these tools should be available as part of the official releases, so I hope we will start to see more Plugins taking advantage of these new capabilities in the future…

C