Reclassify raster pixels underlying certain polygons using Python

Question

I've got a shapefile with many polygons, each of which exactly overlays a contiguous region of raster cells; an example in yellow:

enter image description here

For some of these, I want to use the polygon as a sort of analysis mask, so that I can reclassify only those cells that are overlaid by the polygon according to one of the polygon's attributes. I have to do this thousands of times to the same raster, so I'm looking for a Python-automated way to do it, using (hopefully) any of GDAL/OGR, SAGA, QGIS, or numpy arrays.

Answer 1

Figured this out myself, posting for posterity.

The command line utility gdal_rasterize is able to "burn-in" values into a raster, given a polygon as the area over which to do so. It accepts SQL queries, which can be used to specify certain polygons in a whole layer. From the GDAL documentation (http://www.gdal.org/gdal_rasterize.html), one can call something like

gdal_rasterize -a ROOF_H -where 'class="A"' -l footprints footprints.shp city_dem.tif

That burns the attribute value in the field ROOF_H from footprints.shp into city_dem.tif, wherever there are polygons satisfying 'class="A"'.