polygonize: add option to avoid rounding of values in float rasters

[lutra]

That's it, the result would be a grid where every polygon had an attribute with the exact value of the underlying float raster.

[Jukka Rahkonen]

It seems that gdal_polygonize.py ​https://trac.osgeo.org/gdal/browser/trunk/gdal/swig/python/scripts/gdal_polygonize.py is calling GDAL algorithm gdal.polygonize as

gdal.Polygonize( srcband, maskband, dst_layer, dst_field, options, callback = prog_func )

From ​https://trac.osgeo.org/gdal/browser/trunk/gdal/alg/polygonize.cpp I can read that there exists an variant "GDALFPolygonize()" which seems to do what is desired in this ticket:

• The source pixel band values are read into a 32bit float buffer. If you want
• to use a (probably faster) version using signed 32bit integer buffer, see
• GDALPolygonize().

If I am right we could close this ticket as fixed because "GDALFPolygonize()" exists. However, most users run polygonize throught gdal_polygonize.py and it does not mention anything about this option. Also the comments in polygonize.cpp give more information about the usage than ​http://www.gdal.org/gdal_polygonize.html even the latter does have a link to the doxygen document of the algorithm.

[Flewellyn]

Modified gdal_polygonize.py

[Flewellyn]

I've attached a modified gdal_polygonize.py script, which adds the "-p"/"--float" option to run the process through GDALFPolygonize() instead of the integer-using GDALPolygonize. I can confirm that it works as expected.

[Even Rouault]

This ticket has been automatically closed because Trac is no longer used for GDAL bug tracking, since the project has migrated to GitHub. If you believe this ticket is still valid, you may file it to ​https://github.com/OSGeo/gdal/issues if it is not already reported there.