ARG (Azavea Raster Grid) Driver

[dzwarg]

I am working on a driver implementation for a raw format that we use in ​GeoTrellis, which we call ARG (​spec).

This is essentially a raw format, with a companion .JSON file. I felt that it was important to re-use the JSON libraries, so I added the ogr/ogrsf_frmts/geojson dir as an include in GNUmakefile and makefile.vc.

The patch for the driver is attached to this ticket.

I'm not familiar with the procedure for adding a new driver to the GDAL trunk, so I appreciate any feedback or critique on the implementation.

[dzwarg]

Initial implementation of Azavea Raster Grid driver.

[dzwarg]

Updates to ARG driver based on feedback from FrankW, hobu, and EvenR

[dzwarg]

Updated implementation:

• Use CPLString for filename paths to protect from buffer overflows, etc
• Emit error messages in Open when a recognized ARG file cannot be opened.
• Clean up JSON file during CreateCopy if .arg datafile cannot be created.
• Accommodate int8 formats as uint8, and emit debug message.
• Check return status of importFromWKT()
• Checks for OGR enabled before building argdataset in raw format

[dzwarg]

Updates to ARG driver based on feedback from EvenR -- build arg if ogr enabled, not bundled with RAW drivers.

[Even Rouault]

Patch committed in r24192.

Build fixes in r24193. Would be good if you could test your stuff before commiting ;-)

David, it would be good if you could write some help page, either a specific frmt_arg.html, or a new paragraph in frmts/frmt_various.html, and update frmts/formats_list.html to point to that documentation page.

[Even Rouault]

r24193 /trunk/gdal/frmts/arg/ (GNUmakefile argdataset.cpp makefile.vc): ARG: Linux and Windows build fixes (#4591)

[Even Rouault]

r24194 /trunk/gdal/frmts/arg/argdataset.cpp: ARG: fix crash on autotest/gcore/misc.py when testing CreateCopy() of a int16 source (#4591)

[dzwarg]

I built and test on Ubuntu: Linux raster 3.0.0-12-server #20-Ubuntu SMP Fri Oct 7 16:36:30 UTC 2011 x86_64 x86_64 x86_64 GNU/Linux

Before committing, I performed:

• make clean
• make
• python arg.py

I see now that I should have run ./run_all.py. Thanks for catching this.

Replying to rouault:

r24193 /trunk/gdal/frmts/arg/ (GNUmakefile argdataset.cpp makefile.vc): ARG: Linux and Windows build fixes (#4591)

[Even Rouault]

Well, I don't understand how it could compile without the raw subdirectory to be added in the makefile. Or perhaps you have an installed version of rawdataset.h in a system include directory ? In which case that's dangerous, because you could desynchronization with changes in your local tree. I'd recommand removing any libgdal-dev package from your system to avoid that.

[dzwarg]

I don't have libgdal-dev installed, but there are a bunch of gdal headers in /usr/local/include.

Would they get placed there when I do a 'sudo make install' ?

Replying to rouault:

Well, I don't understand how it could compile without the raw subdirectory to be added in the makefile. Or perhaps you have an installed version of rawdataset.h in a system include directory ? In which case that's dangerous, because you could desynchronization with changes in your local tree. I'd recommand removing any libgdal-dev package from your system to avoid that.

[Even Rouault]

Would they get placed there when I do a 'sudo make install' ?

Yes