jgrss/warp cleanup

[jgrss]

What is this PR changing?

• Simplifies the mosaic() method and ensures support for multi-band mosaics
• Fixes bug in to_raster(), where nodata values were not being written to file
• Adds bigtiff keyword argument to save() method
• Updates write tests

[jgrss]

@mmann1123 if you have time, could you test this branch out with some of your data?

[mmann1123]

Sure

[mmann1123]

resolves #309

[jgrss]

@mmann1123 I added a few more tests, which exposed a bug in to_raster(). Previously, the nodata value was not properly saved when using to_raster(). I commented on the fix in io.py.

[jgrss]

@mmann1123 I made bigtiff a keyword argument, with a default of None. You should be able to pass data.gw.save(..., bigtiff='IF_SAFER'). Let me know if that doesn't work.

[mmann1123]

@mmann1123 I made bigtiff a keyword argument, with a default of None. You should be able to pass data.gw.save(..., bigtiff='IF_SAFER'). Let me know if that doesn't work.

That works.

[jgrss]

I think the issues i was having had to do with hitting memory limits, mosaicing huge 18 band images must have been bonking it. I've tried to recreate the issues on reasonably large datasets and haven't had the same issue. I think this is working well for almost all purposes.

Were you referring to this?

There's an issue here trying it out on my files and I'm seeing empty rasters after mosaicing. Unfortunately I need to sign off and can't dig into why

[jgrss]

We can postpone this merge if you want to diagnose the large raster mosaicking. Is the file too large to upload and share?

[mmann1123]

I think we merge this, but we can open a large mosaic issue. I was able to resolve my issue using da.maximum() Maybe it is possible I was triggering compute or persist.