Can GeoUtils be used for remote files?

[maxrjones]

Thanks for your work on GeoUtils! I don't have much experiences with the API yet, but the documentation is super impressive and well organized!

Are there any examples of using geoutils to load files in cloud object storage? I have a notebook using rasterio with the netcdf driver and vsis3 to open a NASA dataset in https://github.com/developmentseed/warp-resample-profiling/blob/main/examples/load-rasterio-netcdf-vsis3.ipynb. I tried to modify it to create a GeoUtils Raster object in https://github.com/developmentseed/warp-resample-profiling/blob/main/examples/future-load-geoutils-netcdf-vsis3.ipynb, but received IndexError: band index 1 out of range (not in ()). I'm not sure if it's due to using vsis3, the args in the vrt:// syntax, or some other problem. I'd love to include geoutils in a last minute talk I'm giving tomorrow on different resampling methods in Python, so if you have any tips or tricks I would greatly appreciate it! I'll likely still continue developing the materials after the talk as well if this timing doesn't work out.

[rhugonnet]

Thanks for opening this discussion @maxrjones, and sorry for answering only today, yesterday was quite a busy day.
I hope your talk went well in the meantime!

For reading/writing, we rely on Rasterio/GeoPandas, so it should normally work similarly.

It looks like the issue here comes from the delayed loading step of Raster.load(), in the case of working on S3.
Have you tried if forcing immediate loading works?

src = f"NETCDF:/vsis3/{bucket}/{input_uri}:analysed_sst"
da = rio.open(src)
da = Raster(da, load_data=True)
return da

or shorter but equivalent:

src = f"NETCDF:/vsis3/{bucket}/{input_uri}:analysed_sst"
da = Raster(src, load_data=True)
return da

In any case, you make a good point. I'll add an issue to find the necessary fixes for remote files, and setup additional tests + a couple documentation examples. Tell me if you find anything else while further developing your materials!
It should also become easier to deal with this in the future, through the Xarray accessor.

Maybe @erikmannerfelt has more insights for working with remote files?

[erikmannerfelt]

Hello! I expected it to be as simple as just passing any URL/URI that is accepted by rasterio or a rio dataset and it should work.

@maxrjones, what version of geoutils are you running? And can you test the two suggestions by @rhugonnet?

If that doesn't work I'll try it myself. Either way, I'd classify this as a bug because I see no reason why this shouldn't work!