RasterSource interpretation of file paths

[vpipkt]

Assume we have a proper COG at /tmp/modis.tif.

spark.read.rastersource('/tmp/modis.tif').count()

Throws

Error fetching data for [0,100000000e,646f6d2f706d742f,6669742e7369]
java.lang.UnsupportedOperationException: Reading '/tmp/modis.tif' not supported

Which is confusing and unhelpful.

The user meant

spark.read.rastersource('file:///tmp/modis.tif').count()

Can we

1. Give a clearer error message that the uri was not understood, or
2. Look to see if the path starts with '/' and try prepending 'file://' ?

[metasim]

We should do (2.).

[vpipkt]

I believe the current behavior is now that it won't throw an error but an empty dataframe is returned.

[metasim]

🤦‍♂

[vpipkt]

related... see docs/raster-write.pymd