Inverse problem: Read a table or SQL query into an Xarray Dataset

[alxmrs]

I had a shower thought today. If Zarr is an OLAP DB, then could we treat tables as rasters?

https://github.com/zarr-developers/zarr-specs/issues/290

One perspective is that this is the inverse of what this project is trying to do so far, which is to treat rasters as tables to run SQL queries against them.

One fuzzy idea for a use case like this would be to read a table of data associated with a latlng and translate it to a (sparse) raster.

Similar to #4, I wonder if column-oriented data like parquet, which most headless tabular data stores are represented today, would naturally fit well with Xarray-style chunks.

[alxmrs]

I think this is more or less how xvec works: https://xvec.readthedocs.io/en/stable/

I still could see a contribution towards this issue where we implemented something on top of DataFusion and Xarray.