Comparisons to CARTO's raster processing toolkit

[alxmrs]

https://docs.carto.com/data-and-analysis/analytics-toolbox-for-bigquery/sql-reference/raster

https://carto.com/blog/cloud-native-raster-analysis-bigquery

It seems like this project and CARTO's advanced toolbox aim to solve similar problems. Some quick thoughts on differences:

• Both want to give users the ability to join raster and vector data. CARTO focuses on geometry types (via DB operations), whereas we focus on points (letting geopandas, etc. handle geometry).
• Here, we are storage system agnostic (Zarr, Xee, NetCDF, etc.; most likely in cloud storage) and let the DB operations happen at the application later (a batch dask job). CARTO involves data in a data lake (BQ; requires ingestion via raster-loader) as performs operations at the DB layer. One removes an ETL step, but the other gets the advantages of data locality and mature querying capabilities.
• This is OSS and you have to roll your own implementation. The other is proprietary software, a batteries-included SaaS.