spatialpandas icon indicating copy to clipboard operation
spatialpandas copied to clipboard

Integrate with Kartothek

Open dharhas opened this issue 5 years ago • 1 comments

Is your feature request related to a problem? Please describe.

SpatialPandas helps spatially sort data but we are seeing the need for higher level arbitrary indexing. Two example use cases:

  • Geospatial. We have spatially sorted daily GPS data for the US for multiple days. Getting a small region for a 60-90 day process can get bogged down by the need to read the 60-90 multiple metadata files and construct the task graph.

  • Astronomy. We have spatial data for multiple filters (HSC-Y, HSC-G etc). Again we would have to read multiple metadata files.

Describe the solution you'd like

The above could be fixed by building higher level indexes. I think we can benefit from integrating with kartothek. It enables an O(1) index and creates the necessary task graphs for reading just the partitions required. It could also be used to store the extra metadata spatialpandas currently stores in its own format (if I'm understanding spatialpandas correctly)

I'm at the Dask Dev Conference with some of the Kartothek devs and based on conversations with fjetter this integration should be possible.

dharhas avatar Feb 27 '20 16:02 dharhas

Sounds cool to me!

jbednar avatar Jul 16 '20 22:07 jbednar