Closed thehomebrewnerd closed 7 months ago
Would you consider having an ibis backend? It in return supports 20+ backends (including Spark and Dask). #1913 would be solved for free as well.
@AlpAribal It would be great to find a way to more easily support different dataframe types in Featuretools, especially if it would ease the maintenance burden inside Featuretools. I think that probably requires a much deeper study before we launch into that effort though. In any case, thank you for the suggestion and idea!
Maintenance of support for creating EntitySets from Dask and Spark dataframes has been problematic as breaking changes are introduced in newer versions of
dask
andpyspark
. Additionally, implementation for those dataframe types are still incomplete relative to the pandas implementations, with several features and many primitives still unsupported after several years. We should remove support for these dataframe types and focus Featuretools on what it does best, generating features for pandas dataframes.We should keep the ability to do parallel calculation of a feature matrix using Dask, however.