Support for other DataFrame libraries

[lucas-nelson-uiuc]

Wondering if there's been discussion for supporting other dataframe (or pandas-compatible) libraries? Happy to talk through or help develop support to extend the ecosystem.

[willeppy]

Hey thanks for starting discussion on this, I think would be really cool. A couple of libraries that come to mind, curious which ones you or others are using the most that would be the best to support

• Polars
• Ibis
• Modin -- this might work with minimal overhead since has exact same API as Pandas as my understanding
• Raw numpy matricies (if 2d)

[willeppy]

Generally, in terms of implementation I think would involve...

• Python pandas executor to support different libraries and detect different dataframe types in the front end code that is executed (https://github.com/cmudig/AutoProfiler/blob/main/src/dataAPI/jupyter/PythonExecutor.ts)
• And then implementing the required functions in python like we have for pandas (https://github.com/cmudig/AutoProfiler/blob/main/digautoprofiler/profile_lib.py)