Group by nested column doesn't work

[shcheklein]

This code breaks. Can we be related to flattening of the model, but we still need to handle this properly.

    df = (
        dc
            .read_dataset("analyzer.bmi")
            .group_by(count=func.count(), partition_by="person.z")
            .to_pandas()
    )

Workaround:

    df = (
        dc
            .read_dataset("analyzer.bmi")
            .mutate(z=C("person.z"))
            .group_by(count=func.count(), partition_by="z")
            .to_pandas()
    )

[dmpetrov]

We have a pydantic class partitioning logic that generates a custom class from a subset of columns from another class. I guess, it's not in use in group-by. Or just some silly mistake 🙂

[shcheklein]

Yes, I remember that we added it to handle similar cases.