ddf.drop is inconsistent when passed a set of columns about dask HOT 4 CLOSED

Thanks for the report, put up a PR to fix. I wouldn't expect this to work consistently in pandas though, sets aren't officially supported

from dask.

Is this why dask moved away from pd.api.types.is_list_like(columns) ? Because the pandas docs say list-like for the columns kwarg, so i cant imagine they aren't using this fn under the hood for type checking

from dask.

This doesn't completely work with cudf IIUC, cc @rjzamora

from dask.

This doesn't completely work with cudf IIUC

I believe cudf's drop(columns=...) will accept anything blessed by pd.api.types.is_list_like, but there may be corner cases. That said, from a maintenance perspective, I'd definitely prefer if list was the only type allowed :)

from dask.