Open charlesbluca opened 1 month ago
Looks like we should unblocked to support left anti joins when dataframe.backend="cudf", similar to the case in legacy Dask dataframe:
dataframe.backend="cudf"
https://github.com/dask/dask/blob/df4de6ea53054790b09006c8ea68ef8725d39025/dask/dataframe/multi.py#L565
Note that like the legacy code, we'll fail somewhere down in the comptutation stack if we try this on CPU - not sure if it makes sense to check the backend if how="leftanti" and eagerly raise a NotImplementedError if dataframe.backend != "cudf".
how="leftanti"
NotImplementedError
dataframe.backend != "cudf"
cc @rjzamora
Looks like we should unblocked to support left anti joins when
dataframe.backend="cudf", similar to the case in legacy Dask dataframe:https://github.com/dask/dask/blob/df4de6ea53054790b09006c8ea68ef8725d39025/dask/dataframe/multi.py#L565
Note that like the legacy code, we'll fail somewhere down in the comptutation stack if we try this on CPU - not sure if it makes sense to check the backend if
how="leftanti"and eagerly raise aNotImplementedErrorifdataframe.backend != "cudf".cc @rjzamora