After the deprecation message when creating a dask.Dataframe, I’ve recently activated dask-expr as suggested in the Deprecation message in dask 2024.2.1. The code is used to open a csv file to clean a dataset. As part of this process, the code has casting calls such as df.col = df.col.astype(int). However, after activating the new dask-expr dataframe, casting seems to be completely ignored… even after calling compute() on the dask-expr dataframe. In the older dataframe casting was working as expected. Is this as expected or am I missing something important in the use of this new Dataframe? I’ve searched online and seemingly the API for the old dataframe is fully supported with a few exceptions so I wonder what the root cause could be. Any help appreciated.
Update. This issue seems to be WIP in the project. See Inconsistent casting behaviour with dask-expr Dataframe · Issue #10960 · dask/dask · GitHub