`dask.dataframe` `DeprecationWarning` turns into error (no `dask.dataframe`)

[jakirkham]

Seeing the dask.dataframe deprecation warning now turning into an exception on CI:

________ ERROR collecting dask_cuda/tests/test_cudf_builtin_spilling.py ________
tests/test_cudf_builtin_spilling.py:5: in <module>
    from dask_cuda.device_host_file import DeviceHostFile
/opt/conda/envs/test/lib/python3.10/site-packages/dask_cuda/__init__.py:8: in <module>
    import dask.dataframe.core
/opt/conda/envs/test/lib/python3.10/site-packages/dask/dataframe/__init__.py:173: in <module>
    warnings.warn(
E   DeprecationWarning: The current Dask DataFrame implementation is deprecated. 
E   In a future release, Dask DataFrame will use a new implementation that
E   contains several improvements including a logical query planning.
E   The user-facing DataFrame API will remain unchanged.
E   
E   The new implementation is already available and can be enabled by
E   installing the dask-expr library:
E   
E       $ pip install dask-expr
E   
E   and turning the query planning option on:
E   
E       >>> import dask
E       >>> dask.config.set({'dataframe.query-planning': True})
E       >>> import dask.dataframe as dd
E   
E   API documentation for the new implementation is available at
E   https://docs.dask.org/en/stable/dask-expr-api.html
E   
E   Any feedback can be reported on the Dask issue tracker
E   https://github.com/dask/dask/issues 
E   
E   To disable this warning in the future, set dask config:
E   
E       # via Python
E       >>> dask.config.set({'dataframe.query-planning-warning': False})
E   
E       # via CLI
E       dask config set dataframe.query-planning-warning False
__________ ERROR collecting dask_cuda/tests/test_dask_cuda_worker.py ___________
tests/test_dask_cuda_worker.py:15: in <module>
    from dask_cuda.utils import (
/opt/conda/envs/test/lib/python3.10/site-packages/dask_cuda/__init__.py:14: in <module>
    from .cuda_worker import CUDAWorker
/opt/conda/envs/test/lib/python3.10/site-packages/dask_cuda/cuda_worker.py:24: in <module>
    from .proxify_host_file import ProxifyHostFile
/opt/conda/envs/test/lib/python3.10/site-packages/dask_cuda/proxify_host_file.py:40: in <module>
    from . import proxify_device_objects as pdo
/opt/conda/envs/test/lib/python3.10/site-packages/dask_cuda/proxify_device_objects.py:10: in <module>
    from .proxy_object import ProxyObject, asproxy
/opt/conda/envs/test/lib/python3.10/site-packages/dask_cuda/proxy_object.py:896: in <module>
    @dask.dataframe.core.get_parallel_type.register(ProxyObject)
E   AttributeError: module 'dask' has no attribute 'dataframe'

[pentschev]

This is probably related to the unpinning of Dask versions in rapidsai/rapids-dask-dependency#25 , not sure why those errors didn't show up before today.

@rjzamora should we revert the aforementioned PR for now? If so, can you open a PR with that change?

[rjzamora]

@rjzamora should we revert the aforementioned PR for now? If so, can you open a PR with that change?

Sounds good - I opened rapidsai/rapids-dask-dependency#29 to revert that change.

[jakirkham]

Thanks Peter and Rick! 🙏

Looks like that helped. Will go ahead and close this