The Dask team released version 2022.04.1 last week! It includes contributions from 21 people, including 4 new contributors.
Check out the complete changelog here, some highlights include:
- The deprecated “pyarrow-legacy” engine, which was used to read/write parquet files has now been removed. This engine supported pyarrow<1.0.0 and the removal helps simplify maintenance efforts. Thank you @rjzamora for working on this!
- Along the same lines, PyArrow will soon replace fastparquet as the default engine for working with parquet files. Thanks for adding a warning about this, @jcristharif!
- Dask Array is now more consistent with NumPy, thanks to @jsignell for updating the output types on some elementwise functions, and Tom White for adding some missing NumPy universal functions (ufuncs).
- The team has also continued efforts to improve Dask’s stability. This release contains several updates to Dask’s worker state machine.
Finally, thank you @jrbourbeau for managing the release!