This release contains significant amounts of improvements to the
maintainability of the code and bugfixes. Multiple new additions
were made to the pandas API coverage.
Key Features and Updates
------------------------
* Stability and bugfixes
* Fix __setitem__ when key is unhashable list (57bcfc1)
* Fix slice_shift when index has duplicates (cf47333)
* Pandas API implementations and improvements
* experimental ray implementation of read_pickle, to_pickle (cdf47ac)
* Add support for Series.str.__getitem__ (7631def)
* skiprows support added for read_csv (b8098bd)
* support local variables in query and eval (0a64275)
* Add storage_options param for read_parquet (da2ad79)
* Fixed 'value_counts' implementation (ebd07dd)
* Warn user about heterogeneous data presence during read_csv (15f168c)
* Add attribute api in modin.pandas (9865616)
* XGBoost enhancements
* Add async execution support for Modin xgb.predict (ab58612)
* Fix processing of evals parameter in Modin xgb (9bef861)
* Developer API enhancements
* Add additional parameters for from_partitions (ca43e8d)
* add a way to get custom shapes (984e68f)
* use MODIN_MEMORY to specify memory for dask engine (8a3b105)
* Expansion in testing
* Documentation improvements
* pandas 1.3.3 support (f91ee0a)
* Omnisci backend enhancements
* add value_counts benchmark for OmniSci backend (600fa26)
* Get rid of Ray when using OmniSci engine (ac8e2d1)
* Support columns renaming in arrow execution. (3079783)
* support logical 'and' and 'or' in filters. (ecaab1b)
* fix dtypes for OmniSci dataframes. (02a34eb)
* update Arrow to 3.0 for OmniSci backend (411b322)
Contributors this release
-------------------------
The following users contributed code to Modin since the last release.
@YarShev
@anmyachev
@dchigarev
@vnlitvinov
@gshimansky
@prutskov
@amyskov
@krfricke
@fexolm
@devin-petersohn