Modin 0.15.0
This release includes updated support for pandas 1.4.2, new Batch and Logging APIs, and a plethora
of bug fixes and documentation improvements.
Key Features and Updates
- Stability and Bugfixes
- FIX-#4376: Upgrade pandas to 1.4.2 (#4377)
- FIX-#3615: Relax some deps in development env (#4365)
- FIX-#4370: Fix broken docstring links (#4375)
- FIX-#4392: Align Modin XGBoost with xgb>=1.6 (#4393)
- FIX-#4385: Get rid of
use-deprecated
option inpip
(#4386) - FIX-#3527: Fix parquet partitioning issue causing negative row length partitions (#4368)
- FIX-#4330: Override the memory limit to start ray 1.11.0 on Macs (#4335)
- FIX-#4407: Align
insert
function with pandas in case of numpy array with several columns (#4408) - FIX-#4373: Fix invalid file path when trying
read_csv_glob
withusecols
parameter (#4405) - FIX-#4394: Fix issue with multiindex metadata desync (#4395)
- FIX-#4438: Fix
reindex
function that doesn't preserve initial index metadata (#4442) - FIX-#4425: Add parameters to groupby pct_change (#4429)
- FIX-#4457: Fix
loc
in case when need reindex item (#4457) - FIX-#4414: Add missing f prefix on f-strings found at https://codereview.doctor/ (#4415)
- FIX-#4461: Fix S3 CSV data path (#4462)
- FIX-#4467:
drop_duplicates
no longer removes items based on index values (#4468) - FIX-#4449: Drain the call queue before waiting on result in benchmark mode (#4472)
- FIX-#4518: Fix Modin Logging to report specific Modin warnings/errors (#4519)
- FIX-#4481: Allow clipping with a Modin Series of bounds (#4486)
- FIX-#4504: Support na_action in applymap (#4505)
- FIX-#4503: Stop the memory logging thread after session exit (#4515)
- FIX-#4531: Fix a makedirs race condition in to_parquet (#4533)
- FIX-#4464: Refactor Ray utils and quick fix groupby.count failing on virtual partitions (#4490)
- FIX-#4436: Fix to_pydatetime dtype for timezone None (#4437)
- FIX-#4541: Fix merge_asof with non-unique right index (#4542)
- Performance enhancements
- Benchmarking enhancements
- Refactor Codebase
- REFACTOR-#4284: use variable length unpacking when getting results from
deploy
function (#4285) - REFACTOR-#3642: Move PyArrow storage format usage from main feature to experimental ones (#4374)
- REFACTOR-#4003: Delete the deprecated cloud mortgage example (#4406)
- REFACTOR-#4513: Fix spelling mistakes in docs and docstrings (#4514)
- REFACTOR-#4510: Align experimental and regular IO modules initializations (#4511)
- REFACTOR-#4284: use variable length unpacking when getting results from
- Developer API enhancements
- Update testing suite
- TEST-#4363: Use Ray from pypi in CI (#4364)
- FIX-#4422: get rid of case sensitivity for
warns_that_defaulting_to_pandas
(#4423) - TEST-#4426: Stop passing is_default kwarg to Modin and pandas (#4428)
- FIX-#4439: Fix flake8 CI fail (#4440)
- FIX-#4409: Fix
eval_insert
utility that doesn't actually check results ofinsert
function (#4410) - TEST-#4482: Fix getitem and loc with series of bools (#4483).
- Documentation improvements
- DOCS-#4296: Fix docs warnings (#4297)
- DOCS-#4388: Turn off fail_on_warning option for docs build (#4389)
- DOCS-#4469: Say that commit messages can start with PERF (#4470).
- DOCS-#4466: Recommend GitHub issues over [email protected] (#4474).
- DOCS-#4487: Recommend GitHub issues over [email protected] (#4489).
- Dependencies
- New Features
Contributors
@YarShev
@Garra1980
@prutskov
@alexander3774
@amyskov
@wangxiaoying
@jeffreykennethli
@mvashishtha
@anmyachev
@dchigarev
@devin-petersohn
@jrsacher
@orcahmlee
@naren-ponder
@RehanSD