modin-project/modin 0.17.0

Modin 0.17.0

on GitHub

This release includes support for pyhdk 0.2. It also includes many bug fixes and some performance enhancements.

Key Features and Updates Since 0.16.0

• Stability and Bugfixes
  • FIX-#3764: Ensure df.loc with a scalar out of bounds appends to df (#3765)
  • FIX-#4016, FIX-#4086, FIX-#4039: Fall back to pandas in case of duplicate column names (#4896)
  • FIX-#4023: Fall back to pandas in case of MultiIndex columns (#5149)
  • FIX-#4660: Fix fillna when Modin series object is an argument (#4674)
  • FIX-#5034: Handle lists in df.get() (#5035)
  • FIX-#5097: Stop using deprecated mangle_dup_cols. (#5104)
  • FIX-#5098: Stop using append internally. (#5100)
  • FIX-#5099: Fix PandasQueryCompiler.groupby_mean with timestamp in by (#5140)
  • FIX-#5112: allows empty partition to be passed into query_compiler.dt_prop_map (#5133)
  • FIX-#5128: Fix reading parquet directory from s3. (#5129)
  • FIX-#5150: Sync row labels after read_csv when index_col is False (#5151)
  • FIX-#5158: Synchronize metadata before to_parquet (#5161)
  • FIX-#5168: module 'collections' has no attribute 'Sequence' in dataframe protocol (#5169)
  • FIX-#5174: Pin xgboost < 1.7. (#5175)
  • FIX-#5180: Do not set OMP_NUM_THREADS=1 on modin.pandas init (#5181)
  • FIX-#5184: Fix get_dummies to respect passed columns to be encoded (#5185)
  • FIX-#5188: Fix getitem_bool when the key is Series with empty partition (#5189)
  • FIX-#5206: pin mypy<0.990 (#5207)
  • FIX-#5208: pin ray version under 2.1.0 (#5209)
• Performance enhancements
  • PERF-#5029: Don't use _compute_axis_labels_and_lengths for computing _row_lengths/_column_widths (#5030)
  • PERF-#5087: use cache for widths/lengths/index/columns if possible (#5031)
  • PERF-#5162: precompute new row/column lengths in '._reorder_labels' (#5144)
• Refactor Codebase
  • REFACTOR-#4631: Add mypy checks for modin.distributed (#5109)
  • REFACTOR-#5079: Add mypy checks for modin.core.dataframe.base (#5110)
  • REFACTOR-#5092: Fix future warning for set_axis function (#5093)
• Update testing suite
  • TEST-#4982: Require format for PR descriptions instead of commit descriptions (#5117)
  • TEST-#5124: Disable codecov comments. (#5125)
  • TEST-#5135: Return CI back after accidental removal (#5136)
  • TEST-#5172: Add fuzzydata logs to artifacts (#5173)
• Benchmarking enhancements
  • BENCH: add some cases for join and merge ops from pandas (#5021)
  • TEST-#5102: Add HDK benchmarks to github workflows (#5063)
• Documentation improvements
  • DOCS-#3634: Fix examples related to ProgressBar usage (#5119)
  • DOCS-#5019: Update HDK on native documentation (#5088)
  • DOCS-#5095: Remove release note checkbox from PR template (#5096)
  • DOCS-#5105: Update release procedure (#5106)
• New Features
  • FEAT-#5120: Update to pyhdk 0.2 (#5121)
  • FEAT-#5141: Implement 2D insertion of Modin DFs in .__setitem__ (#5142)
  • FEAT-#5145: Upgrade pandas to 1.5.1 (#5146)

Contributors

@AndreyPavlenko
@Billy2551
@RehanSD
@YarShev
@anmyachev
@dchigarev
@devin-petersohn
@ienkovich
@mvashishtha
@noloerino
@pyrito
@rosdyana
@shalearkane
@suhailrehman
@vnlitvinov