Daft 0.0.16 Release Notes#

Warning

Daft 0.0.15 was yanked from PyPi due to some issues in some wheels failing to build due to an issue with dependencies. 0.0.16 includes all changes in the yanked 0.0.15 package as well as the necessary fixes to the build.

Daft 0.0.16 includes big fixes and tests for handling nulls in dataframes:

  • Adds null tests and fixes for all global operations: sorts, groupbys, aggregates, joins, distinct

  • Type inference improved for DataFrame.from_pydict

Enhancements#

  • Benchmark script fix when using ray runner in local mode #286

  • Refactor benchmarking data schema to flatten along columns #283

  • Null tests for sorts/aggregates/joins/distinct and fixes #281

  • add new kernels for search sorted to handle nulls #279

  • Refactor of Hashing module to C++ for handling of seed and nulls #274

  • Adds a setup_ray step in benchmarking #256

Bug Fixes#

  • Remove dataframe_cookbook tests with 2 CSVs #284

  • Fix type inference in DataFrame.from_pydict #280

  • Search Sorted Offset Fix for Arrow Views #276

Closed Issues#

  • Verify offset chunked arrays work with search sorted #275

  • Fix hashing of floats #271

  • Fix type inference for DataFrame.from_pydict and DataFrame.from_pylist #233