pola-rs/polars rs-0.41.3

Rust Polars 0.41.3

on GitHub

🚀 Performance improvements

• Improve unique performance by adding RangedUniqueKernel for primitive arrays (#17166)
• faster decode on Parquet HybridRLE (#17208)

✨ Enhancements

• Add SQL support for NATURAL joins and the COLUMNS function (#17295)
• Add str.extract_many expression (#17304)
• Support '%' in pathnames for async scan (#17271)
• Support SQL Struct/JSON field access operators (#17226)
• Exclude directories from glob expansion result (#17174)
• Support SQL ORDER BY ALL syntax (#17212)
• Support PostgreSQL ^@ ("starts with"), and ~~,~~*,!~~,!~~* ("like", "ilike") string-matching operators (#17251)
• Support SQL SELECT * ILIKE wildcard syntax (#17169)
• Support SQL temporal functions STRFTIME and STRPTIME, and typed literal syntax (#17245)
• Support date/datetime for hive parts (#17256)
• Expose some more information in translated expression IR to python (#17209)
• Allow no-op round/ceil/floor on integer types (#17241)
• Support loading from datasets where the hive columns are also stored in the file (#17203)
• Implement serde for Null columns (#17218)
• Support Decimal types in write_csv/write_json (#14209)
• Improve SQL support for array indexing, increase test coverage (#16972)
• Support reading byte stream split encoded floats and doubles in parquet (#17099)
• Add float_scientific option to write_csv/sink_csv (#17111)

🐞 Bug fixes

• Raise proper error for mismatching parquet schema instead of panicking (#17321)
• Raise on invalid shape dataframe arithmetic (#17322)
• Fix panic in window case (#17320)
• Raise errors instead of panicking when sink_csv fails (#17313)
• Raise if join keys are passed to cross join (#17305)
• Don't null on oob in list.get for column index (#17276)
• Fix issue where sliced PyArrow record batches were not handled correctly (#17058)
• Don't oob on nulls in list.get (#17262)
• Fix list getter with nulls (#17261)
• Respect nulls_last parameter in aggregate sort_by (#17249)
• Fix literal slice in group by (#17242)
• Fix DataFrame.top_k not handling nulls correctly (#17239)
• Avoid using the regex dependency when the regex feature is not used (#17206)
• properly check the BMI2 uleb128 (#17191)

📖 Documentation

• Minor layout/terminology improvement for selector set ops (#17299)
• Fix polars-plan docs.rs build (#17266)
• Add SQL docs for the CAST and TRY_CAST functions (#17214)

🛠️ Other improvements

• Prefer ParquetError::oos to ParquetError::OutOfSpec (#17314)
• remove seqmacro and u8,u16 bitpack (#17290)
• Fix typo in join validation error message (#17296)
• Use typed iter in list.get (#17286)
• add ability to have pipeline blockers in new streaming engine (#17247)
• Support date/datetime for hive parts (#17256)
• Add elementwise select and with_columns to new streaming engine (#17185)
• chrono's ParseErrorKind is now public (#17201)

Thank you to all our contributors for making this release possible!
@IvanIsCoding, @JamesCE2001, @MarcoGorelli, @SeanTater, @adamreeve, @alexander-beedie, @coastalwhite, @datapythonista, @flisky, @itamarst, @jqnatividad, @lukeshingles, @mcrumiller, @nameexhaustion, @orlp, @ritchie46, @stinodego and @wence-