NewYour coding agent can read the release notes before it upgrades.Set up the MCP server →
PyPI, #425 most downloaded on PyPI
Blazingly fast DataFrame library
Latest stable
2.1.0
Last release 28 days ago
25 Oct 2026
Ships fairly regularly
a new release about every 3 weeks
Nearly every release is documented
notes for 20 of 21 stable releases
11 versions withdrawn
withdrawn after publishing
1 years old
38 releases · first in 2025
Drag across the ruler to compare, or tap a release to open it.
Disallow mixed string <=> numeric arithmetic in eager
List(Array) and Array columns without per-row slices (#30278)is_in (#30276)by keys (#30192)over (#30066)TopK for sorts sliced at a nonzero offset (#29989)coalesce (#30013)str.len_chars (#29972)str.replace for patterns that end in .*$ (#29971)str.contains and SQL LIKE (#29953)Struct data (#29813)arg_min and arg_max (especially on data with nulls) (#29849)from_rows and from_dicts memory usage during nested value conversion (#29739)== and is_in (#29885)sort().first() to min() and single-key sort_by().first() to min_by() (#29860)sort_by().first() to min_by and limit sorts read with head() (#29831)sort_by keys in the optimizer (#29812)CREATE TABLE with column definitions in attached Iceberg catalogs (#30260)CREATE [OR REPLACE] TEMP TABLE (#30256)DROP TABLE [PURGE] on tables of attached catalogs (#30247)overwrite_filter and dynamic partition overwrite to sink_iceberg (#29909)JOIN ... ON subqueries that read only one input (#30206)WITH RECURSIVE when no CTE reads itself (#30166)CURRENT_DATE, CURRENT_TIME and CURRENT_TIMESTAMP in SQL (#30144)reinterpret between Int128 and Decimal (#30133)Expr.truncdiv for division rounding toward zero (#30138)JOIN ... ON and correlated scalar subqueries without an aggregate (#30131)create_namespace_if_not_exists to sink_iceberg (#30087)SQLContext via register_catalog (#30113)pl.delete_iceberg (#30032)read_ipc table function (#30054)read_parquet table function (#29998)read_ndjson table function with named options (#29954)read_csv (#29975)REGEXP_REPLACE and DATE_TRUNC to SQL (#29967)generate_series and range to SQL (#29912)create_if_not_exists to sink_iceberg (#29908)read_csv table function (#29917)Array(Struct) rows field by field (#30290)Array rows (#30286)approx_quantile sketches deterministically (#30263)cov/corr output dtype (#30262)Float16/Float32 (#30158)str.replace for nested repetitions before .*$ (#30258)write_iceberg into Iceberg tables with required columns (#30253)read_avro (#30240)FROM item aliases from acting as tables (#30207)CREATE TABLE of a name in the default catalog (#30235)scan_delta file skipping (#30194)Struct and Array inside list (#30200)Float16/Float32 divided by Decimal (#30093)is_first_distinct and is_last_distinct (#30162)register_catalog for IF EXISTS, re-registration and scopes (#30160)String operands before float arms in true division type resolution (#30018)Note truncated.
Your coding agent can read these notes before it upgrades. Set up the MCP server →
Run SQL window functions over grouped rows on the aggregated rows, and evaluate QUALIFY before the projection
%/DIV (#29536)ScalarColumns in Column::append/extend (#28989)DataFrame::estimated_size (#29711)BytecodeParser method rewriting (#29632)from_dicts records straight into column buffers (#29656)as_list with only a single input (#29592)BytecodeParser (#29625)is_nan/is_not_nan filters to Iceberg (#29579)str.starts_with filters to Iceberg (#29545)BytecodeParser rewrite/dispatch (#29548)!= filters to Iceberg (#29455)is_in expression (#29425)scan_lines (#29310)CloudScheme::Http sources (#29284)sort_in_place (#29343)is_in and Map lookup coercion from dtypes alone (#29487)is_in and Map lookups exactly or not at all (#29486)Map in JSON and NDJSON reading and writing (#29478)scan_external_reader (#29232)struct.eval (#29454)BytecodeParser UDF variable resolution (#29553)%/DIV (#29536)BytecodeParser for Python 3.15 (#29491)scan_fn.io_source (#28897)join keys (#29233)APPROX_QUANTILE to the SQL frontend (#29288)approx_quantile in the streaming engine (#29237)scan_iceberg/delta-related attributes in the visitor for cudf_polars (#29297)collect and collect_batches using RemoteEngine (#28914)String to Time (#29215)is_in prepares a string haystack (#29744)asyncio extra (#29713)DataFrame "orient" behaviour (#29679)list.contains on multi-chunk sliced lists (#29635)is_sorted order nested types as sort does (#29683)is_in (#29593)min_by/max_by in streaming engine (#29628)is_nan/is_not_nan pushdown against Decimal columns in Iceberg (#29610)group_by predicate pushdown (#29499)LRUCache key (#29551)* and / (#29520)str.contains for string containment in map_elements UDFs (#25472)Note truncated.
Read Parquet ENUM type as pl.String
Map dtype (#28984)scan_lines (#29310)CloudScheme::Http sources (#29284)sort_in_place (#29343)posix_fadv to Normal (#29157)mmap with file I/O for parquet scan (#29091)head requests (#29131)scan_fn.io_source (#28897)join keys (#29233)APPROX_QUANTILE to the SQL frontend (#29288)approx_quantile in the streaming engine (#29237)scan_iceberg/delta-related attributes in the visitor for cudf_polars (#29297)collect and collect_batches using RemoteEngine (#28914)String to Time (#29215)organization to Config.enable_monitoring (#29221)Map dtype (#28984)read_database for Arrow-based drivers (#29273)Date and Decimal means between the streaming and in-memory engines (#29359)Struct supertypes per-field (#29261)auto (#29236)Wildcard selector (#29220)AnyValues through their field dtypes (#29185)OVER clause to multi-argument aggregates (#29160)DataFrame construction race (and generator data loss) (#29132)read_database Oracle regression, following introduction of Arrow fast-path (#29142)Decimal true division with integers (#29122)exp/log1p on non-numeric dtypes (#29112)offset_by when target date is out-of-range (#29059)is_empty/ has_nulls (#29152)fast-release profile without LTO (#29168)0.23.45 (#29303)test_fused_many_morsels_and_skew test (#29290)SpillFrames in DataFrameSearchBuffers (#29207)MapChunked storage handling (#29248)lf.collect().schema == lf.collect_schema() check (#29224)object_store for new DnsResolver trait (#29145)map_elements (#24925)Thank you to all our contributors for making this release possible! @0guban0v, @ATL2001, @Aidavdw, @AlessandroKuz, @BitWeaverDev, @DeflateAwning, @EndPositive, @Kevin-Patyk, @MarcoGorelli, @SatvikMishra08, @aarushkandukoori, @abokhalill, @alexander-beedie, @atharva7905k, @ayushh0110, @borchero, @carnarez, @dancsi, @dsprenkels, @freundTech, @fsimkovic, @jonasdedden, @kafka1991, @kdn36, @lun3x, @madsbk, @matthewbayer, @mikhail5555, @mroeschke, @nameexhaustion, @orlp, @r-brink and @ritchie46
Python Polars 2.0.0-rc.2 Pre-release
Pre-release
Compare
Set the default engine for SQL to the streaming engine
meta namespace (#28091)Date feature gating (#29031)Series.sample order when with_replacement=True and shuffle=False (#28990)PostApplyExtraOps (#29049)combine_validities_and_many and add test (#28909)sink_batches lazy: Literal[False] overload a default (#28951)if for cfg(debug_assertions) to fix the benchmark job (#29002)Thank you to all our contributors for making this release possible!
@0guban0v, @0xC61, @Aniket-a14, @LarryHu0217, @MatteoPossamai, @NeejWeej, @TNieuwdorp, @WaterWhisperer, @adamreeve, @dsprenkels, @jonasdedden, @kdn36, @math-hiyoko, @matthewbayer, @nameexhaustion, @orlp, @ritchie46, @severinh and @sivakumar-mahalingam
Python Polars 2.0.0-rc.1 Pre-release
Pre-release
Compare
Thank you to all our contributors for making this release possible! @EndPositive , @dsprenkels , @kdn36 , @lun3x , @nameexhaustion and @orlp
Thank you to all our contributors for making this release possible!
@EndPositive, @dsprenkels, @kdn36, @lun3x, @nameexhaustion and @orlp
Thank you to all our contributors for making this release possible! @Kevin-Patyk, @henrytsanford, @jonasdedden, @orlp and @ritchie46
Thank you to all our contributors for making this release possible! @Kevin-Patyk, @henrytsanford, @jonasdedden, @orlp and @ritchie46
Deprecate rechunk parameter for all read/scan functions
rechunk parameter for all read/scan functions (#28063)Expr.rechunk() (#28692)struct.rename_fields() with an incorrect number of fields (#28672)CloudRetryConfig for rate limit stability (#28885)mask for DataFrame filter (#28762)Partial metadata on filter (#28737)join_where (#28880)RemoteEngine and a common base class for all engines (#28800)scan_iceberg (#28772)sinked_paths_callback on sink_parquet as unstable parameter (#28814)infer_schema_files to CSV inference hint (#28809)struct.drop() (#28666)unpivot column selection aligns with the lazy engine (#28846)select() height (#28751)monitoring as an engine-level parameter (#28821)min and max (#28789)any and all (#28754)by column values in rolling_*_by (#27367)field in struct.with_fields with over (#28678)Utf8 and Binary (#28662)__arrow_c_stream__ (#28642)LargeList (#28632)read_database_uri requirements for SQLAlchemy (#28366)monitoring as an engine-level parameter (#28821)RemoteEngine and a common base class for all engines (#28800)CredentialProviderAWS and CredentialProviderAzure (#28790)ScanCastOptions (#27906)test_grouped_agg_parametric as slow (and add param ids) (#28715)expect in test_group_by_arg_max_boolean_26978 (#28690)struct.drop() in rename_fields() deprecation message (#28687)dist=loadgroup to the pytest config (#28676)may_fail_auto_streaming tests (#28675)test_extension() for streaming engine (#28611)test_hive_join_rewrite_semi_join test to work with streaming engine (#28610)Thank you to all our contributors for making this release possible! @0guban0v, @JakubValtar, @Joosboy, @Kevin-Patyk, @MatteoPossamai, @NicoOhR, @TNieuwdorp, @VedantMadane, @aarushkandukoori, @alexander-beedie, @azimafroozeh, @borchero, @carnarez, @dancsi, @dependabot[bot], @dsprenkels, @gautamvarmadatla, @jonasdedden, @jorenham, @kdn36, @lzcmian, @malhotrashivam, @matthewbayer, @mikhail5555, @nameexhaustion, @orlp, @ritchie46, @subotac and dependabot[bot]
Deprecate casts from Categorical to integer dtypes
Categorical to integer dtypes (#28525)plan_stage argument in show_graph() (#28391)len() to concat/union inputs (#28570)from_arrow of ArrowStreamExportable (#28442)infer_schema_files parameter to scan_csv (#28440)Series(Array) (#28602)StringChunked substring kernels (#28573)Unknown(Int) and Unknown(Float) should result in Unknown(Float) (#28545)nulls_last after Expr.reverse() (#28572)SQLContext.execute() (#28549)nulls_last in function_expr_sortedness (#28544)Thank you to all our contributors for making this release possible! @BitWeaverDev, @Matt711, @Samoilov2004, @borchero, @dancsi, @dsprenkels, @kdn36, @masumi-ryugo, @nameexhaustion and @ritchie46
Optimize not(bool\_f) to not\_bool\_f
NOT IN interaction with NULL values and joins (#28484)SUM and CORR aggregates return NULL for all-null inputs, add TOTAL (#28475)null_count_dtype helper between Delta and Iceberg, fixing SchemaError (#28479)ruff and mypy package versions (#28456)Thank you to all our contributors for making this release possible! @AnirudhRahul, @EndPositive, @Jesse-Bakker, @alexander-beedie, @carnarez, @dancsi, @mdroogh, @mroeschke, @nameexhaustion, @nchammas, @orlp, @r-brink and @ritchie46
Deprecate casting numeric types to categoricals
cat.get_categories() and cat.to_local() (#28299)LazyFrame.profile() (#28275)list/arr.to_struct() calls that don't pass field names (#28243)missing_utf8_is_empty_string to empty_string_is_null (#28173)full_null() with small lengths (#28181)scan_arrow_c_stream (#28340)initial-default in native scan_iceberg (#28324)assert_frame_equal error (#27816)list expression which consistently packs elements together into new List type (#27990)read_database Arrow fast-path for "python-oracledb" (#28230)CastColumnsPolicy (#28084)dt.replace when there were multiple chunks (#28437)is_scalar from the input to the output of .sort() and .sort_by() (#28438)sort().reverse() to sort(descending=True) when maintain_order=True (#28403)write_json() null values in Array columns being written incorrectly as null (#28330)select(len()) (#28355)top_k/bottom_k (#28359)aws_checksum_algorithm when provided (#28328)pyarrow.feather.read_table (#28323)Object dtype instead of creating invalid List(Object) (#28246)ewm_var/ewm_std value (#28235)ADBC shouldn't require "CREATE" privileges unless the table is confirmed not to exist (#28228)rolling_rank position invalid if ost.len() < min_samples (#28142)merge_sorted() docs also for DataFrame (#28183)merge_sorted() null inputs should be nulls first (#28177)dev profile to line-tables-only (#28358)shuffle parameter for sample() (#27460)to_struct() in doctest (#28370)spin dependency to v0.10.1 (#28360)assert_frame_equal error (#27816)optimize (#28318)object_store custom DNS resolver patch to rev-based (#28241)*GroupBy (#27903)classinstmethod and attribute annotation (#27904)typos to v1.48.0 (#28162)Thank you to all our contributors for making this release possible! @0guban0v, @AnirudhRahul, @Bharath-970, @CodingSelim, @EndPositive, @Jesse-Bakker, @Kevin-Patyk, @LarryHu0217, @MarcoGorelli, @MatteoPossamai, @TNieuwdorp, @alex-h-sun, @alexander-beedie, @azimafroozeh, @borchero, @dancsi, @dependabot[bot], @dsprenkels, @harrywhalen, @jorenham, @kdn36, @makinzm, @malhotrashivam, @mdavis-xyz, @mikhail5555, @mkzung, @mroeschke, @nameexhaustion, @orlp, @r-brink, @raphaelroshan, @ritchie46, @tylerriccio33, @useredsa, @wence-, @yannbolliger, @zichen0116 and dependabot[bot]
Deprecate strict parameter of pl.concat, replace with new how='horizontal_extend'
strict parameter of pl.concat, replace with new how='horizontal_extend' (#27965)select(len()) after groupby (#28108)ResourceWarning leaks in database/iceberg tests (#28107)DeferredRefreshableCredentials (#28099)replace when old/new contain Expr or object dtype values (#27433)tokio::spawn in clippy (#28123)Makefile with respect to venv robustness (#28110)pd.TimeDelta segfault (#28125)mypy to the new 2.x release (#28116)pyrefly checks run as part of make pre-commit (#28095)test_select_explode_height_filter_order_by failure when POLARS_MAX_THREADS=2 (#28090)Thank you to all our contributors for making this release possible! @0guban0v, @Kevin-Patyk, @TNieuwdorp, @alexander-beedie, @azimafroozeh, @dsprenkels, @kdn36, @nameexhaustion, @orlp, @raphaelroshan, @ritchie46 and @sar-cheng
One column per month.