The Rust crate polars has its own changelog.

Full Changelog

Merged pull requests:

• python collect_all function #1503 (ritchie46)
• python: remove deprecated code/funcitonallity #1502 (ritchie46)
• Refactor Date/Datetime dtypes #1501 (ritchie46)
• add vertical string concat; closes #1490 #1500 (ritchie46)
• fix bug in outer_join functions, add tests #1498 (marcvanheerden)
• refactor cast #1493 (ritchie46)
• Redesign logical types #1489 (ritchie46)
• python: concat also Series #1484 (ritchie46)
• fix outer join on floats and remove dtype-u64 flag #1483 (ritchie46)
• fill_nan expression and fix #1478 #1480 (ritchie46)
• Allowed non-vec in DF expressions. #1477 (jorgecarleitao)
• fix rank on data with nulls: closes #1473 #1475 (ritchie46)
• Python rename drop_column(s) to drop (union eager/lazy) and pl.exclud… #1472 (ritchie46)
• change memmap2 version from 0.2.0 -> 0.5.0 to support wasm #1470 (LemonPy29)
• add categorical anyvalue #1464 (ritchie46)
• Fix cat append #1462 (ritchie46)
• Optional pyarrow #1459 (ghuls)
• remove time64 dtype #1458 (ritchie46)
• fix agg_std/agg_var for Float32 #1449 (ritchie46)
• sort multiple with date64; closes #1437 #1438 (ritchie46)
• Lazy: rename columns + Join suffix #1433 (ritchie46)
• add reverse collect for numeric ChunkedArray #1430 (ritchie46)
• fix and test boolean Series indexing #1425 (ritchie46)
• elide bound checks on validity in take kernels #1419 (ritchie46)
• Use "sep" instead of "delimiter" in pl.Dataframe().to_csv(). #1418 (ghuls)
• fix rolling_mean on integers: closes #1411 #1413 (ritchie46)
• expose numeric bitwise on Series #1412 (ritchie46)
• improve hashing performance w/ specialized hashers #1405 (ritchie46)
• categorical aggregation output consistency #1403 (ritchie46)
• fix apply with list output type #1402 (ritchie46)
• expose categorical round trip to python and add more dictonary types:… #1399 (ritchie46)
• fix invalid ReplaceDropNulls optimization #1398 (ritchie46)
• fix memcpy of multiple chunks; closes 1396 #1397 (ritchie46)
• Workflow to generate m1 wheels #1394 (tiphaineruy)
• Add nan_to_none flag for converting NaN to None for from_pandas #1393 (mahadeveaswar)
• fix explode #1392 (ritchie46)
• improve performance of rolling_var/ rolling_std #1390 (ritchie46)
• python: More rolling_window functions #1389 (ritchie46)
• improve rolling_window performance #1387 (ritchie46)
• csv-parsing automatically parse dates #1386 (ritchie46)
• Rust; add parquet compression #1385 (ritchie46)
• python improve various ergonomics; Series.shift(expr), Series.shift_a… #1384 (ritchie46)
• add conversion between categorical <-> arrow dictionary #1380 (ritchie46)
• feature gate categorical dtype #1379 (ritchie46)
• Ipc: read_schema and projection option in read_ipc #1378 (ritchie46)
• keep col name in fill_null #1377 (ritchie46)
• python DataFrame.with_column -> pl.Series #1373 (ritchie46)
• proposal: implement __bool__ for pl.Expr #1368 (wseaton)
• fix bug in Lazy::apply #1367 (ritchie46)
• Lazy: fix inconsistency in apply_flat #1366 (ritchie46)
• Flate2 zlib ng #1365 (ritchie46)
• python make division consistent #1364 (ritchie46)
• remove non par sort branches. reduces code bloat #1363 (ritchie46)

Full Changelog

Closed issues:

• ShapeMisMatch error when vstacking two dataframes with equal amount columns, but different number of rows #1452
• split agg_std and agg_mean for floats #1447
• .std().over('groups') raises a PanicException when one of the columns has dtype Float32. #1446
• A numeric type is converted to a string type #1444
• Adding a column without stating df length #1440
• Sorting by multiple columns doesn't work when one of the columns is Date64 #1437
• Include a center option in the rolling function (like pandas) #1436
• Parquet File Size Larger than CSV File Size #1381
• Dead links in source #1370

Full Changelog

Closed issues:

• Adding an optional suffix to overlaping columns when joining dataframes #1432
• collect reverse #1429
• indexing bool column gives NotImplemented error #1422
• use sep instead of delimiter in DataFrame.to_csv #1415

Full Changelog

Closed issues:

• Rolling_mean only working with floats #1411
• Add numeric bitwise operations on Series/ Exprs #1410
• Polars date arithmetic is not correct #1404
• mean, median, mode, ... probably should not work for pl.Categorical #1401
• Initially empty series causes memcpy_values to index OOB #1396
• Filter gets mis-optimized and doesn't filter out false values #1395
• pl.DataFrame.explode() gives an empty name name now for the exploded column. #1391
• m1 wheels generation #1345
• Arrow dictionaries -> Polars Categorical #1308
• RuntimeError: Other("Could not determine output type") #1307
• Add from_pandas flag for converting NaN to None #1164

Full Changelog

Closed issues:

• add rolling_std #1388
• Allow fill_null to accept Expr for Series #1383
• getting AttributeError when using .is_in on polars 0.9.7, but works on polars 0.8.* #1382

Full Changelog

Closed issues:

• Faster gzip decompression #1359

Full Changelog

Closed issues:

• Series with large ints lose precision when divided #1369
• // integer division in python not working #1362
• Add a method to extract a column as a series #1346

Full Changelog

Full Changelog

Closed issues:

• Allow rounding within .agg() #1336
• Filtering of DataFrame based on the column with dates in Date32 format is not working #1332
• Filters sometime fill null string value with "" #1322
• Allow conversion from/to list of dicts #1300

Full Changelog

Full Changelog

Closed issues:

• read_csv: dtypes date32 and date64 are not inferred correctly anymore #1330
• add clip function #1326
• Allow .agg() to work on ungrouped data frames #1324
• Add .with_column_renamed() eager method #1323
• add two list array. #1316
• TypeError: large_list() takes exactly one argument (0 given) #1303
• parquet dict encoded panic #1281
• fix explode on lists with empty values #1177

Full Changelog

Closed issues:

• Add standard error as an aggregation expression (standard deviation / sqrt(number of measurements) ) #1315
• Make impossible datatime aggregations consistent with primitives #1311
• Add an argument strict to constructors, cast, etc to ensure the integer range etc on python polars (and otherwise) instead of silently converting to null. #1293

Full Changelog

Closed issues:

• data = data[range(N)] isn't very performant #1283

Full Changelog

Full Changelog

Closed issues:

• cannot explode dataframe with single list column #1288
• Sort dataframe using Date32 column is causing nulls to appear #1277
• Drop duplicates #1260
• rolling functions in lazy #1185

Full Changelog

Full Changelog

Full Changelog

Full Changelog

Closed issues:

• Aggregation fails when grouped by multiple columns with one column of datatype Int8 or Int16 #1255
• [Rust] cross_join coalesce left to right #1254
• rename all null to none #1247
• expr filter.count() gives wrong result #1242
• Lazy: hard error on duplicate names #1241
• horizontal_sum define null behavior #1173
• the new types with udf in apply function #1165

Full Changelog

Closed issues:

• pyo3_runtime.PanicException: should already be coerced to u64 when joining two DataFrames #1231
• Error when performing modulo operation '%" #1230
• CsvReader ignores the last charactor if there's no newline at EOF #1229
• Bad link in doc #1227

Full Changelog

Full Changelog

Closed issues:

• Remove HSTACK node #1218
• Groupby Rank #1209
• Alias dense_rank to argsort_by + 1 #1207

Full Changelog

Closed issues:

• Predicate pushdown using a filter with null values doesn't correctly trigger the filter #1217
• Lazy: use a lazy function to determine type of apply/map #1203
• [python] missing dependency when using Anaconda #1200
• Add transpose method to dataframe #1176

Full Changelog

Closed issues:

• Python: add rolling apply #1194
• Slow concat with large list of dataframes #1183

Full Changelog

Full Changelog

Full Changelog

Closed issues:

• Expose interpolate on DataFrame in core #1161
• [Python] df.filter fails with "out of range" error #1157

Full Changelog

Closed issues:

• shift_and_fill by groups + other operations by grouping variables #1124

Full Changelog

Closed issues:

• inspect/debug expr #1146

Full Changelog

Closed issues:

• Filter expression should call eval_on_groups #1148
• the trait FromIterator<&&str> is not implemented for polars::prelude::Series #1147
• Shift in aggregation context #1141
• Get/set categorical levels? #1115
• Is it possible to create a dataframe from row-like (Vec<Struct>) data? #1111
• PanicException when filter on sorted columns #1110
• Pickle support for Polars dataframes #1109
• Rosetta stone for groupby between pandas and polars? #1083

Full Changelog

Closed issues:

• csv schema inference scientific float notation #1134
• python: csv read_cvs dtypes accept list #1133

Full Changelog

Full Changelog

Closed issues:

• Extremely slow pivot.count (compared with pivot.sum, pivot.max, pivot.first, ...) #1129
• Only first gzip stream of gzipped CSV/TSV files with multiple gzip streams is read. #1126
• implement __copy__, __deepcopy__ #1120
• pretty print failure output of frame_equal assertions #1112
• PanicException when converting non-Utf8 column to Catergorical. #1107
• read_csv of a compressed file fails when selecting a subset of columns #1026
• csv-parser: remove dependency on csv crate. #956

Full Changelog

Full Changelog

Closed issues:

• numpy boolean vectors get converted as "objects" instead of cast to bool #1105
• PanicException: assertion failed: i < (self.bits.len() << 3) #1098

• feature

  • select columns by regex
  • support >1M columns in IPC reader
  • make DataFrame.sort arguments equal to LazyFrame.sort
  • pl.all() == pl.col("*")

• bug fix

  • fix bugs due to filtering in aggregations: #1101
  • fix bug in wildcard in functions 3163ee5

• feature
  • keep_name expr
  • exclude expr
  • drop_nulls expre
  • explode accepts expression (thus wildcard)
  • groupby: head + tail added
  • df[::2] slicing added

patch release to fix panic #1077

• feature
  • extract jsonpath
  • more object support
• performance
  • improve list take performance
• bug fix
  • don't panic in out of bounds take, but error
  • update offsets in case of utf8lossy
  • fix bug in pyarrow round trip with list types

• feature
  • concat_str function
  • more object support
  • hash and row_hash function/ expr
  • reinterpret function/ expr
  • Series.mode expr/function
  • csv file decompression
  • read_sql support
• performance
  • divide and conquer binary expressions

• feature
  • cross join added
  • dot-product
• performance
  • improve csv-parser performance by ~25%
• bug fix
  • various minor

• feature
  • cross join added
  • dot-product
• performance
  • improve csv-parser performance by ~25%
• bug fix
  • various minor

• feature
  • is_first expr/method
  • asof join added
  • eager io can open multiple sources with ffspec
  • resolve ~ to homedir
  • python arange add step and run eager
• performance
  • use fast csv-parser for more python memory buffers/streams
• bug fix
  • kleene or and and operations
  • maybe fix rayon deadlock
  • concat is a pure function
  • string addition lhs broadcast

• feature
  • correct type hints for python 3.6
  • csv-parser option to ignore comment lines
• performance
  • improve take on DataFrame
  • remove bound checks in buffer creation
  • improve performance of sorting by multiple columns
  • improve argsort performance
• bug fix
  • fix backward/forward fill
  • window groupby context
  • fix is_duplicated dispatch

• bug fix
  • fix UB due to slice in take kernel
  • fix join for dates

• feature
  • from_pandas accept series and date range #875
  • expr: forward_fill, backward_fill #874
  • gzipped file support in csv parser
• performance
  • reduce memory usage of multi-key groupby
  • improve variance and std-dev aggregation
• bug fix
  • cast to large-utf8 before collecting chunks #870
  • various

• performance
  • improve hashing performance for grouping on two keys for 64 bit and 32 and 64 bit data.
  • improve cache coherence take operation of multiple chunks
• bug fix
  • fix replaxing string with None #802

• feature
  • improve compatibility with pyarrow csv parser
• performance
  • improve hashing performance for grouping on two keys for 64 bit and 32 and 64 bit data.
  • improve cache coherence take operation of multiple chunks
  • fast path for categorical unique
  • decrease memory fragmentation and usage of csv-parser
• bug fix
  • split utf8 data only at valid char boundaries #789
  • fix bug in outer join due to new partitioning algorithm

• feature
  • Series.round
  • head/ limit aliases
• performance
  • partitioned hashing

• breaking change

  • str namespace Series.str_* methods to Series.str.
  • dt namespace Series datetime related methods to Series.dt.

• feature

  • DataFrame.rows method
  • apply on object types
  • Series.dt.to_python_datetime
  • Series.dt.timestamp

• bug fix

  • preserve date64 in round trip to parquet #723
  • during arrow conversion coerce categorical to utf8 (this preserves string data) #725
  • fix bug in csv skip rows

• performance

  • improve hashing of string data in groupby and join
  • improve numeric hashing in join
  • fast path for filtering no data and all date (upstream)

• feature

  • window function by multiple group columns

• bug fix

  • fix bug in argsort multiple
  • fix bug in filter with nulls (upstream)

• performance

  • improve numeric hashing in groupby
  • fast paths for filters (upstream)

• feature
  • argsort multiple columns

• feature

  • support more indexing
  • scan_csv low memory argument
  • Series.filter accept list of expressions
  • object type:
    • zip
    • take -> join / groupby agg
    • agg first/ last

• performance

  • change memory usage of csv-parser
  • binary aggregation in parallel
  • determine groupby keys in threadpool

• feature

  • Series literal may have any length
  • change globaly string cache behavior
  • Add Expr.arg_sort
  • Make literals typed

• bug fix

  • Fix Expr.fill_null
  • set offset in null buffers (fixes aggregation with null values)

• performance

  • sample cardinality in groupby and choose algorithm

• feature

  • join allows expression syntax
  • use pyarrow as default ipc backend

• bug fix

  • fix deadlock in window expressions

• bug fix

  • fix bug in cumsum #604

• feature

  • DataFrame.describe method #606
  • Multi-level sorting of a DataFrame #607
  • Expand functionality of Expr.is_in #614
  • Csv-parser low_memory option #615
  • Allow expressions in pl.arange #611

• performance

  • sort().reverse() optimization #605

• bug fix
  • null handling in mean, std, var, and cov aggregations. #595
  • rev-mapping of categorical stored duplicates. #595
  • fix memory surge after csv-parsing #593

• bug fix

  • Throw error on join from different string cache #584
  • fix covariance of array with null values #585

• feature

  • Series describe method #569
  • dsl: take, arg_unique, unique
  • allow lazy expressions in Eager API # 588
  • describe Series

• performance

  • fix accidental expensive appends #592
  • remove chunk_id from ChunkedArray #593

• bug fix

  • ensure column name persist after pyarrow cast #563
  • make sure that agg_list maintains dtype #567
  • fix panic in physical dispatch of Date dtypes

• feature

  • Implicitly Cast dtypes to temporal types in csv parser #560
  • Series describe method #569

• performance

  • Cache and improve window functions performance #570

• bug fix

  • fix bug with pyarrow chunkedarray: #545

• feature

  • DataFrame.apply method
  • Make a Series a Literal
  • Make None a Literal

• performance

  • Update arrow
    • faster iterators
    • faster kernels

• bug fix

  • fix bug in downsample: #537

• feature

  • cast categorical in csv parser: #533
  • add many groupby-context aware operations: #534
  • dowcast by month: #537

• performance

  • improve iterator in no null case: #538
  • remove indirection: #536

• bug fix

  • fix bug in vectorized hashing algorithm that affected groupbys with null values: #523
  • fix bug in downsample: 528
  • change median algorithm: #527

• feature

  • use lazy groupby API/DSL in eager API: #522
  • make sort groupby-context aware: #522

• performance

  • improve sort algorithms for sort and argsort: #526

• performance

  • [python | rust] multi-threaded outer join
  • [python | rust] better performance in groupby on multiple keys (faster hashmap comparisons)
  • [python | rust] better performance in multi column joins

• bug fix

  • [python] make horizontal aggregations null aware

• feature

  • [python | rust] Downsample by week
  • [python | rust] join by unlimited columns
  • [python] Create a list Series directly.
  • [python] Create DataFrame from np.ndarray

• bug fix

  • [python] pandas to polars date64, maintain time information
  • [python] fix bug in Date64 Series.year
  • [python] fix bug Series.mean (did not correct for null values) #484
  • [python | rust ] fix bug in rolling windows #484
  • [python | rust ] fix bug lazy csv parser #459

• feature

  • [python | rust] Series methods
    • Series.week
    • Series.weekday
    • Series.arg_min
    • Series.arg_max
    • Series.shape

• bug fix

  • [python] More pyarrow -> polars conversions.

• feature

  • [python] DataFrame methods: [ shift_and_fill].
  • [python] eager: sum, min, max, mean horizontal aggregation.

• performance
  • [python | rust] arrow arrays have a layer of indirection less; 10/20% performance improvement

• name change: Python bindings module renamed from pypolars to polars

• name change: Python bindings package renamed from py-polars to polars

• feature

  • [python] lazy: DataFrame methods: [ tail, first, last ].
  • [python] eager: DataFrame fold for horizontal aggregation.
  • [python] eager: Series methods: [median, quantile, is_in, to_frame]
  • [python] eager: iterate over groupby and yield groups' DataFrames
  • [python] eager: groupby.get_group('value')
  • [python] add parquet compression
  • [python] shift_and_fill expression
  • [python] implicitly download raw files from the web in read_parquet, read_csv.
  • [python | rust] methods for local peak finding in numerical series
  • [python | rust] faster query optimization due to local memory arena's.
  • [rust] reduce default compile time by making less features default.
  • [python | rust] Series zip_with implicitly cast to supertype.
  • [python | rust] window functions have a min_periods argument to control when to compute a result

• bug fix

  • [python] support file buffers for reading and writing csv and parquet
  • [python | rust] fix csv-parser: allow new-line character in a string field
  • [python | rust] don't let predicate-pushdown pass shift | sort operation to maintain correctness.

• performance
  • [python | rust] use mimalloc global allocator
  • [python | rust] undo performance regression on large number of threads
• bug fix
  • [python | rust] fix accidental over-allocation in csv-parser
  • [python] support agg (dictionary aggregation) for downsample

• performance
  • [python | rust] categorical type groupby keys (use size hint)
  • [python | rust] remove indirection layer in vector hasher
  • [python | rust] improve performance of null array creation
• bug fix
  • [python] implement set_with_mask for Boolean type
  • [python | rust] don't panic (instead return null) in dataframe aggregation std and var
• other
  • [rust] internal refactors

• bug fix
  • [python] fix various pyarrow related bugs

• feature
  • [python] render html tables
• performance
  • [python] default to pyarrow for parquet reading
  • [python | rust] use u32 instead of usize in groupby and join to increase cache coherence and reduce memory pressure.