Vaex

* Fix
* Join issue with missing values or nans [2077](https://github.com/vaexio/vaex/issues/2077)
* Feature
* vaex.agg.describe aggregator, and descripe on groupby [2004](https://github.com/vaexio/vaex/pull/2004)
* df.struct.flatten to turn arrow structs into multiple columns [2072](https://github.com/vaexio/vaex/pull/2072)
* Reuse hashmap for isin (df fingerprint is stable, better for cache) [2089](https://github.com/vaexio/vaex/pull/2089)
* Performance
* Value_counts uses a task to get caching support [2085](https://github.com/vaexio/vaex/pull/2085).

* Fix
* Write to cached filesystem when metadata argument is needed [1993](https://github.com/vaexio/vaex/pull/1993)
* Multi-d sparse groupby would fail for arrow data (e.g. list agg) [2031](https://github.com/vaexio/vaex/pull/2031)
* Exporting arrow with large_string would result in schema conflict [2030](https://github.com/vaexio/vaex/pull/2030)
* expression engine did not roundtrip dict correctly, missing ", " [2039](https://github.com/vaexio/vaex/pull/2039)
* Changed deprecated numpy.float to numpy.float64 [2023](https://github.com/vaexio/vaex/pull/2023)
* Replace pylab with pyplot [2047](https://github.com/vaexio/vaex/pull/2047)
* isin should accept empty array or non-existing values [2064](https://github.com/vaexio/vaex/pull/2064)
* Ordinal_encode with values which extra entries gave wrong results [2059](https://github.com/vaexio/vaex/pull/2059)
* Combining filters with arrow arrays failed converting (gave TypeError) [2038](https://github.com/vaexio/vaex/issues/2038)
* Wrong order of casting and subtracting offset cause overflow [2065](https://github.com/vaexio/vaex/pull/2065)

* Fix
* When aggregation leads to arrow data, non-dense binners failed (e.g. vaex.agg.list) [2017](https://github.com/vaexio/vaex/pull/2017)
* Filtering by boolean column would miss the column as dependency [2016](https://github.com/vaexio/vaex/pull/2016)

* Features
* Progress bar for percentile_approx and median_approx [1889](https://github.com/vaexio/vaex/pull/1889)
* Better casting of strings to datetime [1920](https://github.com/vaexio/vaex/pull/1920)
* We better support numpy scalars now, and more arrow time units. [1921](https://github.com/vaexio/vaex/pull/1921)
* Allow sorting by strings, multiple columns and multiple directions [1963](https://github.com/vaexio/vaex/pull/1963)
* Support JSON in df.export [1974](https://github.com/vaexio/vaex/pull/1974)
* New/better aggregators
* first/last use different type 'sort column. [1848](https://github.com/vaexio/vaex/pull/1848)
* Skew and kurtosis [1946](https://github.com/vaexio/vaex/pull/1946)
* List aggregator [1987](https://github.com/vaexio/vaex/pull/1987)
* Pre-sort by the grouping columns in df.groupby (better performance) [1990](https://github.com/vaexio/vaex/pull/1990)
* Performance
* No copy of hashmap and GIL release [1893](https://github.com/vaexio/vaex/pull/1893) [#1961](https://github.com/vaexio/vaex/pull/1961)
* Store strings in hashmap in arrow array, making map.key_array() faster [1976](https://github.com/vaexio/vaex/pull/1976)
* Fix
* Respect row_limit when the groupby is dense [1894](https://github.com/vaexio/vaex/pull/1894)
* Fingerprint collision possible if filter uses virtual column [1949](https://github.com/vaexio/vaex/pull/1949)
* Apply with filtered data could give wrong dtypes [1936](https://github.com/vaexio/vaex/pull/1936)
* Strings array growing failed when first string was zero length [1956](https://github.com/vaexio/vaex/pull/1956)
* Use less processes for when using multiprocessing. [1979](https://github.com/vaexio/vaex/pull/1979)
* Support chunked arrays and empty chunks in value counts. [1958](https://github.com/vaexio/vaex/pull/1958) [#1975](https://github.com/vaexio/vaex/pull/1975)
* Allow renaming of function, to make join use with functions without name collisions. [1966](https://github.com/vaexio/vaex/pull/1966)
* Join would fail if the rhs had no columns besides the join one [2010](https://github.com/vaexio/vaex/pull/2010)
* hdf5 export fails for concat df with missing columns [1493](https://github.com/vaexio/vaex/pull/1493)
* Allow `col` as column name [1992](https://github.com/vaexio/vaex/issues/1992)

* Features
* Multiple example datasets provided in `vaex.datasets` [1317](https://github.com/vaexio/vaex/pull/1317)
* We do not use asyncio for the default sync execute path [1783](https://github.com/vaexio/vaex/pull/1783)
* Executor works with asyncio with multiple tasks [1784]https://github.com/vaexio/vaex/pull/1784)
* Auto execute context manager makes vaex behave normal with await [1785](https://github.com/vaexio/vaex/pull/1785)
* Support exporting arrow and parquet to file like objects [1790](https://github.com/vaexio/vaex/pull/1790)
* Put lock files in $VAEX_HOME/lock [1797](https://github.com/vaexio/vaex/pull/1797)
* Show progress when converting the included datasets [1798](https://github.com/vaexio/vaex/pull/1798)
* Limit and limit_raise for unique and nunique [1801](https://github.com/vaexio/vaex/pull/1801)
* Lazy ordinal encode [1813](https://github.com/vaexio/vaex/pull/1813)
* Configure logging using settings system[1811](https://github.com/vaexio/vaex/pull/1811)
* Export to JSON [1789](https://github.com/vaexio/vaex/pull/1789)
* Progress bar can be configured using settings system [1815](https://github.com/vaexio/vaex/pull/1815)
* fillna and fillmissing should upcast integers when needed [1869](https://github.com/vaexio/vaex/pull/1869)
* Performance
* Moved mutex use to the C++ layer to avoid GIL issues [1847](https://github.com/vaexio/vaex/pull/1847)
* Many refactors to improve performance [1863](https://github.com/vaexio/vaex/pull/1863) [#1869](https://github.com/vaexio/vaex/pull/1869)
* Fix
* Support empty parquet and arrow files [1791](https://github.com/vaexio/vaex/pull/1791)
* Keep virtual column order when renaming/dropping to not break state transfer [1788](https://github.com/vaexio/vaex/pull/1788)
* blake3 compatibility issues [1818](https://github.com/vaexio/vaex/pull/1818) [db527a6](https://github.com/vaexio/vaex/commit/db527a6942db6ee74d97f1f1e8e5ddb3e8978f0c)
* Avoid frozendict 2.2.0 which can segfault on Python 3.6[1856](https://github.com/vaexio/vaex/pull/1856)
* Use label instead of expression for non-ident column names in binby [1842](https://github.com/vaexio/vaex/pull/1842)
* Development
* Use cmake/scikit-build [1847](https://github.com/vaexio/vaex/pull/1847) [92af1b1](https://github.com/vaexio/vaex/commit/92af1b1fab55dcc36c93e327495ac239c3fef772) [ad88d4b](https://github.com/vaexio/vaex/commit/ad88d4b2525c9fda7798c685985d9391a6b498a5)