Vaex

* Fixes
* allow single columns or expressions in materialize [2249](https://github.com/vaexio/vaex/pull/2249)
* arrow data used in selection would ignore null values or fail [2196](https://github.com/vaexio/vaex/pull/2196)
* build expressions / filters with an arrow string scalar [2244](https://github.com/vaexio/vaex/pull/2244)
* selection-dropna did not work with non-identifier expresssion [2208](https://github.com/vaexio/vaex/pull/2208)
* use `vaex.settings` for thread counts [2231](https://github.com/vaexio/vaex/pull/2231)
* raise an informative exception when `extract` can not run [2232](https://github.com/vaexio/vaex/pull/2232)

* Performance
* Import and version checking improvements [2226](https://github.com/vaexio/vaex/pull/2226)

* Features
* Add support for the arrow export CSV backend in `export_many` [2220](https://github.com/vaexio/vaex/pull/2220)
* Auto detect type when Lazy CSV reading [2224](https://github.com/vaexio/vaex/pull/2224)

* Fixes
* convert options for lazy `from_csv_arrow` [2218](https://github.com/vaexio/vaex/pull/2218)
* support arrow datasets containing multiple fragments without row groups [2221](https://github.com/vaexio/vaex/pull/2221)
* Interchange protocol fixes and updates [2150](https://github.com/vaexio/vaex/pull/2150)

* Features
* Out of core (lazy) and normal CSV support using Apache Arrow CSV reader [1028](https://github.com/vaexio/vaex/pull/1028)
* Support compressed csv reading for pandas and arrow [1028](https://github.com/vaexio/vaex/pull/1028)
* Expose vaex.DataFrame and vaex.Expression for typing [2186](https://github.com/vaexio/vaex/pull/2186)

* Fixes
* Negative indexing of a filtered dataframe [2163](https://github.com/vaexio/vaex/pull/2163)
* nunique aggregator for numeric with selection gave wrong results [2199](https://github.com/vaexio/vaex/pull/2199)
* vaex.from_arrow_dataset had no working implementation [2195](https://github.com/vaexio/vaex/pull/2195)

* Features
* get_column_names accepts a dtypes argument [2160](https://github.com/vaexio/vaex/pull/2160)
* Fixes
* df.extract() was not thread safe [2182](https://github.com/vaexio/vaex/pull/2182)
* uuid4 function was not always restored properly [2181](https://github.com/vaexio/vaex/pull/2181)
* groupby could overflow due to wrong downcasting [2137](https://github.com/vaexio/vaex/pull/2137)
* support unique with selection=True [2164](https://github.com/vaexio/vaex/pull/2164)
* value_counts for strings was sometimes off [2147](https://github.com/vaexio/vaex/pull/2147)
* better arrow support for interchanging categorical columns [2135](https://github.com/vaexio/vaex/pull/2135)

* Fix
* Slicing of empty dataframe [2129](https://github.com/vaexio/vaex/issues/2129)
* Dataframe interchange protocol: Better support for numpy backed categoricals [2133](https://github.com/vaexio/vaex/issues/2133)
* Dataframe interchange protocol: Column.null_count returns NA count to comply with spec [2132](https://github.com/vaexio/vaex/issues/2132)
* Export csv with 'protected' column name [2117](https://github.com/vaexio/vaex/issues/2117)

* Fix
* df.func.where relies more on pyarrow 5's if_else [2096](https://github.com/vaexio/vaex/issues/2096)
* correct $VAEX_PATH_HOME -> $VAEX_PATH [2101](https://github.com/vaexio/vaex/issues/2101)
* Various join fixed when the missing values were present [808](https://github.com/vaexio/vaex/issues/808)
* Various join fixed when the missing values were present [808](https://github.com/vaexio/vaex/issues/808)
* string join on large_list with large_strings. [2112](https://github.com/vaexio/vaex/issues/2112)
* Working arm wheel for osx () [2124](https://github.com/vaexio/vaex/issues/2124)
* Performance
* Do not let arrow validate the dict encoded data. [d6242090a1f480abae669bc5281e803fe06c5d36](https://github.com/vaexio/vaex/commit/d6242090a1f480abae669bc5281e803fe06c5d36)
* Feature
* Add how to dropna, dropinf etc [2104](https://github.com/vaexio/vaex/issues/2104)