- **Breaking**: Remove lots the dataframe-library specific functions and instead work with arrow arrays and record-batches directly using the [Arrow PyCapsule interface](https://arrow.apache.org/docs/format/CDataInterface/PyCapsuleInterface.html>). The core and polars parts of this library are no longer depending on `pyarrow` - instead the [lightweight 'arro3' library](https://github.com/kylebarron/arro3) is used. The pandas-parts still require `pyarrow` which needs the be installed manually. - Support for numpy 2. - Upgrade to h3o 0.7. - Release the GIL more often to allow other threads to run. - The minimum supported python version is now 3.9.
0.21.1
-------------------
- Fixed coordinate messup when converting H3 entities to points (issue 58).
0.21.0
-------------------
- Restrict numpy version to <2 until incompatibilities of rust-numpy with version 2 are resolved - Add `cells_to_localij` and `localij_to_cells` functions. - Polars v1.0.0 compatibility.
0.20.1
-------------------
- Upgrade h3o, rasterh3 and h3arrow dependencies. This includes a fix for converting datasets spanning the antimeridian by splitting and normalizing tiles before generating cells from them.
0.20.0
-------------------
- Migrate from the unmaintained arrow2 arrow implementation to the official arrow-rs from apache. This comes along with a few changes: - `directededges_to_wkb_lines` and `directededges_to_lines` have been removed. Use the linestring-versions instead. - Geometry collections are currently unsupported when working with WKB. This is still work-in-progress within geoarrow-rs. - Add support for the h3o containment mode "covers" - Finally removed the "all_intersecting" parameter in geometry to cells conversion as announced in v0.18
0.19.2
-------------------
- Fix bug which required geopandas geometry columns to be named "geometry" in `geodataframe_to_cells`. - Warn about possible memory exhaustion when encountering a `ArrowIndexError` in `explode_table_include_null` / `geodataframe_to_cells`.