* Use variable and dimension names as-is, instead of casting to strings. [xarray](https://docs.xarray.dev/en/stable/index.html#module-xarray) internally treats all variable and dimension names as Hashable, without assuming they are strings. NetCDF4 files always use string names, so this change should not affect you if you only use NetCDF4 datasets ([pull request 25](https://github.com/csiro-coasts/emsarray/pull/25)).
* Fix error when UGRID start_index is the string “0” or “1”. The conventions imply this should be an integer type, however real datasets use a string value here so a tolerant implementation is useful ([pull request 26](https://github.com/csiro-coasts/emsarray/pull/26), [pull request csiro-coasts/emsarray-data#1](https://github.com/csiro-coasts/emsarray-data/pull/1)).
* Split emsarray.operations in to separate modules ([pull request 27](https://github.com/csiro-coasts/emsarray/pull/27)).
* Add [emsarray.operations.geometry](https://emsarray.readthedocs.io/en/v0.3.0/api/operations/geometry.html) module which can export dataset geometry to GeoJSON and Shapefiles ([pull request #28](https://github.com/csiro-coasts/emsarray/pull/28)).
* Add [Format.drop_geometry()](https://emsarray.readthedocs.io/en/v0.3.0/api/formats/interface.html#emsarray.formats.Format.drop_geometry) method ([pull request 29](https://github.com/csiro-coasts/emsarray/pull/29)).
* Add [emsarray.operations.point_extraction](https://emsarray.readthedocs.io/en/v0.3.0/api/operations/point_extraction.html) module and [emsarray extract-points](https://emsarray.readthedocs.io/en/v0.3.0/cli.html#emsarray-extract-points) command line entry point which can extract point data from a dataset ([pull request 29](https://github.com/csiro-coasts/emsarray/pull/29)).