Xarray 0.10 release

New features, performance improvements, and a few bug fixes

Posted by Joe Hamman on October 31, 2017

The first release candidate of Xarray v0.10 was released today. The release includes more than five months of work and contributions from more than 28 contributors, many of whom are part of the Pangeo community.

The release includes numerous new features, performance improvements, and bug fixes. Full details of the release can be found in the Xarray documentation.

Release Highlights:

  • Indexing now supports broadcasting over dimensions, similar to NumPy’s vectorized indexing (but better!).
  • resample() has a new groupby-like API like pandas. Of particular note, Xarray’s resample tools now also support up- and down-sampling.
  • apply_ufunc() facilitates wrapping and parallelizing functions written for NumPy arrays.
  • Performance improvements, particularly for dask and open_mfdataset().

We’re excited for the Pangeo community to test out this release candidate and encourage users to report any issues on GitHub. If all goes well, the final v0.10.0 release will come out in about a week.

To install, use:

pip install --pre --upgrade --upgrade-strategy=only-if-needed xarray

Part of Pangeo’s Roadmap

This is the first release of Xarray since the NSF funded Pangeo project began. While the development work that contributed to this release was by no means done entirely by Pangeo members, it does address two of the key Xarray development objectives described in the Pangeo project description: 1) improvements of distributed computing and efficiency with dask, and 2) enhancement of current Xarray tools for applying arbitrary functions on Xarray/dask objects.

Update 2017-11-13: The second release candidate was released today and is now available on PyPi. This release included a number of minor bug fixes to new features in v0.10.

Update 2017-11-27: The final release was published last week and is now available on PyPi and conda.