What’s New

Contents

What’s New#

v2024.09.0 (Sept 11, 2024)#

This release drops support for Python 3.9, and adds support for grouping by multiple arrays, while providing numerous performance improvements and bug fixes.

Thanks to the 33 contributors to this release: Alfonso Ladino, Andrew Scherer, Anurag Nayak, David Hoese, Deepak Cherian, Diogo Teles Sant’Anna, Dom, Elliott Sales de Andrade, Eni, Holly Mandel, Illviljan, Jack Kelly, Julius Busecke, Justus Magin, Kai Mühlbauer, Manish Kumar Gupta, Matt Savoie, Maximilian Roos, Michele Claus, Miguel Jimenez, Niclas Rieger, Pascal Bourgault, Philip Chmielowiec, Spencer Clark, Stephan Hoyer, Tao Xin, Tiago Sanona, TimothyCera-NOAA, Tom Nicholas, Tom White, Virgile Andreani, oliverhiggs and tiago

New Features#

Performance#

Breaking changes#

  • Support for python 3.9 has been dropped (PR8937)

  • The minimum versions of some dependencies were changed

    Package

    Old

    New

    boto3

    1.26

    1.28

    cartopy

    0.21

    0.22

    dask-core

    2023.4

    2023.9

    distributed

    2023.4

    2023.9

    h5netcdf

    1.1

    1.2

    iris

    3.4

    3.7

    numba

    0.56

    0.57

    numpy

    1.23

    1.24

    pandas

    2.0

    2.1

    scipy

    1.10

    1.11

    typing_extensions

    4.5

    4.7

    zarr

    2.14

    2.16

Bug fixes#

  • Fix bug with rechunking to a frequency when some periods contain no data (GH9360). By Deepak Cherian.

  • Fix bug causing DataTree.from_dict to be sensitive to insertion order (GH9276, PR9292). By Tom Nicholas.

  • Fix resampling error with monthly, quarterly, or yearly frequencies with cftime when the time bins straddle the date “0001-01-01”. For example, this can happen in certain circumstances when the time coordinate contains the date “0001-01-01”. (GH9108, PR9116) By Spencer Clark and Deepak Cherian.

  • Fix issue with passing parameters to ZarrStore.open_store when opening datatree in zarr format (GH9376, PR9377). By Alfonso Ladino

  • Fix deprecation warning that was raised when calling np.array on an xr.DataArray in NumPy 2.0 (GH9312, PR9393) By Andrew Scherer.

  • Fix support for using pandas.DateOffset, pandas.Timedelta, and datetime.timedelta objects as resample frequencies (GH9408, PR9413). By Oliver Higgs.

Internal Changes#

v2024.07.0 (Jul 30, 2024)#

This release extends the API for groupby operations with various grouper objects <groupby.groupers_>, and includes improvements to the documentation and numerous bugfixes.

Thanks to the 22 contributors to this release: Alfonso Ladino, ChrisCleaner, David Hoese, Deepak Cherian, Dieter Werthmüller, Illviljan, Jessica Scheick, Joel Jaeschke, Justus Magin, K. Arthur Endsley, Kai Mühlbauer, Mark Harfouche, Martin Raspaud, Mathijs Verhaegh, Maximilian Roos, Michael Niklas, Michał Górny, Moritz Schreiber, Pontus Lurcock, Spencer Clark, Stephan Hoyer and Tom Nicholas

New Features#

Breaking changes#

  • The base and loffset parameters to Dataset.resample() and DataArray.resample() are now removed. These parameters have been deprecated since v2023.03.0. Using the origin or offset parameters is recommended as a replacement for using the base parameter and using time offset arithmetic is recommended as a replacement for using the loffset parameter. (PR9233) By Deepak Cherian.

  • The squeeze kwarg to groupby is now ignored. This has been the source of some quite confusing behaviour and has been deprecated since v2024.01.0. groupby` behavior is now always consistent with the existing .groupby(..., squeeze=False) behavior. No errors will be raised if squeeze=False. (PR9280) By Deepak Cherian.

Bug fixes#

Documentation#

Internal Changes#

v2024.06.0 (Jun 13, 2024)#

This release brings various performance optimizations and compatibility with the upcoming numpy 2.0 release.

Thanks to the 22 contributors to this release: Alfonso Ladino, David Hoese, Deepak Cherian, Eni Awowale, Ilan Gold, Jessica Scheick, Joe Hamman, Justus Magin, Kai Mühlbauer, Mark Harfouche, Mathias Hauser, Matt Savoie, Maximilian Roos, Mike Thramann, Nicolas Karasiak, Owen Littlejohns, Paul Ockenfuß, Philippe THOMY, Scott Henderson, Spencer Clark, Stephan Hoyer and Tom Nicholas

Performance#

Bug fixes#

  • Preserve conversion of timezone-aware pandas Datetime arrays to numpy object arrays (GH9026, PR9042). By Ilan Gold.

  • DataArrayResample.interpolate() and DatasetResample.interpolate() method now support arbitrary kwargs such as order for polynomial interpolation (GH8762). By Nicolas Karasiak.

Documentation#

  • Add link to CF Conventions on packed data and sentence on type determination in the I/O user guide (GH9041, PR9045). By Kai Mühlbauer.

Internal Changes#

v2024.05.0 (May 12, 2024)#

This release brings support for pandas ExtensionArray objects, optimizations when reading Zarr, the ability to concatenate datasets without pandas indexes, more compatibility fixes for the upcoming numpy 2.0, and the migration of most of the xarray-datatree project code into xarray main!

Thanks to the 18 contributors to this release: Aimilios Tsouvelekakis, Andrey Akinshin, Deepak Cherian, Eni Awowale, Ilan Gold, Illviljan, Justus Magin, Mark Harfouche, Matt Savoie, Maximilian Roos, Noah C. Benson, Pascal Bourgault, Ray Bell, Spencer Clark, Tom Nicholas, ignamv, owenlittlejohns, and saschahofmann.

New Features#

Breaking changes#

  • The PyNIO backend has been deleted (GH4491, PR7301). By Deepak Cherian.

  • The minimum versions of some dependencies were changed, in particular our minimum supported pandas version is now Pandas 2.

    Package

    Old

    New

    dask-core

    2022.12

    2023.4

    distributed

    2022.12

    2023.4

    h5py

    3.7

    3.8

    matplotlib-base

    3.6

    3.7

    packaging

    22.0

    23.1

    pandas

    1.5

    2.0

    pydap

    3.3

    3.4

    sparse

    0.13

    0.14

    typing_extensions

    4.4

    4.5

    zarr

    2.13

    2.14

Bug fixes#

Internal Changes#

v2024.03.0 (Mar 29, 2024)#

This release brings performance improvements for grouped and resampled quantile calculations, CF decoding improvements, minor optimizations to distributed Zarr writes, and compatibility fixes for Numpy 2.0 and Pandas 3.0.

Thanks to the 18 contributors to this release: Anderson Banihirwe, Christoph Hasse, Deepak Cherian, Etienne Schalk, Justus Magin, Kai Mühlbauer, Kevin Schwarzwald, Mark Harfouche, Martin, Matt Savoie, Maximilian Roos, Ray Bell, Roberto Chang, Spencer Clark, Tom Nicholas, crusaderky, owenlittlejohns, saschahofmann

New Features#

Breaking changes#

Bug fixes#

Internal Changes#

v2024.02.0 (Feb 19, 2024)#

This release brings size information to the text repr, changes to the accepted frequency strings, and various bug fixes.

Thanks to our 12 contributors:

Anderson Banihirwe, Deepak Cherian, Eivind Jahren, Etienne Schalk, Justus Magin, Marco Wolsza, Mathias Hauser, Matt Savoie, Maximilian Roos, Rambaud Pierrick, Tom Nicholas

New Features#

Breaking changes#

Deprecations#

  • The dt.weekday_name parameter wasn’t functional on modern pandas versions and has been removed. (GH8610, PR8664) By Sam Coleman.

Bug fixes#

Documentation#

  • Fix variables arg typo in Dataset.sortby() docstring (GH8663, PR8670) By Tom Vo.

  • Fixed documentation where the use of the depreciated pandas frequency string prevented the documentation from being built. (PR8638) By Sam Coleman.

Internal Changes#

v2024.01.1 (23 Jan, 2024)#

This release is to fix a bug with the rendering of the documentation, but it also includes changes to the handling of pandas frequency strings.

Breaking changes#

  • Following pandas, infer_freq() will return "YE", instead of "Y" (formerly "A"). This is to be consistent with the deprecation of the latter frequency string in pandas 2.2. This is a follow up to PR8415 (GH8612, PR8642). By Mathias Hauser.

Deprecations#

Documentation#

  • Pin sphinx-book-theme to 1.0.1 to fix a rendering issue with the sidebar in the docs. (GH8619, PR8632) By Tom Nicholas.

v2024.01.0 (17 Jan, 2024)#

This release brings support for weights in correlation and covariance functions, a new DataArray.cumulative aggregation, improvements to xr.map_blocks, an update to our minimum dependencies, and various bugfixes.

Thanks to our 17 contributors to this release:

Abel Aoun, Deepak Cherian, Illviljan, Johan Mathe, Justus Magin, Kai Mühlbauer, Llorenç Lledó, Mark Harfouche, Markel, Mathias Hauser, Maximilian Roos, Michael Niklas, Niclas Rieger, Sébastien Celles, Tom Nicholas, Trinh Quoc Anh, and crusaderky.

New Features#

  • xr.cov() and xr.corr() now support using weights (GH8527, PR7392). By Llorenç Lledó.

  • Accept the compression arguments new in netCDF 1.6.0 in the netCDF4 backend. See netCDF4 documentation for details. Note that some new compression filters needs plugins to be installed which may not be available in all netCDF distributions. By Markel García-Díez. (GH6929, PR7551)

  • Add DataArray.cumulative() & Dataset.cumulative() to compute cumulative aggregations, such as sum, along a dimension — for example da.cumulative('time').sum(). This is similar to pandas’ .expanding, and mostly equivalent to .cumsum methods, or to DataArray.rolling() with a window length equal to the dimension size. By Maximilian Roos. (PR8512)

  • Decode/Encode netCDF4 enums and store the enum definition in dataarrays’ dtype metadata. If multiple variables share the same enum in netCDF4, each dataarray will have its own enum definition in their respective dtype metadata. By Abel Aoun. (GH8144, PR8147)

Breaking changes#

  • The minimum versions of some dependencies were changed (PR8586):

    Package

    Old

    New

    cartopy

    0.20

    0.21

    dask-core

    2022.7

    2022.12

    distributed

    2022.7

    2022.12

    flox

    0.5

    0.7

    iris

    3.2

    3.4

    matplotlib-base

    3.5

    3.6

    numpy

    1.22

    1.23

    numba

    0.55

    0.56

    packaging

    21.3

    22.0

    seaborn

    0.11

    0.12

    scipy

    1.8

    1.10

    typing_extensions

    4.3

    4.4

    zarr

    2.12

    2.13

Deprecations#

Bug fixes#

Internal Changes#

  • The implementation of map_blocks() has changed to minimize graph size and duplication of data. This should be a strict improvement even though the graphs are not always embarrassingly parallel any more. Please open an issue if you spot a regression. (PR8412, GH8409). By Deepak Cherian.

  • Remove null values before plotting. (PR8535). By Jimmy Westling.

  • Redirect cumulative reduction functions internally through the ChunkManagerEntryPoint, potentially allowing ffill() and bfill() to use non-dask chunked array types. (PR8019) By Tom Nicholas.

v2023.12.0 (2023 Dec 08)#

This release brings new hypothesis strategies for testing, significantly faster rolling aggregations as well as ffill and bfill with numbagg, a new Dataset.eval() method, and improvements to reading and writing Zarr arrays (including a new "a-" mode).

Thanks to our 16 contributors:

Anderson Banihirwe, Ben Mares, Carl Andersson, Deepak Cherian, Doug Latornell, Gregorio L. Trevisan, Illviljan, Jens Hedegaard Nielsen, Justus Magin, Mathias Hauser, Max Jones, Maximilian Roos, Michael Niklas, Patrick Hoefler, Ryan Abernathey, Tom Nicholas

New Features#

  • Added hypothesis strategies for generating xarray.Variable objects containing arbitrary data, useful for parametrizing downstream tests. Accessible under testing.strategies, and documented in a new page on testing in the User Guide. (GH6911, PR8404) By Tom Nicholas.

  • rolling() uses numbagg for most of its computations by default. Numbagg is up to 5x faster than bottleneck where parallelization is possible. Where parallelization isn’t possible — for example a 1D array — it’s about the same speed as bottleneck, and 2-5x faster than pandas’ default functions. (PR8493). numbagg is an optional dependency, so requires installing separately.

  • Use a concise format when plotting datetime arrays. (PR8449). By Jimmy Westling.

  • Avoid overwriting unchanged existing coordinate variables when appending with Dataset.to_zarr() by setting mode='a-'. By Ryan Abernathey and Deepak Cherian.

  • rank() now operates on dask-backed arrays, assuming the core dim has exactly one chunk. (PR8475). By Maximilian Roos.

  • Add a Dataset.eval() method, similar to the pandas’ method of the same name. (PR7163). This is currently marked as experimental and doesn’t yet support the numexpr engine.

  • Dataset.drop_vars() & DataArray.drop_vars() allow passing a callable, similar to Dataset.where() & Dataset.sortby() & others. (PR8511). By Maximilian Roos.

Breaking changes#

  • Explicitly warn when creating xarray objects with repeated dimension names. Such objects will also now raise when DataArray.get_axis_num() is called, which means many functions will raise. This latter change is technically a breaking change, but whilst allowed, this behaviour was never actually supported! (GH3731, PR8491) By Tom Nicholas.

Deprecations#

  • As part of an effort to standardize the API, we’re renaming the dims keyword arg to dim for the minority of functions which current use dims. This started with xarray.dot() & DataArray.dot() and we’ll gradually roll this out across all functions. The warnings are currently PendingDeprecationWarning, which are silenced by default. We’ll convert these to DeprecationWarning in a future release. By Maximilian Roos.

  • Raise a FutureWarning warning that the type of Dataset.dims() will be changed from a mapping of dimension names to lengths to a set of dimension names. This is to increase consistency with DataArray.dims(). To access a mapping of dimension names to lengths please use Dataset.sizes(). The same change also applies to DatasetGroupBy.dims. (GH8496, PR8500) By Tom Nicholas.

  • Dataset.drop() & DataArray.drop() are now deprecated, since pending deprecation for several years. DataArray.drop_sel() & DataArray.drop_var() replace them for labels & variables respectively. (PR8497) By Maximilian Roos.

Bug fixes#

Documentation#

  • Added illustration of updating the time coordinate values of a resampled dataset using time offset arithmetic. This is the recommended technique to replace the use of the deprecated loffset parameter in resample (PR8479). By Doug Latornell.

  • Improved error message when attempting to get a variable which doesn’t exist from a Dataset. (PR8474) By Maximilian Roos.

  • Fix default value of combine_attrs in xarray.combine_by_coords() (PR8471) By Gregorio L. Trevisan.

Internal Changes#

v2023.11.0 (Nov 16, 2023)#

Tip

This is our 10th year anniversary release! Thank you for your love and support.

This release brings the ability to use opt_einsum for xarray.dot() by default, support for auto-detecting region when writing partial datasets to Zarr, and the use of h5py drivers with h5netcdf.

Thanks to the 19 contributors to this release: Aman Bagrecha, Anderson Banihirwe, Ben Mares, Deepak Cherian, Dimitri Papadopoulos Orfanos, Ezequiel Cimadevilla Alvarez, Illviljan, Justus Magin, Katelyn FitzGerald, Kai Muehlbauer, Martin Durant, Maximilian Roos, Metamess, Sam Levang, Spencer Clark, Tom Nicholas, mgunyho, templiert

New Features#

Breaking changes#

  • drop support for cdms2. Please use xcdat instead (PR8441). By Justus Magin.

  • Following pandas, infer_freq() will return "Y", "YS", "QE", "ME", "h", "min", "s", "ms", "us", or "ns" instead of "A", "AS", "Q", "M", "H", "T", "S", "L", "U", or "N". This is to be consistent with the deprecation of the latter frequency strings (GH8394, PR8415). By Spencer Clark.

  • Bump minimum tested pint version to >=0.22. By Deepak Cherian.

  • Minimum supported versions for the following packages have changed: h5py >=3.7, h5netcdf>=1.1. By Kai Mühlbauer.

Deprecations#

Bug fixes#

Documentation#

v2023.10.1 (19 Oct, 2023)#

This release updates our minimum numpy version in pyproject.toml to 1.22, consistent with our documentation below.

v2023.10.0 (19 Oct, 2023)#

This release brings performance enhancements to reading Zarr datasets, the ability to use numbagg for reductions, an expansion in API for rolling_exp, fixes two regressions with datetime decoding, and many other bugfixes and improvements. Groupby reductions will also use numbagg if flox>=0.8.1 and numbagg are both installed.

Thanks to our 13 contributors: Anderson Banihirwe, Bart Schilperoort, Deepak Cherian, Illviljan, Kai Mühlbauer, Mathias Hauser, Maximilian Roos, Michael Niklas, Pieter Eendebak, Simon Høxbro Hansen, Spencer Clark, Tom White, olimcc

New Features#

Breaking changes#

Deprecations#

Bug fixes#

  • DataArray.rename() & Dataset.rename() would emit a warning when the operation was a no-op. (GH8266) By Simon Hansen.

  • Fixed a regression introduced in the previous release checking time-like units when encoding/decoding masked data (GH8269, PR8277). By Kai Mühlbauer.

  • Fix datetime encoding precision loss regression introduced in the previous release for datetimes encoded with units requiring floating point values, and a reference date not equal to the first value of the datetime array (GH8271, PR8272). By Spencer Clark.

  • Fix excess metadata requests when using a Zarr store. Prior to this, metadata was re-read every time data was retrieved from the array, now metadata is retrieved only once when they array is initialized. (GH8290, PR8297). By Oliver McCormack.

  • Fix to_zarr ending in a ReadOnlyError when consolidated metadata was used and the write_empty_chunks was provided. (GH8323, PR8326) By Matthijs Amesz.

Documentation#

Internal Changes#

v2023.09.0 (Sep 26, 2023)#

This release continues work on the new xarray.Coordinates object, allows to provide preferred_chunks when reading from netcdf files, enables xarray.apply_ufunc() to handle missing core dimensions and fixes several bugs.

Thanks to the 24 contributors to this release: Alexander Fischer, Amrest Chinkamol, Benoit Bovy, Darsh Ranjan, Deepak Cherian, Gianfranco Costamagna, Gregorio L. Trevisan, Illviljan, Joe Hamman, JR, Justus Magin, Kai Mühlbauer, Kian-Meng Ang, Kyle Sunden, Martin Raspaud, Mathias Hauser, Mattia Almansi, Maximilian Roos, András Gunyhó, Michael Niklas, Richard Kleijn, Riulinchen, Tom Nicholas and Wiktor Kraśnicki.

We welcome the following new contributors to Xarray!: Alexander Fischer, Amrest Chinkamol, Darsh Ranjan, Gianfranco Costamagna, Gregorio L. Trevisan, Kian-Meng Ang, Riulinchen and Wiktor Kraśnicki.

New Features#

Breaking changes#

  • The Coordinates constructor now creates a (pandas) index by default for each dimension coordinate. To keep the previous behavior (no index created), pass an empty dictionary to indexes. The constructor now also extracts and add the indexes from another Coordinates object passed via coords (PR8107). By Benoît Bovy.

  • Static typing of xlim and ylim arguments in plotting functions now must be tuple[float, float] to align with matplotlib requirements. (GH7802, PR8030). By Michael Niklas.

Deprecations#

Bug fixes#

Documentation#

Internal Changes#

  • Many error messages related to invalid dimensions or coordinates now always show the list of valid dims/coords (PR8079). By András Gunyhó.

  • Refactor of encoding and decoding times/timedeltas to preserve nanosecond resolution in arrays that contain missing values (PR7827). By Kai Mühlbauer.

  • Transition .rolling_exp functions to use .apply_ufunc internally rather than .reduce, as the start of a broader effort to move non-reducing functions away from `.reduce, (PR8114). By Maximilian Roos.

  • Test range of fill_value’s in test_interpolate_pd_compat (GH8146, PR8189). By Kai Mühlbauer.

v2023.08.0 (Aug 18, 2023)#

This release brings changes to minimum dependencies, allows reading of datasets where a dimension name is associated with a multidimensional variable (e.g. finite volume ocean model output), and introduces a new xarray.Coordinates object.

Thanks to the 16 contributors to this release: Anderson Banihirwe, Articoking, Benoit Bovy, Deepak Cherian, Harshitha, Ian Carroll, Joe Hamman, Justus Magin, Peter Hill, Rachel Wegener, Riley Kuttruff, Thomas Nicholas, Tom Nicholas, ilgast, quantsnus, vallirep

Announcements#

The xarray.Variable class is being refactored out to a new project title ‘namedarray’. See the design doc for more details. Reach out to us on this [discussion topic](pydata/xarray#8080) if you have any thoughts.

New Features#

  • Coordinates can now be constructed independently of any Dataset or DataArray (it is also returned by the Dataset.coords and DataArray.coords properties). Coordinates objects are useful for passing both coordinate variables and indexes to new Dataset / DataArray objects, e.g., via their constructor or via Dataset.assign_coords(). We may also wrap coordinate variables in a Coordinates object in order to skip the automatic creation of (pandas) indexes for dimension coordinates. The Coordinates.from_pandas_multiindex constructor may be used to create coordinates directly from a pandas.MultiIndex object (it is preferred over passing it directly as coordinate data, which may be deprecated soon). Like Dataset and DataArray objects, Coordinates objects may now be used in align() and merge(). (GH6392, PR7368). By Benoît Bovy.

  • Visually group together coordinates with the same indexes in the index section of the text repr (PR7225). By Justus Magin.

  • Allow creating Xarray objects where a multidimensional variable shares its name with a dimension. Examples include output from finite volume models like FVCOM. (GH2233, PR7989) By Deepak Cherian and Benoit Bovy.

  • When outputting Dataset objects as Zarr via Dataset.to_zarr(), user can now specify that chunks that will contain no valid data will not be written. Originally, this could be done by specifying "write_empty_chunks": True in the encoding parameter; however, this setting would not carry over when appending new data to an existing dataset. (GH8009) Requires zarr>=2.11.

Breaking changes#

  • The minimum versions of some dependencies were changed (PR8022):

    Package

    Old

    New

    boto3

    1.20

    1.24

    cftime

    1.5

    1.6

    dask-core

    2022.1

    2022.7

    distributed

    2022.1

    2022.7

    hfnetcdf

    0.13

    1.0

    iris

    3.1

    3.2

    lxml

    4.7

    4.9

    netcdf4

    1.5.7

    1.6.0

    numpy

    1.21

    1.22

    pint

    0.18

    0.19

    pydap

    3.2

    3.3

    rasterio

    1.2

    1.3

    scipy

    1.7

    1.8

    toolz

    0.11

    0.12

    typing_extensions

    4.0

    4.3

    zarr

    2.10

    2.12

    numbagg

    0.1

    0.2.1

Documentation#

Internal Changes#

v2023.07.0 (July 17, 2023)#

This release brings improvements to the documentation on wrapping numpy-like arrays, improved docstrings, and bug fixes.

Deprecations#

Bug fixes#

Documentation#

Internal Changes#

  • Allow chunked non-dask arrays (i.e. Cubed arrays) in groupby operations. (PR7941) By Tom Nicholas.

v2023.06.0 (June 21, 2023)#

This release adds features to curvefit, improves the performance of concatenation, and fixes various bugs.

Thank to our 13 contributors to this release: Anderson Banihirwe, Deepak Cherian, dependabot[bot], Illviljan, Juniper Tyree, Justus Magin, Martin Fleischmann, Mattia Almansi, mgunyho, Rutger van Haasteren, Thomas Nicholas, Tom Nicholas, Tom White.

New Features#

Breaking changes#

Deprecations#

Performance#

Bug fixes#

Documentation#

Internal Changes#

v2023.05.0 (May 18, 2023)#

This release adds some new methods and operators, updates our deprecation policy for python versions, fixes some bugs with groupby, and introduces experimental support for alternative chunked parallel array computation backends via a new plugin system!

Note: If you are using a locally-installed development version of xarray then pulling the changes from this release may require you to re-install. This avoids an error where xarray cannot detect dask via the new entrypoints system introduced in PR7019. See GH7856 for details.

Thanks to our 14 contributors: Alan Brammer, crusaderky, David Stansby, dcherian, Deeksha, Deepak Cherian, Illviljan, James McCreight, Joe Hamman, Justus Magin, Kyle Sunden, Max Hollmann, mgunyho, and Tom Nicholas

New Features#

Breaking changes#

Performance#

Bug fixes#

Internal Changes#

  • Experimental support for wrapping chunked array libraries other than dask. A new ABC is defined - xr.core.parallelcompat.ChunkManagerEntrypoint - which can be subclassed and then registered by alternative chunked array implementations. (GH6807, PR7019) By Tom Nicholas.

v2023.04.2 (April 20, 2023)#

This is a patch release to fix a bug with binning (GH7766)

Bug fixes#

Documentation#

v2023.04.1 (April 18, 2023)#

This is a patch release to fix a bug with binning (GH7759)

Bug fixes#

  • Fix binning by unsorted arrays. (GH7759)

v2023.04.0 (April 14, 2023)#

This release includes support for pandas v2, allows refreshing of backend engines in a session, and removes deprecated backends for rasterio and cfgrib.

Thanks to our 19 contributors: Chinemere, Tom Coleman, Deepak Cherian, Harshitha, Illviljan, Jessica Scheick, Joe Hamman, Justus Magin, Kai Mühlbauer, Kwonil-Kim, Mary Gathoni, Michael Niklas, Pierre, Scott Henderson, Shreyal Gupta, Spencer Clark, mccloskey, nishtha981, veenstrajelmer

We welcome the following new contributors to Xarray!: Mary Gathoni, Harshitha, veenstrajelmer, Chinemere, nishtha981, Shreyal Gupta, Kwonil-Kim, mccloskey.

New Features#

Breaking changes#

Deprecations#

Performance#

Bug fixes#

Documentation#

Internal Changes#

  • Don’t assume that arrays read from disk will be Numpy arrays. This is a step toward enabling reads from a Zarr store using the Kvikio or TensorStore libraries. (PR6874). By Deepak Cherian.

  • Remove internal support for reading GRIB files through the cfgrib backend. cfgrib now uses the external backend interface, so no existing code should break. By Deepak Cherian.

  • Implement CF coding functions in VariableCoders (PR7719). By Kai Mühlbauer

  • Added a config.yml file with messages for the welcome bot when a Github user creates their first ever issue or pull request or has their first PR merged. (GH7685, PR7685) By Nishtha P.

  • Ensure that only nanosecond-precision pd.Timestamp objects continue to be used internally under pandas version 2.0.0. This is mainly to ease the transition to this latest version of pandas. It should be relaxed when addressing GH7493. By Spencer Clark (GH7707, PR7731).

v2023.03.0 (March 22, 2023)#

This release brings many bug fixes, and some new features. The maximum pandas version is pinned to <2 until we can support the new pandas datetime types. Thanks to our 19 contributors: Abel Aoun, Alex Goodman, Deepak Cherian, Illviljan, Jody Klymak, Joe Hamman, Justus Magin, Mary Gathoni, Mathias Hauser, Mattia Almansi, Mick, Oriol Abril-Pla, Patrick Hoefler, Paul Ockenfuß, Pierre, Shreyal Gupta, Spencer Clark, Tom Nicholas, Tom Vo

New Features#

Breaking changes#

Deprecations#

  • Following pandas, the base and loffset parameters of xr.DataArray.resample() and xr.Dataset.resample() have been deprecated and will be removed in a future version of xarray. Using the origin or offset parameters is recommended as a replacement for using the base parameter and using time offset arithmetic is recommended as a replacement for using the loffset parameter (PR8459). By Spencer Clark.

Bug fixes#

Documentation#

Internal Changes#

v2023.02.0 (Feb 7, 2023)#

This release brings a major upgrade to xarray.concat(), many bug fixes, and a bump in supported dependency versions. Thanks to our 11 contributors: Aron Gergely, Deepak Cherian, Illviljan, James Bourbeau, Joe Hamman, Justus Magin, Hauke Schulz, Kai Mühlbauer, Ken Mankoff, Spencer Clark, Tom Nicholas.

Breaking changes#

  • Support for python 3.8 has been dropped and the minimum versions of some dependencies were changed (PR7461):

    Package

    Old

    New

    python

    3.8

    3.9

    numpy

    1.20

    1.21

    pandas

    1.3

    1.4

    dask

    2021.11

    2022.1

    distributed

    2021.11

    2022.1

    h5netcdf

    0.11

    0.13

    lxml

    4.6

    4.7

    numba

    5.4

    5.5

Deprecations#

Bug fixes#

Documentation#

v2023.01.0 (Jan 17, 2023)#

This release includes a number of bug fixes. Thanks to the 14 contributors to this release: Aron Gergely, Benoit Bovy, Deepak Cherian, Ian Carroll, Illviljan, Joe Hamman, Justus Magin, Mark Harfouche, Matthew Roeschke, Paige Martin, Pierre, Sam Levang, Tom White, stefank0.

Breaking changes#

Bug fixes#

Internal Changes#

  • Add the pre-commit hook absolufy-imports to convert relative xarray imports to absolute imports (PR7204, PR7370). By Jimmy Westling.

v2022.12.0 (2022 Dec 2)#

This release includes a number of bug fixes and experimental support for Zarr V3. Thanks to the 16 contributors to this release: Deepak Cherian, Francesco Zanetta, Gregory Lee, Illviljan, Joe Hamman, Justus Magin, Luke Conibear, Mark Harfouche, Mathias Hauser, Mick, Mike Taves, Sam Levang, Spencer Clark, Tom Nicholas, Wei Ji, templiert

New Features#

Breaking changes#

  • The minimum versions of some dependencies were changed (PR7300):

    Package

    Old

    New

    boto

    1.18

    1.20

    cartopy

    0.19

    0.20

    distributed

    2021.09

    2021.11

    dask

    2021.09

    2021.11

    h5py

    3.1

    3.6

    hdf5

    1.10

    1.12

    matplotlib-base

    3.4

    3.5

    nc-time-axis

    1.3

    1.4

    netcdf4

    1.5.3

    1.5.7

    packaging

    20.3

    21.3

    pint

    0.17

    0.18

    pseudonetcdf

    3.1

    3.2

    typing_extensions

    3.10

    4.0

Deprecations#

Bug fixes#

Documentation#

  • Add example of reading and writing individual groups to a single netCDF file to I/O docs page. (PR7338) By Tom Nicholas.

Internal Changes#

v2022.11.0 (Nov 4, 2022)#

This release brings a number of bugfixes and documentation improvements. Both text and HTML reprs now have a new “Indexes” section, which we expect will help with development of new Index objects. This release also features more support for the Python Array API.

Many thanks to the 16 contributors to this release: Daniel Goman, Deepak Cherian, Illviljan, Jessica Scheick, Justus Magin, Mark Harfouche, Maximilian Roos, Mick, Patrick Naylor, Pierre, Spencer Clark, Stephan Hoyer, Tom Nicholas, Tom White

New Features#

Breaking changes#

  • repr(ds) may not show the same result because it doesn’t load small, lazy data anymore. Use ds.head().load() when wanting to see just a sample of the data. (GH6722, PR7203). By Jimmy Westling.

  • Many arguments of plotmethods have been made keyword-only.

  • xarray.plot.plot module renamed to xarray.plot.dataarray_plot to prevent shadowing of the plot method. (GH6949, PR7052). By Michael Niklas.

Deprecations#

  • Positional arguments for all plot methods have been deprecated (GH6949, PR7052). By Michael Niklas.

  • xarray.plot.FacetGrid.axes has been renamed to xarray.plot.FacetGrid.axs because it’s not clear if axes refers to single or multiple Axes instances. This aligns with matplotlib.pyplot.subplots. (PR7194) By Jimmy Westling.

Bug fixes#

  • Explicitly opening a file multiple times (e.g., after modifying it on disk) now reopens the file from scratch for h5netcdf and scipy netCDF backends, rather than reusing a cached version (GH4240, GH4862). By Stephan Hoyer.

  • Fixed bug where Dataset.coarsen.construct() would demote non-dimension coordinates to variables. (PR7233) By Tom Nicholas.

  • Raise a TypeError when trying to plot empty data (GH7156, PR7228). By Michael Niklas.

Documentation#

Internal Changes#

  • Doctests fail on any warnings (PR7166) By Maximilian Roos.

  • Improve import time by lazy loading dask.distributed (:pull: 7172).

  • Explicitly specify longdouble=False in cftime.date2num() when encoding times to preserve existing behavior and prevent future errors when it is eventually set to True by default in cftime (PR7171). By Spencer Clark.

  • Improved import time by lazily importing backend modules, matplotlib, dask.array and flox. (GH6726, PR7179) By Michael Niklas.

  • Emit a warning under the development version of pandas when we convert non-nanosecond precision datetime or timedelta values to nanosecond precision. This was required in the past, because pandas previously was not compatible with non-nanosecond precision values. However pandas is currently working towards removing this restriction. When things stabilize in pandas we will likely consider relaxing this behavior in xarray as well (GH7175, PR7201). By Spencer Clark.

v2022.10.0 (Oct 14 2022)#

This release brings numerous bugfixes, a change in minimum supported versions, and a new scatter plot method for DataArrays.

Many thanks to 11 contributors to this release: Anderson Banihirwe, Benoit Bovy, Dan Adriaansen, Illviljan, Justus Magin, Lukas Bindreiter, Mick, Patrick Naylor, Spencer Clark, Thomas Nicholas

New Features#

  • Add scatter plot for datarrays. Scatter plots now also supports 3d plots with the z argument. (PR6778) By Jimmy Westling.

  • Include the variable name in the error message when CF decoding fails to allow for easier identification of problematic variables (GH7145, PR7147). By Spencer Clark.

Breaking changes#

  • The minimum versions of some dependencies were changed:

    Package

    Old

    New

    cftime

    1.4

    1.5

    distributed

    2021.08

    2021.09

    dask

    2021.08

    2021.09

    iris

    2.4

    3.1

    nc-time-axis

    1.2

    1.3

    numba

    0.53

    0.54

    numpy

    1.19

    1.20

    pandas

    1.2

    1.3

    packaging

    20.0

    21.0

    scipy

    1.6

    1.7

    sparse

    0.12

    0.13

    typing_extensions

    3.7

    3.10

    zarr

    2.8

    2.10

Bug fixes#

v2022.09.0 (September 30, 2022)#

This release brings a large number of bugfixes and documentation improvements, as well as an external interface for setting custom indexes!

Many thanks to our 40 contributors:

Anderson Banihirwe, Andrew Ronald Friedman, Bane Sullivan, Benoit Bovy, ColemanTom, Deepak Cherian, Dimitri Papadopoulos Orfanos, Emma Marshall, Fabian Hofmann, Francesco Nattino, ghislainp, Graham Inggs, Hauke Schulz, Illviljan, James Bourbeau, Jody Klymak, Julia Signell, Justus Magin, Keewis, Ken Mankoff, Luke Conibear, Mathias Hauser, Max Jones, mgunyho, Michael Delgado, Mick, Mike Taves, Oliver Lopez, Patrick Naylor, Paul Hockett, Pierre Manchon, Ray Bell, Riley Brady, Sam Levang, Spencer Clark, Stefaan Lippens, Tom Nicholas, Tom White, Travis A. O’Brien, and Zachary Moon.

New Features#

Bug fixes#

Documentation#

Internal Changes#

v2022.06.0 (July 21, 2022)#

This release brings a number of bug fixes and improvements, most notably a major internal refactor of the indexing functionality, the use of flox in groupby operations, and experimental support for the new Python Array API standard. It also stops testing support for the abandoned PyNIO.

Much effort has been made to preserve backwards compatibility as part of the indexing refactor. We are aware of one unfixed issue.

Please also see the whats-new.2022.06.0rc0 for a full list of changes.

Many thanks to our 18 contributors: Bane Sullivan, Deepak Cherian, Dimitri Papadopoulos Orfanos, Emma Marshall, Hauke Schulz, Illviljan, Julia Signell, Justus Magin, Keewis, Mathias Hauser, Michael Delgado, Mick, Pierre Manchon, Ray Bell, Spencer Clark, Stefaan Lippens, Tom White, Travis A. O’Brien,

New Features#

Bug fixes#

Internal Changes#

  • xarray.core.groupby, xarray.core.rolling, xarray.core.rolling_exp, xarray.core.weighted and xarray.core.resample modules are no longer imported by default. (PR6702)

v2022.06.0rc0 (9 June 2022)#

This pre-release brings a number of bug fixes and improvements, most notably a major internal refactor of the indexing functionality and the use of flox in groupby operations. It also stops testing support for the abandoned PyNIO.

Install it using

mamba create -n <name> python=3.10 xarray
python -m pip install --pre --upgrade --no-deps xarray

Many thanks to the 39 contributors:

Abel Soares Siqueira, Alex Santana, Anderson Banihirwe, Benoit Bovy, Blair Bonnett, Brewster Malevich, brynjarmorka, Charles Stern, Christian Jauvin, Deepak Cherian, Emma Marshall, Fabien Maussion, Greg Behm, Guelate Seyo, Illviljan, Joe Hamman, Joseph K Aicher, Justus Magin, Kevin Paul, Louis Stenger, Mathias Hauser, Mattia Almansi, Maximilian Roos, Michael Bauer, Michael Delgado, Mick, ngam, Oleh Khoma, Oriol Abril-Pla, Philippe Blain, PLSeuJ, Sam Levang, Spencer Clark, Stan West, Thomas Nicholas, Thomas Vogt, Tom White, Xianxiang Li

Known Regressions#

  • reset_coords(drop=True) does not create indexes (GH6607)

New Features#

Breaking changes#

  • PyNIO support is now untested. The minimum versions of some dependencies were changed:

    Package

    Old

    New

    cftime

    1.2

    1.4

    dask

    2.30

    2021.4

    distributed

    2.30

    2021.4

    h5netcdf

    0.8

    0.11

    matplotlib-base

    3.3

    3.4

    numba

    0.51

    0.53

    numpy

    1.18

    1.19

    pandas

    1.1

    1.2

    pint

    0.16

    0.17

    rasterio

    1.1

    1.2

    scipy

    1.5

    1.6

    sparse

    0.11

    0.12

    zarr

    2.5

    2.8

  • The Dataset and DataArray rename`` methods do not implicitly add or drop indexes. (PR5692). By Benoît Bovy.

  • Many arguments like keep_attrs, axis, and skipna are now keyword only for all reduction operations like .mean. By Deepak Cherian, Jimmy Westling.

  • Xarray’s ufuncs have been removed, now that they can be replaced by numpy’s ufuncs in all supported versions of numpy. By Maximilian Roos.

  • xr.polyval() now uses the coord argument directly instead of its index coordinate. (PR6548) By Michael Niklas.

Bug fixes#

Documentation#

Performance#

Internal Changes#

  • Many internal changes due to the explicit indexes refactor. See the corresponding pull-request on GitHub for more details. (PR5692). By Benoît Bovy.

v2022.03.0 (2 March 2022)#

This release brings a number of small improvements, as well as a move to calendar versioning (GH6176).

Many thanks to the 16 contributors to the v2022.02.0 release!

Aaron Spring, Alan D. Snow, Anderson Banihirwe, crusaderky, Illviljan, Joe Hamman, Jonas Gliß, Lukas Pilz, Martin Bergemann, Mathias Hauser, Maximilian Roos, Romain Caneill, Stan West, Stijn Van Hoey, Tobias Kölling, and Tom Nicholas.

New Features#

Breaking changes#

Deprecations#

Bug fixes#

Documentation#

  • Delete files of datasets saved to disk while building the documentation and enable building on Windows via sphinx-build (PR6237). By Stan West.

Internal Changes#

v0.21.1 (31 January 2022)#

This is a bugfix release to resolve (GH6216, PR6207).

Bug fixes#

v0.21.0 (27 January 2022)#

Many thanks to the 20 contributors to the v0.21.0 release!

Abel Aoun, Anderson Banihirwe, Ant Gib, Chris Roat, Cindy Chiao, Deepak Cherian, Dominik Stańczak, Fabian Hofmann, Illviljan, Jody Klymak, Joseph K Aicher, Mark Harfouche, Mathias Hauser, Matthew Roeschke, Maximilian Roos, Michael Delgado, Pascal Bourgault, Pierre, Ray Bell, Romain Caneill, Tim Heap, Tom Nicholas, Zeb Nicholls, joseph nowak, keewis.

New Features#

Breaking changes#

  • Rely on matplotlib’s default datetime converters instead of pandas’ (GH6102, PR6109). By Jimmy Westling.

  • Improve repr readability when there are a large number of dimensions in datasets or dataarrays by wrapping the text once the maximum display width has been exceeded. (GH5546, PR5662) By Jimmy Westling.

Deprecations#

  • Removed the lock kwarg from the zarr and pydap backends, completing the deprecation cycle started in GH5256. By Tom Nicholas.

  • Support for python 3.7 has been dropped. (PR5892) By Jimmy Westling.

Bug fixes#

Internal Changes#

v0.20.2 (9 December 2021)#

This is a bugfix release to resolve (GH3391, GH5715). It also includes performance improvements in unstacking to a sparse array and a number of documentation improvements.

Many thanks to the 20 contributors:

Aaron Spring, Alexandre Poux, Deepak Cherian, Enrico Minack, Fabien Maussion, Giacomo Caria, Gijom, Guillaume Maze, Illviljan, Joe Hamman, Joseph Hardin, Kai Mühlbauer, Matt Henderson, Maximilian Roos, Michael Delgado, Robert Gieseke, Sebastian Weigand and Stephan Hoyer.

Breaking changes#

  • Use complex nan when interpolating complex values out of bounds by default (instead of real nan) (PR6019). By Alexandre Poux.

Performance#

Bug fixes#

  • xr.map_blocks() and xr.corr() now work when dask is not installed (GH3391, GH5715, PR5731). By Gijom.

  • Fix plot.line crash for data of shape (1, N) in _title_for_slice on format_item (PR5948). By Sebastian Weigand.

  • Fix a regression in the removal of duplicate backend entrypoints (GH5944, PR5959) By Kai Mühlbauer.

  • Fix an issue that datasets from being saved when time variables with units that cftime can parse but pandas can not were present (PR6049). By Tim Heap.

Documentation#

Internal Changes#

  • Use importlib to replace functionality of pkg_resources in backend plugins tests. (PR5959). By Kai Mühlbauer.

v0.20.1 (5 November 2021)#

This is a bugfix release to fix GH5930.

Bug fixes#

Documentation#

v0.20.0 (1 November 2021)#

This release brings improved support for pint arrays, methods for weighted standard deviation, variance, and sum of squares, the option to disable the use of the bottleneck library, significantly improved performance of unstack, as well as many bugfixes and internal changes.

Many thanks to the 40 contributors to this release!:

Aaron Spring, Akio Taniguchi, Alan D. Snow, arfy slowy, Benoit Bovy, Christian Jauvin, crusaderky, Deepak Cherian, Giacomo Caria, Illviljan, James Bourbeau, Joe Hamman, Joseph K Aicher, Julien Herzen, Kai Mühlbauer, keewis, lusewell, Martin K. Scherer, Mathias Hauser, Max Grover, Maxime Liquet, Maximilian Roos, Mike Taves, Nathan Lis, pmav99, Pushkar Kopparla, Ray Bell, Rio McMahon, Scott Staniewicz, Spencer Clark, Stefan Bender, Taher Chegini, Thomas Nicholas, Tomas Chor, Tom Augspurger, Victor Negîrneac, Zachary Blackwood, Zachary Moon, and Zeb Nicholls.

New Features#

Breaking changes#

  • The minimum versions of some dependencies were changed:

    Package

    Old

    New

    cftime

    1.1

    1.2

    dask

    2.15

    2.30

    distributed

    2.15

    2.30

    lxml

    4.5

    4.6

    matplotlib-base

    3.2

    3.3

    numba

    0.49

    0.51

    numpy

    1.17

    1.18

    pandas

    1.0

    1.1

    pint

    0.15

    0.16

    scipy

    1.4

    1.5

    seaborn

    0.10

    0.11

    sparse

    0.8

    0.11

    toolz

    0.10

    0.11

    zarr

    2.4

    2.5

  • The __repr__ of a xarray.Dataset’s coords and data_vars ignore xarray.set_option(display_max_rows=...) and show the full output when called directly as, e.g., ds.data_vars or print(ds.data_vars) (GH5545, PR5580). By Stefan Bender.

Deprecations#

Bug fixes#

  • Fix ZeroDivisionError from saving dask array with empty dimension (:issue: 5741). By Joseph K Aicher.

  • Fixed performance bug where cftime import attempted within various core operations if cftime not installed (PR5640). By Luke Sewell

  • Fixed bug when combining named DataArrays using combine_by_coords(). (PR5834). By Tom Nicholas.

  • When a custom engine was used in open_dataset() the engine wasn’t initialized properly, causing missing argument errors or inconsistent method signatures. (PR5684) By Jimmy Westling.

  • Numbers are properly formatted in a plot’s title (GH5788, PR5789). By Maxime Liquet.

  • Faceted plots will no longer raise a pint.UnitStrippedWarning when a pint.Quantity array is plotted, and will correctly display the units of the data in the colorbar (if there is one) (PR5886). By Tom Nicholas.

  • With backends, check for path-like objects rather than pathlib.Path type, use os.fspath (PR5879). By Mike Taves.

  • open_mfdataset() now accepts a single pathlib.Path object (:issue: 5881). By Panos Mavrogiorgos.

  • Improved performance of Dataset.unstack() (PR5906). By Tom Augspurger.

Documentation#

  • Users are instructed to try use_cftime=True if a TypeError occurs when combining datasets and one of the types involved is a subclass of cftime.datetime (PR5776). By Zeb Nicholls.

  • A clearer error is now raised if a user attempts to assign a Dataset to a single key of another Dataset. (PR5839) By Tom Nicholas.

Internal Changes#

v0.19.0 (23 July 2021)#

This release brings improvements to plotting of categorical data, the ability to specify how attributes are combined in xarray operations, a new high-level unify_chunks() function, as well as various deprecations, bug fixes, and minor improvements.

Many thanks to the 29 contributors to this release!:

Andrew Williams, Augustus, Aureliana Barghini, Benoit Bovy, crusaderky, Deepak Cherian, ellesmith88, Elliott Sales de Andrade, Giacomo Caria, github-actions[bot], Illviljan, Joeperdefloep, joooeey, Julia Kent, Julius Busecke, keewis, Mathias Hauser, Matthias Göbel, Mattia Almansi, Maximilian Roos, Peter Andreas Entschev, Ray Bell, Sander, Santiago Soler, Sebastian, Spencer Clark, Stephan Hoyer, Thomas Hirtz, Thomas Nicholas.

New Features#

Breaking changes#

  • The default mode for Dataset.to_zarr() when region is set has changed to the new mode="r+", which only allows for overriding pre-existing array values. This is a safer default than the prior mode="a", and allows for higher performance writes (PR5252). By Stephan Hoyer.

  • The main parameter to combine_by_coords() is renamed to data_objects instead of datasets so anyone calling this method using a named parameter will need to update the name accordingly (GH3248, PR4696). By Augustus Ijams.

Deprecations#

Bug fixes#

Internal Changes#

  • Run CI on the first & last python versions supported only; currently 3.7 & 3.9. (PR5433) By Maximilian Roos.

  • Publish test results & timings on each PR. (PR5537) By Maximilian Roos.

  • Explicit indexes refactor: add a xarray.Index.query() method in which one may eventually provide a custom implementation of label-based data selection (not ready yet for public use). Also refactor the internal, pandas-specific implementation into PandasIndex.query() and PandasMultiIndex.query() (PR5322). By Benoit Bovy.

v0.18.2 (19 May 2021)#

This release reverts a regression in xarray’s unstacking of dask-backed arrays.

v0.18.1 (18 May 2021)#

This release is intended as a small patch release to be compatible with the new 2021.5.0 dask.distributed release. It also includes a new drop_duplicates method, some documentation improvements, the beginnings of our internal Index refactoring, and some bug fixes.

Thank you to all 16 contributors!

Anderson Banihirwe, Andrew, Benoit Bovy, Brewster Malevich, Giacomo Caria, Illviljan, James Bourbeau, Keewis, Maximilian Roos, Ravin Kumar, Stephan Hoyer, Thomas Nicholas, Tom Nicholas, Zachary Moon.

New Features#

Bug fixes#

  • Opening netCDF files from a path that doesn’t end in .nc without supplying an explicit engine works again (GH5295), fixing a bug introduced in 0.18.0. By Stephan Hoyer

Documentation#

Internal Changes#

  • Explicit indexes refactor: add an xarray.Index base class and Dataset.xindexes / DataArray.xindexes properties. Also rename PandasIndexAdapter to PandasIndex, which now inherits from xarray.Index (PR5102). By Benoit Bovy.

  • Replace SortedKeysDict with python’s dict, given dicts are now ordered. By Maximilian Roos.

  • Updated the release guide for developers. Now accounts for actions that are automated via github actions. (PR5274). By Tom Nicholas.

v0.18.0 (6 May 2021)#

This release brings a few important performance improvements, a wide range of usability upgrades, lots of bug fixes, and some new features. These include a plugin API to add backend engines, a new theme for the documentation, curve fitting methods, and several new plotting functions.

Many thanks to the 38 contributors to this release: Aaron Spring, Alessandro Amici, Alex Marandon, Alistair Miles, Ana Paula Krelling, Anderson Banihirwe, Aureliana Barghini, Baudouin Raoult, Benoit Bovy, Blair Bonnett, David Trémouilles, Deepak Cherian, Gabriel Medeiros Abrahão, Giacomo Caria, Hauke Schulz, Illviljan, Mathias Hauser, Matthias Bussonnier, Mattia Almansi, Maximilian Roos, Ray Bell, Richard Kleijn, Ryan Abernathey, Sam Levang, Spencer Clark, Spencer Jones, Tammas Loughran, Tobias Kölling, Todd, Tom Nicholas, Tom White, Victor Negîrneac, Xianxiang Li, Zeb Nicholls, crusaderky, dschwoerer, johnomotani, keewis

New Features#

Breaking changes#

  • The minimum versions of some dependencies were changed:

    Package

    Old

    New

    boto3

    1.12

    1.13

    cftime

    1.0

    1.1

    dask

    2.11

    2.15

    distributed

    2.11

    2.15

    matplotlib

    3.1

    3.2

    numba

    0.48

    0.49

  • open_dataset() and open_dataarray() now accept only the first argument as positional, all others need to be passed are keyword arguments. This is part of the refactor to support external backends (GH4309, PR4989). By Alessandro Amici.

  • Functions that are identities for 0d data return the unchanged data if axis is empty. This ensures that Datasets where some variables do not have the averaged dimensions are not accidentally changed (GH4885, PR5207). By David Schwörer.

  • DataArray.coarsen and Dataset.coarsen no longer support passing keep_attrs via its constructor. Pass keep_attrs via the applied function, i.e. use ds.coarsen(...).mean(keep_attrs=False) instead of ds.coarsen(..., keep_attrs=False).mean(). Further, coarsen now keeps attributes per default (PR5227). By Mathias Hauser.

  • switch the default of the merge() combine_attrs parameter to "override". This will keep the current behavior for merging the attrs of variables but stop dropping the attrs of the main objects (PR4902). By Justus Magin.

Deprecations#

Bug fixes#

Documentation#

Internal Changes#

  • Enable displaying mypy error codes and ignore only specific error codes using # type: ignore[error-code] (PR5096). By Mathias Hauser.

  • Replace uses of raises_regex with the more standard pytest.raises(Exception, match="foo"); (PR5188), (PR5191). By Maximilian Roos.

v0.17.0 (24 Feb 2021)#

This release brings a few important performance improvements, a wide range of usability upgrades, lots of bug fixes, and some new features. These include better cftime support, a new quiver plot, better unstack performance, more efficient memory use in rolling operations, and some python packaging improvements. We also have a few documentation improvements (and more planned!).

Many thanks to the 36 contributors to this release: Alessandro Amici, Anderson Banihirwe, Aureliana Barghini, Ayrton Bourn, Benjamin Bean, Blair Bonnett, Chun Ho Chow, DWesl, Daniel Mesejo-León, Deepak Cherian, Eric Keenan, Illviljan, Jens Hedegaard Nielsen, Jody Klymak, Julien Seguinot, Julius Busecke, Kai Mühlbauer, Leif Denby, Martin Durant, Mathias Hauser, Maximilian Roos, Michael Mann, Ray Bell, RichardScottOZ, Spencer Clark, Tim Gates, Tom Nicholas, Yunus Sevinchan, alexamici, aurghs, crusaderky, dcherian, ghislainp, keewis, rhkleijn

Breaking changes#

  • xarray no longer supports python 3.6

    The minimum version policy was changed to also apply to projects with irregular releases. As a result, the minimum versions of some dependencies have changed:

    Package

    Old

    New

    Python

    3.6

    3.7

    setuptools

    38.4

    40.4

    numpy

    1.15

    1.17

    pandas

    0.25

    1.0

    dask

    2.9

    2.11

    distributed

    2.9

    2.11

    bottleneck

    1.2

    1.3

    h5netcdf

    0.7

    0.8

    iris

    2.2

    2.4

    netcdf4

    1.4

    1.5

    pseudonetcdf

    3.0

    3.1

    rasterio

    1.0

    1.1

    scipy

    1.3

    1.4

    seaborn

    0.9

    0.10

    zarr

    2.3

    2.4

    (GH4688, PR4720, PR4907, PR4942)

  • As a result of PR4684 the default units encoding for datetime-like values (np.datetime64[ns] or cftime.datetime) will now always be set such that int64 values can be used. In the past, no units finer than “seconds” were chosen, which would sometimes mean that float64 values were required, which would lead to inaccurate I/O round-trips.

  • Variables referred to in attributes like bounds and grid_mapping can be set as coordinate variables. These attributes are moved to DataArray.encoding from DataArray.attrs. This behaviour is controlled by the decode_coords kwarg to open_dataset() and open_mfdataset(). The full list of decoded attributes is in Weather and climate data (PR2844, GH3689)

  • As a result of PR4911 the output from calling DataArray.sum() or DataArray.prod() on an integer array with skipna=True and a non-None value for min_count will now be a float array rather than an integer array.

Deprecations#

New Features#

Bug fixes#

Documentation#

Internal Changes#

  • Speed up of the continuous integration tests on azure.

    • Switched to mamba and use matplotlib-base for a faster installation of all dependencies (PR4672).

    • Use pytest.mark.skip instead of pytest.mark.xfail for some tests that can currently not succeed (PR4685).

    • Run the tests in parallel using pytest-xdist (PR4694).

    By Justus Magin and Mathias Hauser.

  • Use pyproject.toml instead of the setup_requires option for setuptools (PR4897). By Justus Magin.

  • Replace all usages of assert x.identical(y) with assert_identical(x,  y) for clearer error messages (PR4752). By Maximilian Roos.

  • Speed up attribute style access (e.g. ds.somevar instead of ds["somevar"]) and tab completion in IPython (GH4741, PR4742). By Richard Kleijn.

  • Added the set_close method to Dataset and DataArray for backends to specify how to voluntary release all resources. (PR#4809) By Alessandro Amici.

  • Update type hints to work with numpy v1.20 (PR4878). By Mathias Hauser.

  • Ensure warnings cannot be turned into exceptions in testing.assert_equal() and the other assert_* functions (PR4864). By Mathias Hauser.

  • Performance improvement when constructing DataArrays. Significantly speeds up repr for Datasets with large number of variables. By Deepak Cherian.

v0.16.2 (30 Nov 2020)#

This release brings the ability to write to limited regions of zarr files, open zarr files with open_dataset() and open_mfdataset(), increased support for propagating attrs using the keep_attrs flag, as well as numerous bugfixes and documentation improvements.

Many thanks to the 31 contributors who contributed to this release: Aaron Spring, Akio Taniguchi, Aleksandar Jelenak, alexamici, Alexandre Poux, Anderson Banihirwe, Andrew Pauling, Ashwin Vishnu, aurghs, Brian Ward, Caleb, crusaderky, Dan Nowacki, darikg, David Brochart, David Huard, Deepak Cherian, Dion Häfner, Gerardo Rivera, Gerrit Holl, Illviljan, inakleinbottle, Jacob Tomlinson, James A. Bednar, jenssss, Joe Hamman, johnomotani, Joris Van den Bossche, Julia Kent, Julius Busecke, Kai Mühlbauer, keewis, Keisuke Fujii, Kyle Cranmer, Luke Volpatti, Mathias Hauser, Maximilian Roos, Michaël Defferrard, Michal Baumgartner, Nick R. Papior, Pascal Bourgault, Peter Hausamann, PGijsbers, Ray Bell, Romain Martinez, rpgoldman, Russell Manser, Sahid Velji, Samnan Rahee, Sander, Spencer Clark, Stephan Hoyer, Thomas Zilio, Tobias Kölling, Tom Augspurger, Wei Ji, Yash Saboo, Zeb Nicholls,

Deprecations#

New Features#

Bug fixes#

  • Fix bug where reference times without padded years (e.g. since 1-1-1) would lose their units when being passed by encode_cf_datetime (GH4422, PR4506). Such units are ambiguous about which digit represents the years (is it YMD or DMY?). Now, if such formatting is encountered, it is assumed that the first digit is the years, they are padded appropriately (to e.g. since 0001-1-1) and a warning that this assumption is being made is issued. Previously, without cftime, such times would be silently parsed incorrectly (at least based on the CF conventions) e.g. “since 1-1-1” would be parsed (via pandas and dateutil) to since 2001-1-1. By Zeb Nicholls.

  • Fix DataArray.plot.step(). By Deepak Cherian.

  • Fix bug where reading a scalar value from a NetCDF file opened with the h5netcdf backend would raise a ValueError when decode_cf=True (GH4471, PR4485). By Gerrit Holl.

  • Fix bug where datetime64 times are silently changed to incorrect values if they are outside the valid date range for ns precision when provided in some other units (GH4427, PR4454). By Andrew Pauling

  • Fix silently overwriting the engine key when passing open_dataset() a file object to an incompatible netCDF (GH4457). Now incompatible combinations of files and engines raise an exception instead. By Alessandro Amici.

  • The min_count argument to DataArray.sum() and DataArray.prod() is now ignored when not applicable, i.e. when skipna=False or when skipna=None and the dtype does not have a missing value (GH4352). By Mathias Hauser.

  • combine_by_coords() now raises an informative error when passing coordinates with differing calendars (GH4495). By Mathias Hauser.

  • DataArray.rolling and Dataset.rolling now also keep the attributes and names of of (wrapped) DataArray objects, previously only the global attributes were retained (GH4497, PR4510). By Mathias Hauser.

  • Improve performance where reading small slices from huge dimensions was slower than necessary (PR4560). By Dion Häfner.

  • Fix bug where dask_gufunc_kwargs was silently changed in apply_ufunc() (PR4576). By Kai Mühlbauer.

Documentation#

Internal Changes#

v0.16.1 (2020-09-20)#

This patch release fixes an incompatibility with a recent pandas change, which was causing an issue indexing with a datetime64. It also includes improvements to rolling, to_dataframe, cov & corr methods and bug fixes. Our documentation has a number of improvements, including fixing all doctests and confirming their accuracy on every commit.

Many thanks to the 36 contributors who contributed to this release:

Aaron Spring, Akio Taniguchi, Aleksandar Jelenak, Alexandre Poux, Caleb, Dan Nowacki, Deepak Cherian, Gerardo Rivera, Jacob Tomlinson, James A. Bednar, Joe Hamman, Julia Kent, Kai Mühlbauer, Keisuke Fujii, Mathias Hauser, Maximilian Roos, Nick R. Papior, Pascal Bourgault, Peter Hausamann, Romain Martinez, Russell Manser, Samnan Rahee, Sander, Spencer Clark, Stephan Hoyer, Thomas Zilio, Tobias Kölling, Tom Augspurger, alexamici, crusaderky, darikg, inakleinbottle, jenssss, johnomotani, keewis, and rpgoldman.

Breaking changes#

New Features#

Bug fixes#

Documentation#

Internal Changes#

v0.16.0 (2020-07-11)#

This release adds xarray.cov & xarray.corr for covariance & correlation respectively; the idxmax & idxmin methods, the polyfit method & xarray.polyval for fitting polynomials, as well as a number of documentation improvements, other features, and bug fixes. Many thanks to all 44 contributors who contributed to this release:

Akio Taniguchi, Andrew Williams, Aurélien Ponte, Benoit Bovy, Dave Cole, David Brochart, Deepak Cherian, Elliott Sales de Andrade, Etienne Combrisson, Hossein Madadi, Huite, Joe Hamman, Kai Mühlbauer, Keisuke Fujii, Maik Riechert, Marek Jacob, Mathias Hauser, Matthieu Ancellin, Maximilian Roos, Noah D Brenowitz, Oriol Abril, Pascal Bourgault, Phillip Butcher, Prajjwal Nijhara, Ray Bell, Ryan Abernathey, Ryan May, Spencer Clark, Spencer Hill, Srijan Saurav, Stephan Hoyer, Taher Chegini, Todd, Tom Nicholas, Yohai Bar Sinai, Yunus Sevinchan, arabidopsis, aurghs, clausmichele, dmey, johnomotani, keewis, raphael dussin, risebell

Breaking changes#

New Features#

Enhancements#

Bug fixes#

Documentation#

Internal Changes#

v0.15.1 (23 Mar 2020)#

This release brings many new features such as Dataset.weighted() methods for weighted array reductions, a new jupyter repr by default, and the start of units integration with pint. There’s also the usual batch of usability improvements, documentation additions, and bug fixes.

Breaking changes#

New Features#

Bug fixes#

Documentation#

Internal Changes#

  • Remove the internal import_seaborn function which handled the deprecation of the seaborn.apionly entry point (GH3747). By Mathias Hauser.

  • Don’t test pint integration in combination with datetime objects. (GH3778, PR3788) By Justus Magin.

  • Change test_open_mfdataset_list_attr to only run with dask installed (GH3777, PR3780). By Bruno Pagani.

  • Preserve the ability to index with method="nearest" with a CFTimeIndex with pandas versions greater than 1.0.1 (GH3751). By Spencer Clark.

  • Greater flexibility and improved test coverage of subtracting various types of objects from a CFTimeIndex. By Spencer Clark.

  • Update Azure CI MacOS image, given pending removal. By Maximilian Roos

  • Remove xfails for scipy 1.0.1 for tests that append to netCDF files (PR3805). By Mathias Hauser.

  • Remove conversion to pandas.Panel, given its removal in pandas in favor of xarray’s objects. By Maximilian Roos

v0.15.0 (30 Jan 2020)#

This release brings many improvements to xarray’s documentation: our examples are now binderized notebooks (click here) and we have new example notebooks from our SciPy 2019 sprint (many thanks to our contributors!).

This release also features many API improvements such as a new TimedeltaAccessor and support for CFTimeIndex in interpolate_na()); as well as many bug fixes.

Breaking changes#

  • Bumped minimum tested versions for dependencies:

    • numpy 1.15

    • pandas 0.25

    • dask 2.2

    • distributed 2.2

    • scipy 1.3

  • Remove compat and encoding kwargs from DataArray, which have been deprecated since 0.12. (PR3650). Instead, specify the encoding kwarg when writing to disk or set the DataArray.encoding attribute directly. By Maximilian Roos.

  • xarray.dot(), DataArray.dot(), and the @ operator now use align="inner" (except when xarray.set_options(arithmetic_join="exact"); GH3694) by Mathias Hauser.

New Features#

Bug fixes#

Documentation#

Internal Changes#

  • Make sure dask names change when rechunking by different chunk sizes. Conversely, make sure they stay the same when rechunking by the same chunk size. (GH3350) By Deepak Cherian.

  • 2x to 5x speed boost (on small arrays) for Dataset.isel(), DataArray.isel(), and DataArray.__getitem__() when indexing by int, slice, list of int, scalar ndarray, or 1-dimensional ndarray. (PR3533) by Guido Imperiale.

  • Removed internal method Dataset._from_vars_and_coord_names, which was dominated by Dataset._construct_direct. (PR3565) By Maximilian Roos.

  • Replaced versioneer with setuptools-scm. Moved contents of setup.py to setup.cfg. Removed pytest-runner from setup.py, as per deprecation notice on the pytest-runner project. (PR3714) by Guido Imperiale.

  • Use of isort is now enforced by CI. (PR3721) by Guido Imperiale

v0.14.1 (19 Nov 2019)#

Breaking changes#

  • Broken compatibility with cftime < 1.0.3 . By Deepak Cherian.

    Warning

    cftime version 1.0.4 is broken (cftime/126); please use version 1.0.4.2 instead.

  • All leftover support for dates from non-standard calendars through netcdftime, the module included in versions of netCDF4 prior to 1.4 that eventually became the cftime package, has been removed in favor of relying solely on the standalone cftime package (PR3450). By Spencer Clark.

New Features#

Bug fixes#

Documentation#

Internal Changes#

v0.14.0 (14 Oct 2019)#

Breaking changes#

  • This release introduces a rolling policy for minimum dependency versions: Minimum dependency versions.

    Several minimum versions have been increased:

    Package

    Old

    New

    Python

    3.5.3

    3.6

    numpy

    1.12

    1.14

    pandas

    0.19.2

    0.24

    dask

    0.16 (tested: 2.4)

    1.2

    bottleneck

    1.1 (tested: 1.2)

    1.2

    matplotlib

    1.5 (tested: 3.1)

    3.1

    Obsolete patch versions (x.y.Z) are not tested anymore. The oldest supported versions of all optional dependencies are now covered by automated tests (before, only the very latest versions were tested).

    (GH3222, GH3293, GH3340, GH3346, GH3358). By Guido Imperiale.

  • Dropped the drop=False optional parameter from Variable.isel(). It was unused and doesn’t make sense for a Variable. (PR3375). By Guido Imperiale.

  • Remove internal usage of collections.OrderedDict. After dropping support for Python <=3.5, most uses of OrderedDict in xarray were no longer necessary. We have removed the internal use of the OrderedDict in favor of Python’s builtin dict object which is now ordered itself. This change will be most obvious when interacting with the attrs property on Dataset and DataArray objects. (GH3380, PR3389). By Joe Hamman.

New functions/methods#

Enhancements#

Bug fixes#

  • Reintroduce support for weakref (broken in v0.13.0). Support has been reinstated for DataArray and Dataset objects only. Internal xarray objects remain unaddressable by weakref in order to save memory (GH3317). By Guido Imperiale.

  • Line plots with the x or y argument set to a 1D non-dimensional coord now plot the correct data for 2D DataArrays (GH3334). By Tom Nicholas.

  • Make concat() more robust when merging variables present in some datasets but not others (GH508). By Deepak Cherian.

  • The default behaviour of reducing across all dimensions for DataArrayGroupBy objects has now been properly removed as was done for DatasetGroupBy in 0.13.0 (GH3337). Use xarray.ALL_DIMS if you need to replicate previous behaviour. Also raise nicer error message when no groups are created (GH1764). By Deepak Cherian.

  • Fix error in concatenating unlabeled dimensions (PR3362). By Deepak Cherian.

  • Warn if the dim kwarg is passed to rolling operations. This is redundant since a dimension is specified when the DatasetRolling or DataArrayRolling object is created. (PR3362). By Deepak Cherian.

Documentation#

v0.13.0 (17 Sep 2019)#

This release includes many exciting changes: wrapping of NEP18 compliant numpy-like arrays; new scatter() plotting method that can scatter two DataArrays in a Dataset against each other; support for converting pandas DataFrames to xarray objects that wrap pydata/sparse; and more!

Breaking changes#

  • This release increases the minimum required Python version from 3.5.0 to 3.5.3 (GH3089). By Guido Imperiale.

  • The isel_points and sel_points methods are removed, having been deprecated since v0.10.0. These are redundant with the isel / sel methods. See Vectorized Indexing for the details By Maximilian Roos

  • The inplace kwarg for public methods now raises an error, having been deprecated since v0.11.0. By Maximilian Roos

  • concat() now requires the dim argument. Its indexers, mode and concat_over kwargs have now been removed. By Deepak Cherian

  • Passing a list of colors in cmap will now raise an error, having been deprecated since v0.6.1.

  • Most xarray objects now define __slots__. This reduces overall RAM usage by ~22% (not counting the underlying numpy buffers); on CPython 3.7/x64, a trivial DataArray has gone down from 1.9kB to 1.5kB.

    Caveats:

    • Pickle streams produced by older versions of xarray can’t be loaded using this release, and vice versa.

    • Any user code that was accessing the __dict__ attribute of xarray objects will break. The best practice to attach custom metadata to xarray objects is to use the attrs dictionary.

    • Any user code that defines custom subclasses of xarray classes must now explicitly define __slots__ itself. Subclasses that don’t add any attributes must state so by defining __slots__ = () right after the class header. Omitting __slots__ will now cause a FutureWarning to be logged, and will raise an error in a later release.

    (GH3250) by Guido Imperiale.

  • The default dimension for Dataset.groupby(), Dataset.resample(), DataArray.groupby() and DataArray.resample() reductions is now the grouping or resampling dimension.

  • DataArray.to_dataset() requires name to be passed as a kwarg (previously ambiguous positional arguments were deprecated)

  • Reindexing with variables of a different dimension now raise an error (previously deprecated)

  • xarray.broadcast_array is removed (previously deprecated in favor of broadcast())

  • Variable.expand_dims is removed (previously deprecated in favor of Variable.set_dims())

New functions/methods#

Enhancements#

  • Multiple enhancements to concat() and open_mfdataset(). By Deepak Cherian

    • Added compat='override'. When merging, this option picks the variable from the first dataset and skips all comparisons.

    • Added join='override'. When aligning, this only checks that index sizes are equal among objects and skips checking indexes for equality.

    • concat() and open_mfdataset() now support the join kwarg. It is passed down to align().

    • concat() now calls merge() on variables that are not concatenated (i.e. variables without concat_dim when data_vars or coords are "minimal"). concat() passes its new compat kwarg down to merge(). (GH2064)

    Users can avoid a common bottleneck when using open_mfdataset() on a large number of files with variables that are known to be aligned and some of which need not be concatenated. Slow equality comparisons can now be avoided, for e.g.:

    data = xr.open_mfdataset(files, concat_dim='time', data_vars='minimal',
                         coords='minimal', compat='override', join='override')

  • In to_zarr(), passing mode is not mandatory if append_dim is set, as it will automatically be set to 'a' internally. By David Brochart.

  • Added the ability to initialize an empty or full DataArray with a single value. (GH277) By Gerardo Rivera.

  • to_netcdf() now supports the invalid_netcdf kwarg when used with engine="h5netcdf". It is passed to h5netcdf.File. By Ulrich Herter.

  • xarray.Dataset.drop now supports keyword arguments; dropping index labels by using both dim and labels or using a DataArrayCoordinates object are deprecated (GH2910). By Gregory Gundersen.

  • Added examples of Dataset.set_index() and DataArray.set_index(), as well are more specific error messages when the user passes invalid arguments (GH3176). By Gregory Gundersen.

  • Dataset.filter_by_attrs() now filters the coordinates as well as the variables. By Spencer Jones.

Bug fixes#

Documentation#

v0.12.3 (10 July 2019)#

New functions/methods#

Enhancements#

Bug fixes#

  • Resolved deprecation warnings from newer versions of matplotlib and dask.

  • Compatibility fixes for the upcoming pandas 0.25 and NumPy 1.17 releases. By Stephan Hoyer.

  • Fix summaries for multiindex coordinates (GH3079). By Jonas Hörsch.

  • Fix HDF5 error that could arise when reading multiple groups from a file at once (GH2954). By Stephan Hoyer.

v0.12.2 (29 June 2019)#

New functions/methods#

  • Two new functions, combine_nested() and combine_by_coords(), allow for combining datasets along any number of dimensions, instead of the one-dimensional list of datasets supported by concat().

    The new combine_nested will accept the datasets as a nested list-of-lists, and combine by applying a series of concat and merge operations. The new combine_by_coords instead uses the dimension coordinates of datasets to order them.

    open_mfdataset() can use either combine_nested or combine_by_coords to combine datasets along multiple dimensions, by specifying the argument combine='nested' or combine='by_coords'.

    The older function auto_combine has been deprecated, because its functionality has been subsumed by the new functions. To avoid FutureWarnings switch to using combine_nested or combine_by_coords, (or set the combine argument in open_mfdataset). (GH2159) By Tom Nicholas.

  • rolling_exp() and rolling_exp() added, similar to pandas’ pd.DataFrame.ewm method. Calling .mean on the resulting object will return an exponentially weighted moving average. By Maximilian Roos.

  • New DataArray.str for string related manipulations, based on pandas.Series.str. By 0x0L.

  • Added strftime method to .dt accessor, making it simpler to hand a datetime DataArray to other code expecting formatted dates and times. (GH2090). strftime() is also now available on CFTimeIndex. By Alan Brammer and Ryan May.

  • GroupBy.quantile is now a method of GroupBy objects (GH3018). By David Huard.

  • Argument and return types are added to most methods on DataArray and Dataset, allowing static type checking both within xarray and external libraries. Type checking with mypy is enabled in CI (though not required yet). By Guido Imperiale and Maximilian Roos.

Enhancements to existing functionality#

Bug fixes#

v0.12.1 (4 April 2019)#

Enhancements#

  • Allow expand_dims method to support inserting/broadcasting dimensions with size > 1. (GH2710) By Martin Pletcher.

Bug fixes#

v0.12.0 (15 March 2019)#

Highlights include:

Deprecations#

  • The compat argument to Dataset and the encoding argument to DataArray are deprecated and will be removed in a future release. (GH1188) By Maximilian Roos.

Other enhancements#

Bug fixes#

  • Silenced warnings that appear when using pandas 0.24. By Stephan Hoyer

  • Interpolating via resample now internally specifies bounds_error=False as an argument to scipy.interpolate.interp1d, allowing for interpolation from higher frequencies to lower frequencies. Datapoints outside the bounds of the original time coordinate are now filled with NaN (GH2197). By Spencer Clark.

  • Line plots with the x argument set to a non-dimensional coord now plot the correct data for 1D DataArrays. (GH2725). By Tom Nicholas.

  • Subtracting a scalar cftime.datetime object from a CFTimeIndex now results in a pandas.TimedeltaIndex instead of raising a TypeError (GH2671). By Spencer Clark.

  • backend_kwargs are no longer ignored when using open_dataset with pynio engine (:issue:’2380’) By Jonathan Joyce.

  • Fix open_rasterio creating a WKT CRS instead of PROJ.4 with rasterio 1.0.14+ (GH2715). By David Hoese.

  • Masking data arrays with xarray.DataArray.where() now returns an array with the name of the original masked array (GH2748 and GH2457). By Yohai Bar-Sinai.

  • Fixed error when trying to reduce a DataArray using a function which does not require an axis argument. (GH2768) By Tom Nicholas.

  • Concatenating a sequence of DataArray with varying names sets the name of the output array to None, instead of the name of the first input array. If the names are the same it sets the name to that, instead to the name of the first DataArray in the list as it did before. (GH2775). By Tom Nicholas.

  • Per the CF conventions section on calendars, specifying 'standard' as the calendar type in cftime_range() now correctly refers to the 'gregorian' calendar instead of the 'proleptic_gregorian' calendar (GH2761).

v0.11.3 (26 January 2019)#

Bug fixes#

  • Saving files with times encoded with reference dates with timezones (e.g. ‘2000-01-01T00:00:00-05:00’) no longer raises an error (GH2649). By Spencer Clark.

  • Fixed performance regression with open_mfdataset (GH2662). By Tom Nicholas.

  • Fixed supplying an explicit dimension in the concat_dim argument to to open_mfdataset (GH2647). By Ben Root.

v0.11.2 (2 January 2019)#

Removes inadvertently introduced setup dependency on pytest-runner (GH2641). Otherwise, this release is exactly equivalent to 0.11.1.

Warning

This is the last xarray release that will support Python 2.7. Future releases will be Python 3 only, but older versions of xarray will always be available for Python 2.7 users. For the more details, see:

v0.11.1 (29 December 2018)#

This minor release includes a number of enhancements and bug fixes, and two (slightly) breaking changes.

Breaking changes#

  • Minimum rasterio version increased from 0.36 to 1.0 (for open_rasterio)

  • Time bounds variables are now also decoded according to CF conventions (GH2565). The previous behavior was to decode them only if they had specific time attributes, now these attributes are copied automatically from the corresponding time coordinate. This might break downstream code that was relying on these variables to be brake downstream code that was relying on these variables to be not decoded. By Fabien Maussion.

Enhancements#

  • Ability to read and write consolidated metadata in zarr stores (GH2558). By Ryan Abernathey.

  • CFTimeIndex uses slicing for string indexing when possible (like pandas.DatetimeIndex), which avoids unnecessary copies. By Stephan Hoyer

  • Enable passing rasterio.io.DatasetReader or rasterio.vrt.WarpedVRT to open_rasterio instead of file path string. Allows for in-memory reprojection, see (GH2588). By Scott Henderson.

  • Like pandas.DatetimeIndex, CFTimeIndex now supports “dayofyear” and “dayofweek” accessors (GH2597). Note this requires a version of cftime greater than 1.0.2. By Spencer Clark.

  • The option 'warn_for_unclosed_files' (False by default) has been added to allow users to enable a warning when files opened by xarray are deallocated but were not explicitly closed. This is mostly useful for debugging; we recommend enabling it in your test suites if you use xarray for IO. By Stephan Hoyer

  • Support Dask HighLevelGraphs by Matthew Rocklin.

  • DataArray.resample() and Dataset.resample() now supports the loffset kwarg just like pandas. By Deepak Cherian

  • Datasets are now guaranteed to have a 'source' encoding, so the source file name is always stored (GH2550). By Tom Nicholas.

  • The apply methods for DatasetGroupBy, DataArrayGroupBy, DatasetResample and DataArrayResample now support passing positional arguments to the applied function as a tuple to the args argument. By Matti Eskelinen.

  • 0d slices of ndarrays are now obtained directly through indexing, rather than extracting and wrapping a scalar, avoiding unnecessary copying. By Daniel Wennberg.

  • Added support for fill_value with shift() and shift() By Maximilian Roos

Bug fixes#

  • Ensure files are automatically closed, if possible, when no longer referenced by a Python variable (GH2560). By Stephan Hoyer

  • Fixed possible race conditions when reading/writing to disk in parallel (GH2595). By Stephan Hoyer

  • Fix h5netcdf saving scalars with filters or chunks (GH2563). By Martin Raspaud.

  • Fix parsing of _Unsigned attribute set by OPENDAP servers. (GH2583). By Deepak Cherian

  • Fix failure in time encoding when exporting to netCDF with versions of pandas less than 0.21.1 (GH2623). By Spencer Clark.

  • Fix MultiIndex selection to update label and level (GH2619). By Keisuke Fujii.

v0.11.0 (7 November 2018)#

Breaking changes#

  • Finished deprecations (changed behavior with this release):

    • Dataset.T has been removed as a shortcut for Dataset.transpose(). Call Dataset.transpose() directly instead.

    • Iterating over a Dataset now includes only data variables, not coordinates. Similarly, calling len and bool on a Dataset now includes only data variables.

    • DataArray.__contains__ (used by Python’s in operator) now checks array data, not coordinates.

    • The old resample syntax from before xarray 0.10, e.g., data.resample('1D', dim='time', how='mean'), is no longer supported will raise an error in most cases. You need to use the new resample syntax instead, e.g., data.resample(time='1D').mean() or data.resample({'time': '1D'}).mean().

  • New deprecations (behavior will be changed in xarray 0.12):

  • Refactored storage backends:

    • Xarray’s storage backends now automatically open and close files when necessary, rather than requiring opening a file with autoclose=True. A global least-recently-used cache is used to store open files; the default limit of 128 open files should suffice in most cases, but can be adjusted if necessary with xarray.set_options(file_cache_maxsize=...). The autoclose argument to open_dataset and related functions has been deprecated and is now a no-op.

      This change, along with an internal refactor of xarray’s storage backends, should significantly improve performance when reading and writing netCDF files with Dask, especially when working with many files or using Dask Distributed. By Stephan Hoyer

  • Support for non-standard calendars used in climate science:

    • Xarray will now always use cftime.datetime objects, rather than by default trying to coerce them into np.datetime64[ns] objects. A CFTimeIndex will be used for indexing along time coordinates in these cases.

    • A new method to_datetimeindex() has been added to aid in converting from a CFTimeIndex to a pandas.DatetimeIndex for the remaining use-cases where using a CFTimeIndex is still a limitation (e.g. for resample or plotting).

    • Setting the enable_cftimeindex option is now a no-op and emits a FutureWarning.

Enhancements#

  • xarray.DataArray.plot.line() can now accept multidimensional coordinate variables as input. hue must be a dimension name in this case. (GH2407) By Deepak Cherian.

  • Added support for Python 3.7. (GH2271). By Joe Hamman.

  • Added support for plotting data with pandas.Interval coordinates, such as those created by groupby_bins() By Maximilian Maahn.

  • Added shift() for shifting the values of a CFTimeIndex by a specified frequency. (GH2244). By Spencer Clark.

  • Added support for using cftime.datetime coordinates with differentiate(), differentiate(), interp(), and interp(). By Spencer Clark

  • There is now a global option to either always keep or always discard dataset and dataarray attrs upon operations. The option is set with xarray.set_options(keep_attrs=True), and the default is to use the old behaviour. By Tom Nicholas.

  • Added a new backend for the GRIB file format based on ECMWF cfgrib python driver and ecCodes C-library. (GH2475) By Alessandro Amici, sponsored by ECMWF.

  • Resample now supports a dictionary mapping from dimension to frequency as its first argument, e.g., data.resample({'time': '1D'}).mean(). This is consistent with other xarray functions that accept either dictionaries or keyword arguments. By Stephan Hoyer.

  • The preferred way to access tutorial data is now to load it lazily with xarray.tutorial.open_dataset(). xarray.tutorial.load_dataset() calls Dataset.load() prior to returning (and is now deprecated). This was changed in order to facilitate using tutorial datasets with dask. By Joe Hamman.

  • DataArray can now use xr.set_option(keep_attrs=True) and retain attributes in binary operations, such as (+, -, * ,/). Default behaviour is unchanged (Attributes will be dismissed). By Michael Blaschek

Bug fixes#

  • FacetGrid now properly uses the cbar_kwargs keyword argument. (GH1504, GH1717) By Deepak Cherian.

  • Addition and subtraction operators used with a CFTimeIndex now preserve the index’s type. (GH2244). By Spencer Clark.

  • We now properly handle arrays of datetime.datetime and datetime.timedelta provided as coordinates. (GH2512) By Deepak Cherian.

  • xarray.DataArray.roll correctly handles multidimensional arrays. (GH2445) By Keisuke Fujii.

  • xarray.plot() now properly accepts a norm argument and does not override the norm’s vmin and vmax. (GH2381) By Deepak Cherian.

  • xarray.DataArray.std() now correctly accepts ddof keyword argument. (GH2240) By Keisuke Fujii.

  • Restore matplotlib’s default of plotting dashed negative contours when a single color is passed to DataArray.contour() e.g. colors='k'. By Deepak Cherian.

  • Fix a bug that caused some indexing operations on arrays opened with open_rasterio to error (GH2454). By Stephan Hoyer.

  • Subtracting one CFTimeIndex from another now returns a pandas.TimedeltaIndex, analogous to the behavior for DatetimeIndexes (GH2484). By Spencer Clark.

  • Adding a TimedeltaIndex to, or subtracting a TimedeltaIndex from a CFTimeIndex is now allowed (GH2484). By Spencer Clark.

  • Avoid use of Dask’s deprecated get= parameter in tests by Matthew Rocklin.

  • An OverflowError is now accurately raised and caught during the encoding process if a reference date is used that is so distant that the dates must be encoded using cftime rather than NumPy (GH2272). By Spencer Clark.

  • Chunked datasets can now roundtrip to Zarr storage continually with to_zarr and open_zarr (GH2300). By Lily Wang.

v0.10.9 (21 September 2018)#

This minor release contains a number of backwards compatible enhancements.

Announcements of note:

  • Xarray is now a NumFOCUS fiscally sponsored project! Read the announcement for more details.

  • We have a new Development roadmap that outlines our future development plans.

  • Dataset.apply now properly documents the way func is called. By Matti Eskelinen.

Enhancements#

  • differentiate() and differentiate() are newly added. (GH1332) By Keisuke Fujii.

  • Default colormap for sequential and divergent data can now be set via set_options() (GH2394) By Julius Busecke.

  • min_count option is newly supported in sum(), prod() and sum(), and prod(). (GH2230) By Keisuke Fujii.

  • plot() now accepts the kwargs xscale, yscale, xlim, ylim, xticks, yticks just like pandas. Also xincrease=False, yincrease=False now use matplotlib’s axis inverting methods instead of setting limits. By Deepak Cherian. (GH2224)

  • DataArray coordinates and Dataset coordinates and data variables are now displayed as a b … y z rather than a b c d …. (GH1186) By Seth P.

  • A new CFTimeIndex-enabled cftime_range() function for use in generating dates from standard or non-standard calendars. By Spencer Clark.

  • When interpolating over a datetime64 axis, you can now provide a datetime string instead of a datetime64 object. E.g. da.interp(time='1991-02-01') (GH2284) By Deepak Cherian.

  • A clear error message is now displayed if a set or dict is passed in place of an array (GH2331) By Maximilian Roos.

  • Applying unstack to a large DataArray or Dataset is now much faster if the MultiIndex has not been modified after stacking the indices. (GH1560) By Maximilian Maahn.

  • You can now control whether or not to offset the coordinates when using the roll method and the current behavior, coordinates rolled by default, raises a deprecation warning unless explicitly setting the keyword argument. (GH1875) By Andrew Huang.

  • You can now call unstack without arguments to unstack every MultiIndex in a DataArray or Dataset. By Julia Signell.

  • Added the ability to pass a data kwarg to copy to create a new object with the same metadata as the original object but using new values. By Julia Signell.

Bug fixes#

  • xarray.plot.imshow() correctly uses the origin argument. (GH2379) By Deepak Cherian.

  • Fixed DataArray.to_iris() failure while creating DimCoord by falling back to creating AuxCoord. Fixed dependency on var_name attribute being set. (GH2201) By Thomas Voigt.

  • Fixed a bug in zarr backend which prevented use with datasets with invalid chunk size encoding after reading from an existing store (GH2278). By Joe Hamman.

  • Tests can be run in parallel with pytest-xdist By Tony Tung.

  • Follow up the renamings in dask; from dask.ghost to dask.overlap By Keisuke Fujii.

  • Now raises a ValueError when there is a conflict between dimension names and level names of MultiIndex. (GH2299) By Keisuke Fujii.

  • Follow up the renamings in dask; from dask.ghost to dask.overlap By Keisuke Fujii.

  • Now apply_ufunc() raises a ValueError when the size of input_core_dims is inconsistent with the number of arguments. (GH2341) By Keisuke Fujii.

  • Fixed Dataset.filter_by_attrs() behavior not matching netCDF4.Dataset.get_variables_by_attributes(). When more than one key=value is passed into Dataset.filter_by_attrs() it will now return a Dataset with variables which pass all the filters. (GH2315) By Andrew Barna.

v0.10.8 (18 July 2018)#

Breaking changes#

  • Xarray no longer supports python 3.4. Additionally, the minimum supported versions of the following dependencies has been updated and/or clarified:

    • pandas: 0.18 -> 0.19

    • NumPy: 1.11 -> 1.12

    • Dask: 0.9 -> 0.16

    • Matplotlib: unspecified -> 1.5

    (GH2204). By Joe Hamman.

Enhancements#

Bug fixes#

  • Fixed a bug in zarr backend which prevented use with datasets with incomplete chunks in multiple dimensions (GH2225). By Joe Hamman.

  • Fixed a bug in to_netcdf() which prevented writing datasets when the arrays had different chunk sizes (GH2254). By Mike Neish.

  • Fixed masking during the conversion to cdms2 objects by to_cdms2() (GH2262). By Stephane Raynaud.

  • Fixed a bug in 2D plots which incorrectly raised an error when 2D coordinates weren’t monotonic (GH2250). By Fabien Maussion.

  • Fixed warning raised in to_netcdf() due to deprecation of effective_get in dask (GH2238). By Joe Hamman.

v0.10.7 (7 June 2018)#

Enhancements#

Bug fixes#

  • Fixed a bug in rasterio backend which prevented use with distributed. The rasterio backend now returns pickleable objects (GH2021). By Joe Hamman.

v0.10.6 (31 May 2018)#

The minor release includes a number of bug-fixes and backwards compatible enhancements.

Enhancements#

  • New PseudoNetCDF backend for many Atmospheric data formats including GEOS-Chem, CAMx, NOAA arlpacked bit and many others. See io.PseudoNetCDF for more details. By Barron Henderson.

  • The Dataset constructor now aligns DataArray arguments in data_vars to indexes set explicitly in coords, where previously an error would be raised. (GH674) By Maximilian Roos.

  • sel(), isel() & reindex(), (and their Dataset counterparts) now support supplying a dict as a first argument, as an alternative to the existing approach of supplying kwargs. This allows for more robust behavior of dimension names which conflict with other keyword names, or are not strings. By Maximilian Roos.

  • rename() now supports supplying **kwargs, as an alternative to the existing approach of supplying a dict as the first argument. By Maximilian Roos.

  • cumsum() and cumprod() now support aggregation over multiple dimensions at the same time. This is the default behavior when dimensions are not specified (previously this raised an error). By Stephan Hoyer

  • DataArray.dot() and dot() are partly supported with older dask<0.17.4. (related to GH2203) By Keisuke Fujii.

  • Xarray now uses Versioneer to manage its version strings. (GH1300). By Joe Hamman.

Bug fixes#

  • Fixed a regression in 0.10.4, where explicitly specifying dtype='S1' or dtype=str in encoding with to_netcdf() raised an error (GH2149). Stephan Hoyer

  • apply_ufunc() now directly validates output variables (GH1931). By Stephan Hoyer.

  • Fixed a bug where to_netcdf(..., unlimited_dims='bar') yielded NetCDF files with spurious 0-length dimensions (i.e. b, a, and r) (GH2134). By Joe Hamman.

  • Removed spurious warnings with Dataset.update(Dataset) (GH2161) and array.equals(array) when array contains NaT (GH2162). By Stephan Hoyer.

  • Aggregations with Dataset.reduce() (including mean, sum, etc) no longer drop unrelated coordinates (GH1470). Also fixed a bug where non-scalar data-variables that did not include the aggregation dimension were improperly skipped. By Stephan Hoyer

  • Fix stack() with non-unique coordinates on pandas 0.23 (GH2160). By Stephan Hoyer

  • Selecting data indexed by a length-1 CFTimeIndex with a slice of strings now behaves as it does when using a length-1 DatetimeIndex (i.e. it no longer falsely returns an empty array when the slice includes the value in the index) (GH2165). By Spencer Clark.

  • Fix DataArray.groupby().reduce() mutating coordinates on the input array when grouping over dimension coordinates with duplicated entries (GH2153). By Stephan Hoyer

  • Fix Dataset.to_netcdf() cannot create group with engine="h5netcdf" (GH2177). By Stephan Hoyer

v0.10.4 (16 May 2018)#

The minor release includes a number of bug-fixes and backwards compatible enhancements. A highlight is CFTimeIndex, which offers support for non-standard calendars used in climate modeling.

Documentation#

Enhancements#

  • Add an option for using a CFTimeIndex for indexing times with non-standard calendars and/or outside the Timestamp-valid range; this index enables a subset of the functionality of a standard pandas.DatetimeIndex. See Non-standard calendars and dates outside the nanosecond-precision range for full details. (GH789, GH1084, GH1252) By Spencer Clark with help from Stephan Hoyer.

  • Allow for serialization of cftime.datetime objects (GH789, GH1084, GH2008, GH1252) using the standalone cftime library. By Spencer Clark.

  • Support writing lists of strings as netCDF attributes (GH2044). By Dan Nowacki.

  • to_netcdf() with engine='h5netcdf' now accepts h5py encoding settings compression and compression_opts, along with the NetCDF4-Python style settings gzip=True and complevel. This allows using any compression plugin installed in hdf5, e.g. LZF (GH1536). By Guido Imperiale.

  • dot() on dask-backed data will now call dask.array.einsum(). This greatly boosts speed and allows chunking on the core dims. The function now requires dask >= 0.17.3 to work on dask-backed data (GH2074). By Guido Imperiale.

  • plot.line() learned new kwargs: xincrease, yincrease that change the direction of the respective axes. By Deepak Cherian.

  • Added the parallel option to open_mfdataset(). This option uses dask.delayed to parallelize the open and preprocessing steps within open_mfdataset. This is expected to provide performance improvements when opening many files, particularly when used in conjunction with dask’s multiprocessing or distributed schedulers (GH1981). By Joe Hamman.

  • New compute option in to_netcdf(), to_zarr(), and save_mfdataset() to allow for the lazy computation of netCDF and zarr stores. This feature is currently only supported by the netCDF4 and zarr backends. (GH1784). By Joe Hamman.

Bug fixes#

v0.10.3 (13 April 2018)#

The minor release includes a number of bug-fixes and backwards compatible enhancements.

Enhancements#

Bug fixes#

  • Fixed decode_cf function to operate lazily on dask arrays (GH1372). By Ryan Abernathey.

  • Fixed labeled indexing with slice bounds given by xarray objects with datetime64 or timedelta64 dtypes (GH1240). By Stephan Hoyer.

  • Attempting to convert an xarray.Dataset into a numpy array now raises an informative error message. By Stephan Hoyer.

  • Fixed a bug in decode_cf_datetime where int32 arrays weren’t parsed correctly (GH2002). By Fabien Maussion.

  • When calling xr.auto_combine() or xr.open_mfdataset() with a concat_dim, the resulting dataset will have that one-element dimension (it was silently dropped, previously) (GH1988). By Ben Root.

v0.10.2 (13 March 2018)#

The minor release includes a number of bug-fixes and enhancements, along with one possibly backwards incompatible change.

Backwards incompatible changes#

  • The addition of __array_ufunc__ for xarray objects (see below) means that NumPy ufunc methods (e.g., np.add.reduce) that previously worked on xarray.DataArray objects by converting them into NumPy arrays will now raise NotImplementedError instead. In all cases, the work-around is simple: convert your objects explicitly into NumPy arrays before calling the ufunc (e.g., with .values).

Enhancements#

  • Added dot(), equivalent to numpy.einsum(). Also, dot() now supports dims option, which specifies the dimensions to sum over. (GH1951) By Keisuke Fujii.

  • Support for writing xarray datasets to netCDF files (netcdf4 backend only) when using the dask.distributed scheduler (GH1464). By Joe Hamman.

  • Support lazy vectorized-indexing. After this change, flexible indexing such as orthogonal/vectorized indexing, becomes possible for all the backend arrays. Also, lazy transpose is now also supported. (GH1897) By Keisuke Fujii.

  • Implemented NumPy’s __array_ufunc__ protocol for all xarray objects (GH1617). This enables using NumPy ufuncs directly on xarray.Dataset objects with recent versions of NumPy (v1.13 and newer):

    In [1]: ds = xr.Dataset({"a": 1})

In [2]: np.sin(ds)
Out[2]: 
<xarray.Dataset> Size: 8B
Dimensions:  ()
Data variables:
    a        float64 8B 0.8415

    This obliviates the need for the xarray.ufuncs module, which will be deprecated in the future when xarray drops support for older versions of NumPy. By Stephan Hoyer.

  • Improve rolling() logic. DataArrayRolling() object now supports construct() method that returns a view of the DataArray / Dataset object with the rolling-window dimension added to the last axis. This enables more flexible operation, such as strided rolling, windowed rolling, ND-rolling, short-time FFT and convolution. (GH1831, GH1142, GH819) By Keisuke Fujii.

  • line() learned to make plots with data on x-axis if so specified. (GH575) By Deepak Cherian.

Bug fixes#

v0.10.1 (25 February 2018)#

The minor release includes a number of bug-fixes and backwards compatible enhancements.

Documentation#

Enhancements#

New functions and methods:

Plotting enhancements:

Other enhancements:

  • Reduce methods such as DataArray.sum() now handles object-type array.

    In [3]: da = xr.DataArray(np.array([True, False, np.nan], dtype=object), dims="x")

In [4]: da.sum()
Out[4]: 
<xarray.DataArray ()> Size: 8B
array(1)

    (GH1866) By Keisuke Fujii.

  • Reduce methods such as DataArray.sum() now accepts dtype arguments. (GH1838) By Keisuke Fujii.

  • Added nodatavals attribute to DataArray when using open_rasterio(). (GH1736). By Alan Snow.

  • Use pandas.Grouper class in xarray resample methods rather than the deprecated pandas.TimeGrouper class (GH1766). By Joe Hamman.

  • Experimental support for parsing ENVI metadata to coordinates and attributes in xarray.open_rasterio(). By Matti Eskelinen.

  • Reduce memory usage when decoding a variable with a scale_factor, by converting 8-bit and 16-bit integers to float32 instead of float64 (PR1840), and keeping float16 and float32 as float32 (GH1842). Correspondingly, encoded variables may also be saved with a smaller dtype. By Zac Hatfield-Dodds.

  • Speed of reindexing/alignment with dask array is orders of magnitude faster when inserting missing values (GH1847). By Stephan Hoyer.

  • Fix axis keyword ignored when applying np.squeeze to DataArray (GH1487). By Florian Pinault.

  • netcdf4-python has moved the its time handling in the netcdftime module to a standalone package (netcdftime). As such, xarray now considers netcdftime an optional dependency. One benefit of this change is that it allows for encoding/decoding of datetimes with non-standard calendars without the netcdf4-python dependency (GH1084). By Joe Hamman.

New functions/methods

Bug fixes#

  • Rolling aggregation with center=True option now gives the same result with pandas including the last element (GH1046). By Keisuke Fujii.

  • Support indexing with a 0d-np.ndarray (GH1921). By Keisuke Fujii.

  • Added warning in api.py of a netCDF4 bug that occurs when the filepath has 88 characters (GH1745). By Liam Brannigan.

  • Fixed encoding of multi-dimensional coordinates in to_netcdf() (GH1763). By Mike Neish.

  • Fixed chunking with non-file-based rasterio datasets (GH1816) and refactored rasterio test suite. By Ryan Abernathey

  • Bug fix in open_dataset(engine=’pydap’) (GH1775) By Keisuke Fujii.

  • Bug fix in vectorized assignment (GH1743, GH1744). Now item assignment to __setitem__() checks

  • Bug fix in vectorized assignment (GH1743, GH1744). Now item assignment to DataArray.__setitem__() checks coordinates of target, destination and keys. If there are any conflict among these coordinates, IndexError will be raised. By Keisuke Fujii.

  • Properly point DataArray.__dask_scheduler__ to dask.threaded.get. By Matthew Rocklin.

  • Bug fixes in DataArray.plot.imshow(): all-NaN arrays and arrays with size one in some dimension can now be plotted, which is good for exploring satellite imagery (GH1780). By Zac Hatfield-Dodds.

  • Fixed UnboundLocalError when opening netCDF file (GH1781). By Stephan Hoyer.

  • The variables, attrs, and dimensions properties have been deprecated as part of a bug fix addressing an issue where backends were unintentionally loading the datastores data and attributes repeatedly during writes (GH1798). By Joe Hamman.

  • Compatibility fixes to plotting module for NumPy 1.14 and pandas 0.22 (GH1813). By Joe Hamman.

  • Bug fix in encoding coordinates with {'_FillValue': None} in netCDF metadata (GH1865). By Chris Roth.

  • Fix indexing with lists for arrays loaded from netCDF files with engine='h5netcdf (GH1864). By Stephan Hoyer.

  • Corrected a bug with incorrect coordinates for non-georeferenced geotiff files (GH1686). Internally, we now use the rasterio coordinate transform tool instead of doing the computations ourselves. A parse_coordinates kwarg has been added to open_rasterio() (set to True per default). By Fabien Maussion.

  • The colors of discrete colormaps are now the same regardless if seaborn is installed or not (GH1896). By Fabien Maussion.

  • Fixed dtype promotion rules in where() and concat() to match pandas (GH1847). A combination of strings/numbers or unicode/bytes now promote to object dtype, instead of strings or unicode. By Stephan Hoyer.

  • Fixed bug where isnull() was loading data stored as dask arrays (GH1937). By Joe Hamman.

v0.10.0 (20 November 2017)#

This is a major release that includes bug fixes, new features and a few backwards incompatible changes. Highlights include:

  • Indexing now supports broadcasting over dimensions, similar to NumPy’s vectorized indexing (but better!).

  • resample() has a new groupby-like API like pandas.

  • apply_ufunc() facilitates wrapping and parallelizing functions written for NumPy arrays.

  • Performance improvements, particularly for dask and open_mfdataset().

Breaking changes#

  • xarray now supports a form of vectorized indexing with broadcasting, where the result of indexing depends on dimensions of indexers, e.g., array.sel(x=ind) with ind.dims == ('y',). Alignment between coordinates on indexed and indexing objects is also now enforced. Due to these changes, existing uses of xarray objects to index other xarray objects will break in some cases.

    The new indexing API is much more powerful, supporting outer, diagonal and vectorized indexing in a single interface. The isel_points and sel_points methods are deprecated, since they are now redundant with the isel / sel methods. See Vectorized Indexing for the details (GH1444, GH1436). By Keisuke Fujii and Stephan Hoyer.

  • A new resampling interface to match pandas’ groupby-like API was added to Dataset.resample() and DataArray.resample() (GH1272). Timeseries resampling is fully supported for data with arbitrary dimensions as is both downsampling and upsampling (including linear, quadratic, cubic, and spline interpolation).

    Old syntax:

    In [5]: ds.resample("24H", dim="time", how="max")
Out[5]: 
<xarray.Dataset>
[...]

    New syntax:

    In [6]: ds.resample(time="24H").max()
Out[6]: 
<xarray.Dataset>
[...]

    Note that both versions are currently supported, but using the old syntax will produce a warning encouraging users to adopt the new syntax. By Daniel Rothenberg.

  • Calling repr() or printing xarray objects at the command line or in a Jupyter Notebook will not longer automatically compute dask variables or load data on arrays lazily loaded from disk (GH1522). By Guido Imperiale.

  • Supplying coords as a dictionary to the DataArray constructor without also supplying an explicit dims argument is no longer supported. This behavior was deprecated in version 0.9 but will now raise an error (GH727).

  • Several existing features have been deprecated and will change to new behavior in xarray v0.11. If you use any of them with xarray v0.10, you should see a FutureWarning that describes how to update your code:

    • Dataset.T has been deprecated an alias for Dataset.transpose() (GH1232). In the next major version of xarray, it will provide short- cut lookup for variables or attributes with name 'T'.

    • DataArray.__contains__ (e.g., key in data_array) currently checks for membership in DataArray.coords. In the next major version of xarray, it will check membership in the array data found in DataArray.values instead (GH1267).

    • Direct iteration over and counting a Dataset (e.g., [k for k in ds], ds.keys(), ds.values(), len(ds) and if ds) currently includes all variables, both data and coordinates. For improved usability and consistency with pandas, in the next major version of xarray these will change to only include data variables (GH884). Use ds.variables, ds.data_vars or ds.coords as alternatives.

  • Changes to minimum versions of dependencies:

    • Old numpy < 1.11 and pandas < 0.18 are no longer supported (GH1512). By Keisuke Fujii.

    • The minimum supported version bottleneck has increased to 1.1 (GH1279). By Joe Hamman.

Enhancements#

New functions/methods

Performance improvements

  • concat() was computing variables that aren’t in memory (e.g. dask-based) multiple times; open_mfdataset() was loading them multiple times from disk. Now, both functions will instead load them at most once and, if they do, store them in memory in the concatenated array/dataset (GH1521). By Guido Imperiale.

  • Speed-up (x 100) of xarray.conventions.decode_cf_datetime. By Christian Chwala.

IO related improvements

  • Unicode strings (str on Python 3) are now round-tripped successfully even when written as character arrays (e.g., as netCDF3 files or when using engine='scipy') (GH1638). This is controlled by the _Encoding attribute convention, which is also understood directly by the netCDF4-Python interface. See String encoding for full details. By Stephan Hoyer.

  • Support for data_vars and coords keywords from concat() added to open_mfdataset() (GH438). Using these keyword arguments can significantly reduce memory usage and increase speed. By Oleksandr Huziy.

  • Support for pathlib.Path objects added to open_dataset(), open_mfdataset(), xarray.to_netcdf, and save_mfdataset() (GH799):

    In [10]: from pathlib import Path  # In Python 2, use pathlib2!

In [11]: data_dir = Path("data/")

In [12]: one_file = data_dir / "dta_for_month_01.nc"

In [13]: xr.open_dataset(one_file)
Out[13]: 
<xarray.Dataset>
[...]

    By Willi Rath.

  • You can now explicitly disable any default _FillValue (NaN for floating point values) by passing the encoding {'_FillValue': None} (GH1598). By Stephan Hoyer.

  • More attributes available in attrs dictionary when raster files are opened with open_rasterio(). By Greg Brener.

  • Support for NetCDF files using an _Unsigned attribute to indicate that a a signed integer data type should be interpreted as unsigned bytes (GH1444). By Eric Bruning.

  • Support using an existing, opened netCDF4 Dataset with NetCDF4DataStore. This permits creating an Dataset from a netCDF4 Dataset that has been opened using other means (GH1459). By Ryan May.

  • Changed PydapDataStore to take a Pydap dataset. This permits opening Opendap datasets that require authentication, by instantiating a Pydap dataset with a session object. Also added xarray.backends.PydapDataStore.open() which takes a url and session object (GH1068). By Philip Graae.

  • Support reading and writing unlimited dimensions with h5netcdf (GH1636). By Joe Hamman.

Other improvements

  • Added _ipython_key_completions_ to xarray objects, to enable autocompletion for dictionary-like access in IPython, e.g., ds['tem + tab -> ds['temperature'] (GH1628). By Keisuke Fujii.

  • Support passing keyword arguments to load, compute, and persist methods. Any keyword arguments supplied to these methods are passed on to the corresponding dask function (GH1523). By Joe Hamman.

  • Encoding attributes are now preserved when xarray objects are concatenated. The encoding is copied from the first object (GH1297). By Joe Hamman and Gerrit Holl.

  • Support applying rolling window operations using bottleneck’s moving window functions on data stored as dask arrays (GH1279). By Joe Hamman.

  • Experimental support for the Dask collection interface (GH1674). By Matthew Rocklin.

Bug fixes#

  • Suppress RuntimeWarning issued by numpy for “invalid value comparisons” (e.g. NaN). Xarray now behaves similarly to pandas in its treatment of binary and unary operations on objects with NaNs (GH1657). By Joe Hamman.

  • Unsigned int support for reduce methods with skipna=True (GH1562). By Keisuke Fujii.

  • Fixes to ensure xarray works properly with pandas 0.21:

    By Stephan Hoyer.

  • open_rasterio() method now shifts the rasterio coordinates so that they are centered in each pixel (GH1468). By Greg Brener.

  • rename() method now doesn’t throw errors if some Variable is renamed to the same name as another Variable as long as that other Variable is also renamed (GH1477). This method now does throw when two Variables would end up with the same name after the rename (since one of them would get overwritten in this case). By Prakhar Goel.

  • Fix xarray.testing.assert_allclose() to actually use atol and rtol arguments when called on DataArray objects (GH1488). By Stephan Hoyer.

  • xarray quantile methods now properly raise a TypeError when applied to objects with data stored as dask arrays (GH1529). By Joe Hamman.

  • Fix positional indexing to allow the use of unsigned integers (GH1405). By Joe Hamman and Gerrit Holl.

  • Creating a Dataset now raises MergeError if a coordinate shares a name with a dimension but is comprised of arbitrary dimensions (GH1120). By Joe Hamman.

  • open_rasterio() method now skips rasterio’s crs attribute if its value is None (GH1520). By Leevi Annala.

  • Fix xarray.DataArray.to_netcdf() to return bytes when no path is provided (GH1410). By Joe Hamman.

  • Fix xarray.save_mfdataset() to properly raise an informative error when objects other than Dataset are provided (GH1555). By Joe Hamman.

  • xarray.Dataset.copy() would not preserve the encoding property (GH1586). By Guido Imperiale.

  • xarray.concat() would eagerly load dask variables into memory if the first argument was a numpy variable (GH1588). By Guido Imperiale.

  • Fix bug in to_netcdf() when writing in append mode (GH1215). By Joe Hamman.

  • Fix netCDF4 backend to properly roundtrip the shuffle encoding option (GH1606). By Joe Hamman.

  • Fix bug when using pytest class decorators to skipping certain unittests. The previous behavior unintentionally causing additional tests to be skipped (GH1531). By Joe Hamman.

  • Fix pynio backend for upcoming release of pynio with Python 3 support (GH1611). By Ben Hillman.

  • Fix seaborn import warning for Seaborn versions 0.8 and newer when the apionly module was deprecated. (GH1633). By Joe Hamman.

  • Fix COMPAT: MultiIndex checking is fragile (GH1833). By Florian Pinault.

  • Fix rasterio backend for Rasterio versions 1.0alpha10 and newer. (GH1641). By Chris Holden.

Bug fixes after rc1#

  • Suppress warning in IPython autocompletion, related to the deprecation of .T attributes (GH1675). By Keisuke Fujii.

  • Fix a bug in lazily-indexing netCDF array. (GH1688) By Keisuke Fujii.

  • (Internal bug) MemoryCachedArray now supports the orthogonal indexing. Also made some internal cleanups around array wrappers (GH1429). By Keisuke Fujii.

  • (Internal bug) MemoryCachedArray now always wraps np.ndarray by NumpyIndexingAdapter. (GH1694) By Keisuke Fujii.

  • Fix importing xarray when running Python with -OO (GH1706). By Stephan Hoyer.

  • Saving a netCDF file with a coordinates with a spaces in its names now raises an appropriate warning (GH1689). By Stephan Hoyer.

  • Fix two bugs that were preventing dask arrays from being specified as coordinates in the DataArray constructor (GH1684). By Joe Hamman.

  • Fixed apply_ufunc with dask='parallelized' for scalar arguments (GH1697). By Stephan Hoyer.

  • Fix “Chunksize cannot exceed dimension size” error when writing netCDF4 files loaded from disk (GH1225). By Stephan Hoyer.

  • Validate the shape of coordinates with names matching dimensions in the DataArray constructor (GH1709). By Stephan Hoyer.

  • Raise NotImplementedError when attempting to save a MultiIndex to a netCDF file (GH1547). By Stephan Hoyer.

  • Remove netCDF dependency from rasterio backend tests. By Matti Eskelinen

Bug fixes after rc2#

  • Fixed unexpected behavior in Dataset.set_index() and DataArray.set_index() introduced by pandas 0.21.0. Setting a new index with a single variable resulted in 1-level pandas.MultiIndex instead of a simple pandas.Index (GH1722). By Benoit Bovy.

  • Fixed unexpected memory loading of backend arrays after print. (GH1720). By Keisuke Fujii.

v0.9.6 (8 June 2017)#

This release includes a number of backwards compatible enhancements and bug fixes.

Enhancements#

Bug fixes#

  • Fix error from repeated indexing of datasets loaded from disk (GH1374). By Stephan Hoyer.

  • Fix a bug where .isel_points wrongly assigns unselected coordinate to data_vars. By Keisuke Fujii.

  • Tutorial datasets are now checked against a reference MD5 sum to confirm successful download (GH1392). By Matthew Gidden.

  • DataArray.chunk() now accepts dask specific kwargs like Dataset.chunk() does. By Fabien Maussion.

  • Support for engine='pydap' with recent releases of Pydap (3.2.2+), including on Python 3 (GH1174).

Documentation#

Testing#

  • Fix test suite failure caused by changes to pandas.cut function (GH1386). By Ryan Abernathey.

  • Enhanced tests suite by use of @network decorator, which is controlled via --run-network-tests command line argument to py.test (GH1393). By Matthew Gidden.

v0.9.5 (17 April, 2017)#

Remove an inadvertently introduced print statement.

v0.9.3 (16 April, 2017)#

This minor release includes bug-fixes and backwards compatible enhancements.

Enhancements#

Bug fixes#

  • Fix .where() with drop=True when arguments do not have indexes (GH1350). This bug, introduced in v0.9, resulted in xarray producing incorrect results in some cases. By Stephan Hoyer.

  • Fixed writing to file-like objects with to_netcdf() (GH1320). Stephan Hoyer.

  • Fixed explicitly setting engine='scipy' with to_netcdf when not providing a path (GH1321). Stephan Hoyer.

  • Fixed open_dataarray does not pass properly its parameters to open_dataset (GH1359). Stephan Hoyer.

  • Ensure test suite works when runs from an installed version of xarray (GH1336). Use @pytest.mark.slow instead of a custom flag to mark slow tests. By Stephan Hoyer

v0.9.2 (2 April 2017)#

The minor release includes bug-fixes and backwards compatible enhancements.

Enhancements#

  • rolling on Dataset is now supported (GH859).

  • .rolling() on Dataset is now supported (GH859). By Keisuke Fujii.

  • When bottleneck version 1.1 or later is installed, use bottleneck for rolling var, argmin, argmax, and rank computations. Also, rolling median now accepts a min_periods argument (GH1276). By Joe Hamman.

  • When .plot() is called on a 2D DataArray and only one dimension is specified with x= or y=, the other dimension is now guessed (GH1291). By Vincent Noel.

  • Added new method assign_attrs() to DataArray and Dataset, a chained-method compatible implementation of the dict.update method on attrs (GH1281). By Henry S. Harrison.

  • Added new autoclose=True argument to open_mfdataset() to explicitly close opened files when not in use to prevent occurrence of an OS Error related to too many open files (GH1198). Note, the default is autoclose=False, which is consistent with previous xarray behavior. By Phillip J. Wolfram.

  • The repr() of Dataset and DataArray attributes uses a similar format to coordinates and variables, with vertically aligned entries truncated to fit on a single line (GH1319). Hopefully this will stop people writing data.attrs = {} and discarding metadata in notebooks for the sake of cleaner output. The full metadata is still available as data.attrs. By Zac Hatfield-Dodds.

  • Enhanced tests suite by use of @slow and @flaky decorators, which are controlled via --run-flaky and --skip-slow command line arguments to py.test (GH1336). By Stephan Hoyer and Phillip J. Wolfram.

  • New aggregation on rolling objects count() which providing a rolling count of valid values (GH1138).

Bug fixes#

v0.9.1 (30 January 2017)#

Renamed the “Unindexed dimensions” section in the Dataset and DataArray repr (added in v0.9.0) to “Dimensions without coordinates” (GH1199).

v0.9.0 (25 January 2017)#

This major release includes five months worth of enhancements and bug fixes from 24 contributors, including some significant changes that are not fully backwards compatible. Highlights include:

Breaking changes#

  • Index coordinates for each dimensions are now optional, and no longer created by default GH1017. You can identify such dimensions without coordinates by their appearance in list of “Dimensions without coordinates” in the Dataset or DataArray repr:

    In [14]: xr.Dataset({"foo": (("x", "y"), [[1, 2]])})
Out[14]: 
<xarray.Dataset>
Dimensions:  (x: 1, y: 2)
Dimensions without coordinates: x, y
Data variables:
    foo      (x, y) int64 1 2

    This has a number of implications:

    • align() and reindex() can now error, if dimensions labels are missing and dimensions have different sizes.

    • Because pandas does not support missing indexes, methods such as to_dataframe/from_dataframe and stack/unstack no longer roundtrip faithfully on all inputs. Use reset_index() to remove undesired indexes.

    • Dataset.__delitem__ and drop() no longer delete/drop variables that have dimensions matching a deleted/dropped variable.

    • DataArray.coords.__delitem__ is now allowed on variables matching dimension names.

    • .sel and .loc now handle indexing along a dimension without coordinate labels by doing integer based indexing. See Missing coordinate labels for an example.

    • indexes is no longer guaranteed to include all dimensions names as keys. The new method get_index() has been added to get an index for a dimension guaranteed, falling back to produce a default RangeIndex if necessary.

  • The default behavior of merge is now compat='no_conflicts', so some merges will now succeed in cases that previously raised xarray.MergeError. Set compat='broadcast_equals' to restore the previous default. See Merging with ‘no_conflicts’ for more details.

  • Reading values no longer always caches values in a NumPy array GH1128. Caching of .values on variables read from netCDF files on disk is still the default when open_dataset() is called with cache=True. By Guido Imperiale and Stephan Hoyer.

  • Pickling a Dataset or DataArray linked to a file on disk no longer caches its values into memory before pickling (GH1128). Instead, pickle stores file paths and restores objects by reopening file references. This enables preliminary, experimental use of xarray for opening files with dask.distributed. By Stephan Hoyer.

  • Coordinates used to index a dimension are now loaded eagerly into pandas.Index objects, instead of loading the values lazily. By Guido Imperiale.

  • Automatic levels for 2d plots are now guaranteed to land on vmin and vmax when these kwargs are explicitly provided (GH1191). The automated level selection logic also slightly changed. By Fabien Maussion.

  • DataArray.rename() behavior changed to strictly change the DataArray.name if called with string argument, or strictly change coordinate names if called with dict-like argument. By Markus Gonser.

  • By default to_netcdf() add a _FillValue = NaN attributes to float types. By Frederic Laliberte.

  • repr on DataArray objects uses an shortened display for NumPy array data that is less likely to overflow onto multiple pages (GH1207). By Stephan Hoyer.

  • xarray no longer supports python 3.3, versions of dask prior to v0.9.0, or versions of bottleneck prior to v1.0.

Deprecations#

  • Renamed the Coordinate class from xarray’s low level API to IndexVariable. Variable.to_variable and Variable.to_coord have been renamed to to_base_variable() and to_index_variable().

  • Deprecated supplying coords as a dictionary to the DataArray constructor without also supplying an explicit dims argument. The old behavior encouraged relying on the iteration order of dictionaries, which is a bad practice (GH727).

  • Removed a number of methods deprecated since v0.7.0 or earlier: load_data, vars, drop_vars, dump, dumps and the variables keyword argument to Dataset.

  • Removed the dummy module that enabled import xray.

Enhancements#

Bug fixes#

  • groupby_bins now restores empty bins by default (GH1019). By Ryan Abernathey.

  • Fix issues for dates outside the valid range of pandas timestamps (GH975). By Mathias Hauser.

  • Unstacking produced flipped array after stacking decreasing coordinate values (GH980). By Stephan Hoyer.

  • Setting dtype via the encoding parameter of to_netcdf failed if the encoded dtype was the same as the dtype of the original array (GH873). By Stephan Hoyer.

  • Fix issues with variables where both attributes _FillValue and missing_value are set to NaN (GH997). By Marco Zühlke.

  • .where() and .fillna() now preserve attributes (GH1009). By Fabien Maussion.

  • Applying broadcast() to an xarray object based on the dask backend won’t accidentally convert the array from dask to numpy anymore (GH978). By Guido Imperiale.

  • Dataset.concat() now preserves variables order (GH1027). By Fabien Maussion.

  • Fixed an issue with pcolormesh (GH781). A new infer_intervals keyword gives control on whether the cell intervals should be computed or not. By Fabien Maussion.

  • Grouping over an dimension with non-unique values with groupby gives correct groups. By Stephan Hoyer.

  • Fixed accessing coordinate variables with non-string names from .coords. By Stephan Hoyer.

  • rename() now simultaneously renames the array and any coordinate with the same name, when supplied via a dict (GH1116). By Yves Delley.

  • Fixed sub-optimal performance in certain operations with object arrays (GH1121). By Yves Delley.

  • Fix .groupby(group) when group has datetime dtype (GH1132). By Jonas Sølvsteen.

  • Fixed a bug with facetgrid (the norm keyword was ignored, GH1159). By Fabien Maussion.

  • Resolved a concurrency bug that could cause Python to crash when simultaneously reading and writing netCDF4 files with dask (GH1172). By Stephan Hoyer.

  • Fix to make .copy() actually copy dask arrays, which will be relevant for future releases of dask in which dask arrays will be mutable (GH1180). By Stephan Hoyer.

  • Fix opening NetCDF files with multi-dimensional time variables (GH1229). By Stephan Hoyer.

Performance improvements#

  • xarray.Dataset.isel_points and xarray.Dataset.sel_points now use vectorised indexing in numpy and dask (GH1161), which can result in several orders of magnitude speedup. By Jonathan Chambers.

v0.8.2 (18 August 2016)#

This release includes a number of bug fixes and minor enhancements.

Breaking changes#

Enhancements#

Bug fixes#

  • Ensure xarray works with h5netcdf v0.3.0 for arrays with dtype=str (GH953). By Stephan Hoyer.

  • Dataset.__dir__() (i.e. the method python calls to get autocomplete options) failed if one of the dataset’s keys was not a string (GH852). By Maximilian Roos.

  • Dataset constructor can now take arbitrary objects as values (GH647). By Maximilian Roos.

  • Clarified copy argument for reindex() and align(), which now consistently always return new xarray objects (GH927).

  • Fix open_mfdataset with engine='pynio' (GH936). By Stephan Hoyer.

  • groupby_bins sorted bin labels as strings (GH952). By Stephan Hoyer.

  • Fix bug introduced by v0.8.0 that broke assignment to datasets when both the left and right side have the same non-unique index values (GH956).

v0.8.1 (5 August 2016)#

Bug fixes#

  • Fix bug in v0.8.0 that broke assignment to Datasets with non-unique indexes (GH943). By Stephan Hoyer.

v0.8.0 (2 August 2016)#

This release includes four months of new features and bug fixes, including several breaking changes.

Breaking changes#

  • Dropped support for Python 2.6 (GH855).

  • Indexing on multi-index now drop levels, which is consistent with pandas. It also changes the name of the dimension / coordinate when the multi-index is reduced to a single index (GH802).

  • Contour plots no longer add a colorbar per default (GH866). Filled contour plots are unchanged.

  • DataArray.values and .data now always returns an NumPy array-like object, even for 0-dimensional arrays with object dtype (GH867). Previously, .values returned native Python objects in such cases. To convert the values of scalar arrays to Python objects, use the .item() method.

Enhancements#

Bug fixes#

  • Attributes were being retained by default for some resampling operations when they should not. With the keep_attrs=False option, they will no longer be retained by default. This may be backwards-incompatible with some scripts, but the attributes may be kept by adding the keep_attrs=True option. By Jeremy McGibbon.

  • Concatenating xarray objects along an axis with a MultiIndex or PeriodIndex preserves the nature of the index (GH875). By Stephan Hoyer.

  • Fixed bug in arithmetic operations on DataArray objects whose dimensions are numpy structured arrays or recarrays GH861, GH837. By Maciek Swat.

  • decode_cf_timedelta now accepts arrays with ndim >1 (GH842).

    This fixes issue GH665. Filipe Fernandes.

  • Fix a bug where xarray.ufuncs that take two arguments would incorrectly use to numpy functions instead of dask.array functions (GH876). By Stephan Hoyer.

  • Support for pickling functions from xarray.ufuncs (GH901). By Stephan Hoyer.

  • Variable.copy(deep=True) no longer converts MultiIndex into a base Index (GH769). By Benoit Bovy.

  • Fixes for groupby on dimensions with a multi-index (GH867). By Stephan Hoyer.

  • Fix printing datasets with unicode attributes on Python 2 (GH892). By Stephan Hoyer.

  • Fixed incorrect test for dask version (GH891). By Stephan Hoyer.

  • Fixed dim argument for isel_points/sel_points when a pandas.Index is passed. By Stephan Hoyer.

  • contour() now plots the correct number of contours (GH866). By Fabien Maussion.

v0.7.2 (13 March 2016)#

This release includes two new, entirely backwards compatible features and several bug fixes.

Enhancements#

  • New DataArray method DataArray.dot() for calculating the dot product of two DataArrays along shared dimensions. By Dean Pospisil.

  • Rolling window operations on DataArray objects are now supported via a new DataArray.rolling() method. For example:

    In [15]: import xarray as xr
   ....: import numpy as np
   ....: 

In [16]: arr = xr.DataArray(np.arange(0, 7.5, 0.5).reshape(3, 5), dims=("x", "y"))

In [17]: arr
Out[17]: 
<xarray.DataArray (x: 3, y: 5)>
array([[ 0. ,  0.5,  1. ,  1.5,  2. ],
       [ 2.5,  3. ,  3.5,  4. ,  4.5],
       [ 5. ,  5.5,  6. ,  6.5,  7. ]])
Coordinates:
  * x        (x) int64 0 1 2
  * y        (y) int64 0 1 2 3 4

In [18]: arr.rolling(y=3, min_periods=2).mean()
Out[18]: 
<xarray.DataArray (x: 3, y: 5)>
array([[  nan,  0.25,  0.5 ,  1.  ,  1.5 ],
       [  nan,  2.75,  3.  ,  3.5 ,  4.  ],
       [  nan,  5.25,  5.5 ,  6.  ,  6.5 ]])
Coordinates:
  * x        (x) int64 0 1 2
  * y        (y) int64 0 1 2 3 4

    See Rolling window operations for more details. By Joe Hamman.

Bug fixes#

  • Fixed an issue where plots using pcolormesh and Cartopy axes were being distorted by the inference of the axis interval breaks. This change chooses not to modify the coordinate variables when the axes have the attribute projection, allowing Cartopy to handle the extent of pcolormesh plots (GH781). By Joe Hamman.

  • 2D plots now better handle additional coordinates which are not DataArray dimensions (GH788). By Fabien Maussion.

v0.7.1 (16 February 2016)#

This is a bug fix release that includes two small, backwards compatible enhancements. We recommend that all users upgrade.

Enhancements#

  • Numerical operations now return empty objects on no overlapping labels rather than raising ValueError (GH739).

  • Series is now supported as valid input to the Dataset constructor (GH740).

Bug fixes#

  • Restore checks for shape consistency between data and coordinates in the DataArray constructor (GH758).

  • Single dimension variables no longer transpose as part of a broader .transpose. This behavior was causing pandas.PeriodIndex dimensions to lose their type (GH749)

  • Dataset labels remain as their native type on .to_dataset. Previously they were coerced to strings (GH745)

  • Fixed a bug where replacing a DataArray index coordinate would improperly align the coordinate (GH725).

  • DataArray.reindex_like now maintains the dtype of complex numbers when reindexing leads to NaN values (GH738).

  • Dataset.rename and DataArray.rename support the old and new names being the same (GH724).

  • Fix from_dataframe() for DataFrames with Categorical column and a MultiIndex index (GH737).

  • Fixes to ensure xarray works properly after the upcoming pandas v0.18 and NumPy v1.11 releases.

Acknowledgments#

The following individuals contributed to this release:

  • Edward Richards

  • Maximilian Roos

  • Rafael Guedes

  • Spencer Hill

  • Stephan Hoyer

v0.7.0 (21 January 2016)#

This major release includes redesign of DataArray internals, as well as new methods for reshaping, rolling and shifting data. It includes preliminary support for pandas.MultiIndex, as well as a number of other features and bug fixes, several of which offer improved compatibility with pandas.

New name#

The project formerly known as “xray” is now “xarray”, pronounced “x-array”! This avoids a namespace conflict with the entire field of x-ray science. Renaming our project seemed like the right thing to do, especially because some scientists who work with actual x-rays are interested in using this project in their work. Thanks for your understanding and patience in this transition. You can now find our documentation and code repository at new URLs:

To ease the transition, we have simultaneously released v0.7.0 of both xray and xarray on the Python Package Index. These packages are identical. For now, import xray still works, except it issues a deprecation warning. This will be the last xray release. Going forward, we recommend switching your import statements to import xarray as xr.

Breaking changes#

  • The internal data model used by xray.DataArray has been rewritten to fix several outstanding issues (GH367, GH634, this stackoverflow report). Internally, DataArray is now implemented in terms of ._variable and ._coords attributes instead of holding variables in a Dataset object.

    This refactor ensures that if a DataArray has the same name as one of its coordinates, the array and the coordinate no longer share the same data.

    In practice, this means that creating a DataArray with the same name as one of its dimensions no longer automatically uses that array to label the corresponding coordinate. You will now need to provide coordinate labels explicitly. Here’s the old behavior:

    In [19]: xray.DataArray([4, 5, 6], dims="x", name="x")
Out[19]: 
<xray.DataArray 'x' (x: 3)>
array([4, 5, 6])
Coordinates:
  * x        (x) int64 4 5 6

    and the new behavior (compare the values of the x coordinate):

    In [20]: xray.DataArray([4, 5, 6], dims="x", name="x")
Out[20]: 
<xray.DataArray 'x' (x: 3)>
array([4, 5, 6])
Coordinates:
  * x        (x) int64 0 1 2

  • It is no longer possible to convert a DataArray to a Dataset with xray.DataArray.to_dataset if it is unnamed. This will now raise ValueError. If the array is unnamed, you need to supply the name argument.

Enhancements#

  • Basic support for MultiIndex coordinates on xray objects, including indexing, stack() and unstack():

    In [21]: df = pd.DataFrame({"foo": range(3), "x": ["a", "b", "b"], "y": [0, 0, 1]})

In [22]: s = df.set_index(["x", "y"])["foo"]

In [23]: arr = xray.DataArray(s, dims="z")

In [24]: arr
Out[24]: 
<xray.DataArray 'foo' (z: 3)>
array([0, 1, 2])
Coordinates:
  * z        (z) object ('a', 0) ('b', 0) ('b', 1)

In [25]: arr.indexes["z"]
Out[25]: 
MultiIndex(levels=[[u'a', u'b'], [0, 1]],
           labels=[[0, 1, 1], [0, 0, 1]],
           names=[u'x', u'y'])

In [26]: arr.unstack("z")
Out[26]: 
<xray.DataArray 'foo' (x: 2, y: 2)>
array([[  0.,  nan],
       [  1.,   2.]])
Coordinates:
  * x        (x) object 'a' 'b'
  * y        (y) int64 0 1

In [27]: arr.unstack("z").stack(z=("x", "y"))
Out[27]: 
<xray.DataArray 'foo' (z: 4)>
array([  0.,  nan,   1.,   2.])
Coordinates:
  * z        (z) object ('a', 0) ('a', 1) ('b', 0) ('b', 1)

    See Stack and unstack for more details.

    Warning

    xray’s MultiIndex support is still experimental, and we have a long to- do list of desired additions (GH719), including better display of multi-index levels when printing a Dataset, and support for saving datasets with a MultiIndex to a netCDF file. User contributions in this area would be greatly appreciated.

  • Support for reading GRIB, HDF4 and other file formats via PyNIO.

  • Better error message when a variable is supplied with the same name as one of its dimensions.

  • Plotting: more control on colormap parameters (GH642). vmin and vmax will not be silently ignored anymore. Setting center=False prevents automatic selection of a divergent colormap.

  • New xray.Dataset.shift and xray.Dataset.roll methods for shifting/rotating datasets or arrays along a dimension:

    In [28]: array = xray.DataArray([5, 6, 7, 8], dims="x")

In [29]: array.shift(x=2)
Out[29]: 
<xarray.DataArray (x: 4)> Size: 32B
array([nan, nan,  5.,  6.])
Dimensions without coordinates: x

In [30]: array.roll(x=2)
Out[30]: 
<xarray.DataArray (x: 4)> Size: 32B
array([7, 8, 5, 6])
Dimensions without coordinates: x

    Notice that shift moves data independently of coordinates, but roll moves both data and coordinates.

  • Assigning a pandas object directly as a Dataset variable is now permitted. Its index names correspond to the dims of the Dataset, and its data is aligned.

  • Passing a pandas.DataFrame or pandas.Panel to a Dataset constructor is now permitted.

  • New function xray.broadcast for explicitly broadcasting DataArray and Dataset objects against each other. For example:

    In [31]: a = xray.DataArray([1, 2, 3], dims="x")

In [32]: b = xray.DataArray([5, 6], dims="y")

In [33]: a
Out[33]: 
<xarray.DataArray (x: 3)> Size: 24B
array([1, 2, 3])
Dimensions without coordinates: x

In [34]: b
Out[34]: 
<xarray.DataArray (y: 2)> Size: 16B
array([5, 6])
Dimensions without coordinates: y

In [35]: a2, b2 = xray.broadcast(a, b)

In [36]: a2
Out[36]: 
<xarray.DataArray (x: 3, y: 2)> Size: 48B
array([[1, 1],
       [2, 2],
       [3, 3]])
Dimensions without coordinates: x, y

In [37]: b2
Out[37]: 
<xarray.DataArray (x: 3, y: 2)> Size: 48B
array([[5, 6],
       [5, 6],
       [5, 6]])
Dimensions without coordinates: x, y

Bug fixes#

  • Fixes for several issues found on DataArray objects with the same name as one of their coordinates (see Breaking changes for more details).

  • DataArray.to_masked_array always returns masked array with mask being an array (not a scalar value) (GH684)

  • Allows for (imperfect) repr of Coords when underlying index is PeriodIndex (GH645).

  • Fixes for several issues found on DataArray objects with the same name as one of their coordinates (see Breaking changes for more details).

  • Attempting to assign a Dataset or DataArray variable/attribute using attribute-style syntax (e.g., ds.foo = 42) now raises an error rather than silently failing (GH656, GH714).

  • You can now pass pandas objects with non-numpy dtypes (e.g., categorical or datetime64 with a timezone) into xray without an error (GH716).

Acknowledgments#

The following individuals contributed to this release:

  • Antony Lee

  • Fabien Maussion

  • Joe Hamman

  • Maximilian Roos

  • Stephan Hoyer

  • Takeshi Kanmae

  • femtotrader

v0.6.1 (21 October 2015)#

This release contains a number of bug and compatibility fixes, as well as enhancements to plotting, indexing and writing files to disk.

Note that the minimum required version of dask for use with xray is now version 0.6.

API Changes#

  • The handling of colormaps and discrete color lists for 2D plots in xray.DataArray.plot was changed to provide more compatibility with matplotlib’s contour and contourf functions (GH538). Now discrete lists of colors should be specified using colors keyword, rather than cmap.

Enhancements#

  • Faceted plotting through xray.plot.FacetGrid and the xray.plot.plot method. See Faceting for more details and examples.

  • xray.Dataset.sel and xray.Dataset.reindex now support the tolerance argument for controlling nearest-neighbor selection (GH629):

    In [38]: array = xray.DataArray([1, 2, 3], dims="x")

In [39]: array.reindex(x=[0.9, 1.5], method="nearest", tolerance=0.2)
Out[39]: 
<xray.DataArray (x: 2)>
array([  2.,  nan])
Coordinates:
  * x        (x) float64 0.9 1.5

    This feature requires pandas v0.17 or newer.

  • New encoding argument in xray.Dataset.to_netcdf for writing netCDF files with compression, as described in the new documentation section on Writing encoded data.

  • Add xray.Dataset.real and xray.Dataset.imag attributes to Dataset and DataArray (GH553).

  • More informative error message with xray.Dataset.from_dataframe if the frame has duplicate columns.

  • xray now uses deterministic names for dask arrays it creates or opens from disk. This allows xray users to take advantage of dask’s nascent support for caching intermediate computation results. See GH555 for an example.

Bug fixes#

  • Forwards compatibility with the latest pandas release (v0.17.0). We were using some internal pandas routines for datetime conversion, which unfortunately have now changed upstream (GH569).

  • Aggregation functions now correctly skip NaN for data for complex128 dtype (GH554).

  • Fixed indexing 0d arrays with unicode dtype (GH568).

  • xray.DataArray.name and Dataset keys must be a string or None to be written to netCDF (GH533).

  • xray.DataArray.where now uses dask instead of numpy if either the array or other is a dask array. Previously, if other was a numpy array the method was evaluated eagerly.

  • Global attributes are now handled more consistently when loading remote datasets using engine='pydap' (GH574).

  • It is now possible to assign to the .data attribute of DataArray objects.