v0.9.1

v0.9.1 (2024-06-04)

Contributors to this version: Pascal Bourgault (@aulemahal), Trevor James Smith (@Zeitsperre), Juliette Lavoie (@juliettelavoie).

Breaking changes

• xscen now uses a src layout in lieu of a flat layout. (PR/407).

Bug fixes

• Fixed defaults for xr_combine_kwargs in extract_dataset (PR/402).
• Fixed bug with xs.utils.update_attr(GH/404, PR/405).
• Fixed template 1 bugs due to changes in dependencies. ( PR/405).

Internal changes

• cartopy has been pinned above version '0.23.0' in order to address a licensing issue. (PR/403).
• The cookiecutter template has been updated to the latest commit via cruft. (PR/407).
  • GitHub Workflows now point to commits rather than tags.
  • Dependabot will now only update on a monthly schedule.
  • Dependencies have been updated and synchronized.
  • CHANGES.rst is now CHANGELOG.rst (see: KeepAChangelog).
  • The CODE_OF_CONDUCT.rst file adapted to Contributor Covenant v2.1 <https://www.contributor-covenant.org/version/2/1/code_of_conduct/>_.
  • Maintainer-specific directions are now found under releasing.rst

v0.9.0

v0.9.0 (2024-05-07)

Contributors to this version: Trevor James Smith (@Zeitsperre), Pascal Bourgault (@aulemahal), Gabriel Rondeau-Genesse (@RondeauG), Juliette Lavoie (@juliettelavoie), Marco Braun (@vindelico).

New features and enhancements

• xs.reduce_ensemble will now call xclim.ensembles.create_ensemble and xclim.ensembles.make_critera if required. (PR/386).

Breaking changes

• Removed support for the old instances of the region argument in spatial_mean, extract_dataset, and subset. (PR/367).
• Removed xscen.extract.clisops_subset. (PR/367).
• dtr (the function) was renamed to dtr_from_minmax to avoid confusion with the dtr variable. (PR/372).
• The xscen.reduce module has been abandoned. (PR/386).
  • build_reduction_data has been made redundant by xclim.ensembles.make_critera and will be removed in a future release.
  • xscen.reduce.reduce_ensemble has been moved to xscen.ensembles.reduce_ensemble, as a module was no longer necessary.

Internal changes

• Modified xscen.utils.change_unit to always adopt the name from the variables_and_units dictionary if the physical units are equal but their names are not (ex. degC <-> ˚C) (PR/373).
• Updated the cookiecutter template to the latest version. (PR/358):
  • Addresses a handful of misconfigurations in the GitHub Workflows.
  • Added a few free grep-based hooks for finding unwanted artifacts in the code base.
  • Updated ruff to v0.2.0 and black to v24.2.0.
• Added more tests. (PR/366, PR/367, PR/372).
• Refactored xs.spatial.subset into smaller functions. (PR/367).
• An encoding argument was added to xs.config.load_config. (PR/370).
• Various small fixes to the code to address FutureWarnings. (PR/380).
• xs.spatial.subset will try to guess CF coordinate if it can't find "latitude" or "longitude" in ds.cf. (PR/384).
• xs.extract_dataset and xs.DataCatalog.to_dataset will now default to opening datasets with option chunks={}, which tries to respect chunking on disk. (PR/398, GH/368).

Bug fixes

• Fix unstack_dates for the new frequency syntax introduced by pandas v2.2. (PR/359).
• subset_warming_level will not return partial subsets if the warming level is reached at the end of the timeseries. (GH/360, PR/359).
• Loading of training in adjust is now done outside of the periods loop. (PR/366).
• Fixed bug for adding the preprocessing attributes inside the adjust function. (PR/366).
• Fixed a bug to accept group = False in adjust function. (PR/366).
• creep_weights now correctly handles the case where the grid is small, n is large, and mode=wrap. (GH/367).
• Fixed a bug in tasmin_from_dtr and tasmax_from_dtr, when dtr units differed from tasmin/max. (PR/372).
• Fixed a bug where the requested chunking would be ignored when saving a dataset (PR/379).
• The missing value check in health_checks will no longer crasg if a variable has no time dimension. (PR/382).

v0.8.3

Contributors to this version: Juliette Lavoie (@juliettelavoie), Trevor James Smith (@Zeitsperre), Gabriel Rondeau-Genesse (@RondeauG), Pascal Bourgault (@aulemahal).

Announcements

• xscen now has a security disclosure policy. (PR/353).
• Various frequency-related changes to match the new pandas naming conventions. (PR/351).

Internal changes

• Added tests for diagnostics. (PR/352).
• Added a SECURITY.md file to the repository and the documentation. (PR/353).
• Added tox modifier for testing builds against the main development branch of xclim. (PR/351, PR/355).
• Added a requirements_upstream.txt file to the repository to track the development branches of relevant dependencies. (PR/355).
• Added a dedicated GitHub Workflow to evaluate compatibility with upstream dependencies. (PR/355).

Breaking changes

• xscen now requires pandas >= 2.2 and xclim >= 0.48.2. (PR/351).
• Functions that output a dict with keys as xrfreq (such as extract_dataset, compute_indicators) will now return the new nomenclature (e.g. "YS-JAN" instead of "AS-JAN"). (PR/351).
• Going from xrfreq to frequencies or timedeltas will still work, but the opposite (frequency --> xrfreq/timedelta) will now only result in the new pandas nomenclature. (PR/351).

v0.8.2

v0.8.2 (2024-02-12)

Contributors to this version: Trevor James Smith (@Zeitsperre), Pascal Bourgault (@aulemahal)

New features and enhancements

• Added a new argument indicators_kw to xs.ensembles.build_partition_data. (PR/315).
• xscen is Semantic Versioning 2.0.0 <https://semver.org/spec/v2.0.0.html>_ compliant. (PR/319).
• xesmf made an optional dependency, making xscen easier to install with pip. (PR/337).

Internal changes

• Granular permissions and dependency scanning actions have been added to all GitHub CI Workflows. (PR/313).
• Updated the list of dependencies to add missing requirements. (PR/314).
• The cookiecutter template has been updated to the latest commit via cruft. (PR/319):
  • actions-versions-updater.yml has been replaced with Dependabot (it's just better).
  • The OpenSSF scorecard.yml workflow has been added to the GitHub workflows to evaluate package security.
  • Code formatting tools (black, blackdoc, isort) are now hard-pinned. These need to be kept in sync with changes from pre-commit. (Dependabot should perform this task automatically.)
  • The versioning system has been updated to follow the Semantic Versioning 2.0.0 standard.

v0.8.0

Changelog

v0.8.0 (2024-01-16)

Contributors to this version: Gabriel Rondeau-Genesse (@RondeauG), Pascal Bourgault (@aulemahal), Juliette Lavoie (@juliettelavoie), Sarah-Claude Bourdeau-Goulet (@sarahclaude), Trevor James Smith (@Zeitsperre), Marco Braun (@vindelico).

Announcements

• xscen now adheres to PEPs 517/518/621 using the setuptools and setuptools-scm backend for building and packaging. (PR/292).

New features and enhancements

• New function xscen.indicators.select_inds_for_avail_vars to filter the indicators that can be calculated with the variables available in a xarray.Dataset. (PR/291).
• Replaced aggregation function climatological_mean() with climatological_op() offering more types of operations to aggregate over climatological periods. (PR/290)
• Added the ability to search for simulations that reach a given warming level. (PR/251).
• xs.spatial_mean now accepts the region="global" keyword to perform a global average (GH/94, PR/260).
• xs.spatial_mean with method='xESMF' will also automatically segmentize polygons (down to a 1° resolution) to ensure a correct average (PR/260).
• Added documentation for require_all_on in search_data_catalogs. (PR/263).
• xs.save_to_table and xs.io.to_table to transform datasets and arrays to DataFrames, but with support for multi-columns, multi-sheets and localized table of content generation.
• Better xs.extract.resample : support for weighted resampling operations when starting with frequencies coarser than daily and missing timesteps/values handling. (GH/80, GH/93, PR/265).
• New argument attribute_weights to generate_weights to allow for custom weights. (PR/252).
• xs.io.round_bits to round floating point variable up to a number of bits, allowing for a better compression. This can be combined with the saving step through argument "bitround" of save_to_netcdf and save_to_zarr. (PR/266).
• Added annual global tas timeseries for CMIP6's models CMCC-ESM2 (ssp245, ssp370, ssp585), EC-Earth3-CC (ssp245, ssp585), KACE-1-0-G (ssp245, ssp370, ssp585) and TaiESM1 (ssp245, ssp370). Moved global tas database to a netCDF file. (GH/268, PR/270).
• Implemented support for multiple levels and models in xs.subset_warming_level. Better support for DataArray and DataFrame in xs.get_warming_level. (PR/270).
• Added the ability to directly provide an ensemble dataset to xs.ensemble_stats. (PR/299).
• Added support in xs.ensemble_stats for the new robustness-related functions available in xclim. (PR/299).
• New function xs.ensembles.get_partition_input (PR/289).

Breaking changes

• climatological_mean() has been replaced with climatological_op() and will be abandoned in a future version. (PR/290)
• experiment_weights argument in generate_weights was renamed to balance_experiments. (PR/252).
• New argument attribute_weights to generate_weights to allow for custom weights. (PR/252).
• For a sequence of models, the output of xs.get_warming_level is now a list. Revert to a dictionary with output='selected' (PR/270).
• The global average temperature database is now a netCDF, custom databases must follow the same format (PR/270).

Bug fixes

• Fixed a bug in xs.search_data_catalogs when searching for fixed fields and specific experiments/members. (PR/251).
• Fixed a bug in the documentation build configuration that prevented stable/latest and tagged documentation builds from resolving on ReadTheDocs. (PR/256).
• Fixed get_warming_level to avoid incomplete matches. (PR/269).
• search_data_catalogs now eliminates anything that matches any entry in exclusions. (GH/275, PR/280).
• Fixed a bug in xs.scripting.save_and_update where build_path_kwargs was ignored when trying to guess the file format. (PR/282).
• Add a warning to xs.extract._dispatch_historical_to_future. (GH/286, PR/287).
• Modify use_cftime for the calendar conversion in to_dataset. (GH/303, PR/289).

Internal changes

• Continued work on adding tests. (PR/251).
• Fixed pre-commit's pretty-format-json hook so that it ignores notebooks. (PR/254).
• Fixed the labeler so docs/CI isn't automatically added for contributions by new collaborators. (PR/254).
• Made it so that tests are no longer treated as an installable package. (PR/248).
• Renamed the pytest marker from requires_docs to requires_netcdf. (PR/248).
• Included the documentation in the source distribution, while excluding the NetCDF files. (PR/248).
• Reduced the size of the files in /docs/notebooks/samples and changed the notebooks and tests accordingly. (GH/247, PR/248).
• Added a new xscen.testing module with the datablock_3d function previously located in /tests/conftest.py. (PR/248).
• New function xscen.testing.fake_data to generate fake data for testing. (PR/248).
• xESMF 0.8 Regridder and SpatialAverager argument out_chunks is now accepted by xs.regrid_dataset and xs.spatial_mean. (PR/260).
• Testing, Packaging, and CI adjustments. (PR/274):
  • xscen builds now install in a tox environment with conda-provided ESMF in GitHub Workflows.
  • tox now offers a method for installing esmpy from a tag/branch (via ESMF_VERSION environment variable).
  • $ make translate is now called on ReadTheDocs and within tox.
  • Linters are now called by order of most common failures first, to speed up the CI.
  • Manifest.in is much more specific about what is installed.
  • Re-adds a dev recipe to the setup.py.
• Multiple improvements to the docstrings and type annotations. (PR/282).
• pip check in conda builds in GitHub workflows have been temporarily set to always pass. (PR/288).
• The cookiecutter template has been updated to the latest commit via cruft. (PR/292):
  • setup.py has been mostly hollowed-out, save for the babel-related translation function.
  • pyproject.toml has been added, with most package configurations migrated into it.
  • HISTORY.rst has been renamed to CHANGES.rst.
  • actions-version-updater.yml has been added to automate the versioning of the package.
  • pre-commit hooks have been updated to the latest versions; check-toml and toml-sort have been added to cleanup the pyproject.toml file, and check-json-schema has been added to ensure GitHub and ReadTheDocs workflow files are valid.
  • ruff has been added to the linting tools to replace most flake8 and pydocstyle verifications.
  • tox builds are more pure Python environment/PyPI-friendly.
  • xscen now uses Trusted Publishing for TestPyPI and PyPI uploads.
• Linting checks now examine the testing folder, function complexity, and alphabetical order of __all__ lists. (PR/292).
• publish_release_notes now uses better logic for finding and reformatting the CHANGES.rst file. (PR/292).
• bump2version version-bumping utility was replaced by bump-my-version. (PR/292).
• Documentation build checks no longer fail due to broken external links; Notebooks are now nested and numbered. (PR/304).

v0.7.1

Bug fix release to update the dependencies and clean the way for a working conda-forge package!

Also, a bug fix in aggregate.spatial_mean for method= 'cos-lat' and the case of a rectilinear grid.

v0.7.0

v0.7.0 (2023-08-22)

Contributors to this version: Gabriel Rondeau-Genesse (@RondeauG), Pascal Bourgault (@aulemahal), Trevor James Smith (@Zeitsperre), Juliette Lavoie (@juliettelavoie), Marco Braun (@vindelico).

Announcements

• Dropped support for Python 3.8, added support for 3.11. (PR/199, PR/222).
• xscen is now available on conda-forge <https://anaconda.org/conda-forge/xscen>_, and can be installed with conda install -c conda-forge xscen. (PR/241)

New features and enhancements

• xscen now tracks code coverage using coveralls. (PR/187).
• New function get_warming_level to search within the IPCC CMIP global temperatures CSV without requiring data. (GH/208, PR/210).
• File re-structuration from catalogs with xscen.catutils.build_path. (PR/205, PR/237).
• New scripting functions save_and_update and move_and_delete. (PR/214).
• Spatial dimensions can be generalized as X/Y when rechunking and will be mapped to rlon/rlat or lon/lat accordingly. (PR/221).
• New argument var_as_string for get_cat_attrs to return variable names as strings. (PR/233).
• New argument copy for move_and_delete. (PR/233).
• New argument restrict_year for compute_indicators. (PR/233).
• Add more comments in the template. (PR/233, GH/232).
• generate_weights now allows to split weights between experiments, and make them vary along the time/horizon axis. (GH/108, PR/231).
• New independence_level, institution, added to generate_weights. (PR/231).
• Updated produce_horizon so it can accept multiple periods or warming levels. (PR/231, PR/240).
• Add more comments in the template. (PR/233, PR/235, GH/232).
• New function diagnostics.health_checks that can perform multiple checkups on a dataset. (PR/238).

Breaking changes

• Columns date_start and date_end now use a datetime64[ms] dtype. (PR/222).
• The default output of date_parser is now pd.Timestamp (output_dtype='datetime'). (PR/222).
• date_parser(date, end_of_period=True) has time "23:59:59", instead of "23:00". (PR/222, PR/237).
• driving_institution was removed from the "default" xscen columns. (PR/222).
• Folder parsing utilities (parse_directory) moved to xscen.catutils. Signature changed : globpattern removed, dirglob added, new patterns specifications. See doc for all changes. (PR/205).
• compute_indicators now returns all outputs produced by indicators with multiple outputs (such as rain_season). (PR/228).
• In generate_weights, independence_level all was renamed model. (PR/231).
• In response to a bugfix, results for generate_weights(independence_level='GCM') are significantly altered. (GH/230, PR/231).
• Legacy support for stats_kwargs in ensemble_stats was dropped. (PR/231).
• period in produce_horizon has been deprecated and replaced with periods. (PR/231).
• Some automated to_level were updated to reflect more recent changes. (PR/231).
• Removed diagnostics.fix_unphysical_values. (PR/238).

Bug fixes

• Fix bug in unstack_dates with seasonal climatological mean. (GH/202, PR/202).
• Added NotImplemented errors when trying to call climatological_mean and compute_deltas with daily data. (PR/187).
• Minor documentation fixes. (GH/223, PR/225).
• Fixed a bug in unstack_dates where it failed for anything other than seasons. (PR/228).
• cleanup with common_attrs_only now works even when no cat attribute is present in the datasets. (PR/231).

Internal changes

• Removed the pin on xarray's version. (GH/175, PR/199).
• Folder parsing utilities now in pure python, platform independent. New dependency parse. (PR/205).
• Updated ReadTheDocs configuration to prevent --eager installation of xscen (PR/209).
• Implemented a template to be used for unit tests. (PR/187).
• Updated GitHub Actions to remove deprecation warnings. (PR/187).
• Updated the cookiecutter used to generate boilerplate documentation and code via cruft. (PR/212).
• A few changes to subset_warming_level so it doesn't need driving_institution. (PR/215).
• Added more tests. (PR/228).
• In compute_indicators, the logic to manage indicators returning multiple outputs was simplified. (PR/228).

v0.6.0

History

v0.6.0 (2023-05-04)

Contributors to this version: Trevor James Smith (@Zeitsperre), Juliette Lavoie (@juliettelavoie), Pascal Bourgault (@aulemahal), Gabriel Rondeau-Genesse (@RondeauG).

Announcements

• xscen is now offered as a conda package available through Anaconda.org. Refer to the installation documentation for more information. (GH/149, PR/171).
• Deprecation: Release 0.6.0 of xscen will be the last version to support xscen.extract.clisops_subset. Use xscen.spatial.subset instead. (PR/182, PR/184).
• Deprecation: The argument region, used in multiple functions, has been slightly reformatted. Release 0.6.0 of xscen will be the last version to support the old format. (GH/99, GH/101, PR/184).

New features and enhancements

• New 'cos-lat' averaging in spatial_mean. (GH/94, PR/125).
• Support for computing anomalies in compute_deltas. (PR/165).
• Add function diagnostics.measures_improvement_2d. (PR/167).
• Add function regrid.create_bounds_rotated_pole and automatic use in regrid_dataset and spatial_mean. This is temporary, while we wait for a functionning method in cf_xarray. (PR/174, GH/96).
• Add spatial submodule with functions creep_weights and creep_fill for filling NaNs using neighbours. (PR/174).
• Allow passing GeoDataFrame instances in spatial_mean's region argument, not only geospatial file paths. (PR/174).
• Allow searching for periods in catalog.search. (GH/123, PR/170).
• Allow searching and extracting multiple frequencies for a given variable. (GH/168, PR/170).
• New masking feature in extract_dataset. (GH/180, PR/182).
• New function xs.spatial.subset to replace xs.extract.clisops_subset and add method "sel". (GH/180, PR/182).
• Add long_name attribute to diagnostics. ( PR/189).
• Added a new YAML-centric notebook (GH/8, PR/191).
• New utils.standardize_periods to standardize that argument across multiple functions. (GH/87, PR/192).
• New coverage_kwargs argument added to search_data_catalogs to allow modifying the default values of subset_file_coverage. (GH/87, PR/192).

Breaking changes

• 'mean' averaging has been deprecated in spatial_mean. (PR/125).
• 'interp_coord' has been renamed to 'interp_centroid' in spatial_mean. (PR/125).
• The 'datasets' dimension of the output of diagnostics.measures_heatmap is renamed 'realization'. (PR/167).
• _subset_file_coverage was renamed subset_file_coverage and moved to catalog.py to prevent circular imports. (PR/170).
• extract_dataset doesn't fail when a variable is in the dataset, but not variables_and_freqs. (PR/185).
• The argument period, used in multiple function, is now always a single list, while periods is more flexible. (GH/87, PR/192).
• The parameters reference_period and simulation_period of xscen.train and xscen.adjust were renamed period/periods to respect the point above. (GH/87, PR/192).

Bug fixes

• Forbid pandas v1.5.3 in the environment files, as the linux conda build breaks the data catalog parser. (GH/161, PR/162).
• Only return requested variables when using DataCatalog.to_dataset. (PR/163).
• compute_indicators no longer crashes if less than 3 timesteps are produced. (PR/125).
• xarray is temporarily pinned below v2023.3.0 due to an API-breaking change. (GH/175, PR/173).
• `xscen.utils.unstack_fill_nan`` can now handle datasets that have non dimension coordinates. (GH/156, PR/175).
• extract_dataset now skips a simulation way earlier if the frequency doesn't match. (PR/170).
• extract_dataset now correctly tries to extract in reverse timedelta order. (PR/170).
• compute_deltas no longer creates all NaN values if the input dataset is in a non-standard calendar. (PR/188).

Internal changes

• xscen now manages packaging for PyPi and TestPyPI via GitHub workflows. (PR/159).
• Pre-load coordinates in extract.clisops_subset (PR/163).
• Minimal documentation for templates. (PR/163).
• xscen is now indexed in Zenodo, under the ouranos community of projects. (PR/164).
• Added a few relevant Shields to the README.rst. (PR/164).
• Better warning messages in _subset_file_coverage when coverage is insufficient. (PR/125).
• The top-level Makefile now includes a linkcheck recipe, and the ReadTheDocs configuration no longer reinstalls the llvmlite compiler library. (PR/173).
• The checkups on coverage and duplicates can now be skipped in subset_file_coverage. (PR/170).
• Changed the ProjectCatalog docstrings to make it more obvious that it needs to be created empty. (GH/99, PR/184).
• Added parse_config to creep_fill, creep_weights, and reduce_ensemble (PR/191).

v0.5.0

Contributors to this version: Gabriel Rondeau-Genesse (@RondeauG), Juliette Lavoie (@juliettelavoie), Trevor James Smith (@Zeitsperre), Sarah Gammon (@sg2475962) and Pascal Bourgault (@aulemahal).

New features and enhancements

• Possibility of excluding variables read from file from the catalog produced by parse_directory. (PR/107).
• New functions extract.subset_warming_level and aggregate.produce_horizon. (PR/93).
• add round_var to xs.clean_up. (PR/93).
• New "timeout_cleanup" option for save_to_zarr, which removes variables that were in the process of being written when receiving a TimeoutException. (PR/106).
• New scripting.skippable context, allowing the use of CTRL-C to skip code sections. (PR/106).
• Possibility of fields with underscores in the patterns of parse_directory. (PR/111).
• New utils.show_versions function for printing or writing to file the dependency versions of xscen. (GH/109, PR/112).
• Added previously private notebooks to the documentation. (PR/108).
• Notebooks are now tested using pytest with nbval. (PR/108).
• New restrict_warming_level argument for extract.search_data_catalogs to filter dataset that are not in the warming level csv. (GH/105, PR/138).
• Set configuration value programmatically through CONFIG.set. (PR/144).
• New to_dataset method on DataCatalog. The same as to_dask, but exposing more aggregation options. (PR/147).

Breaking changes

• Functions that are called internally can no longer parse the configuration. (PR/133).

Bug fixes

• clean_up now converts the calendar of variables that use "interpolate" in "missing_by_var" at the same time.
  • Hence, when it is a conversion from a 360_day calendar, the random dates are the same for all of the these variables. (GH/102, PR/104).
• properties_and_measures no longer casts month coordinates to string. (PR/106).
• search_data_catalogs no longer crashes if it finds nothing. (GH/42, PR/92).
• Prevented fixed fields from being duplicated during _dispatch_historical_to_future (GH/81, PR/92).
• Added missing parse_config to functions in reduce.py (PR/92).
• Added deepcopy before skipna is popped in spatial_mean (PR/92).
• subset_warming_level now validates that the data exists in the dataset provided (GH/117, PR/119).
• Adapt stack_drop_nan for the newest version of xarray (2022.12.0). (GH/122, PR/126).
• Fix stack_drop_nan not working if intermediate directories don't exist (GH/128).
• Fixed a crash when compute_indicators produced fixed fields (PR/139).

Internal changes

• compute_deltas skips the unstacking step if there is no time dimension and cast object dimensions to string. (PR/9)
• Added the "2sem" frequency to the translations CVs. (PR/111).
• Skip files we can't read in parse_directory. (PR/111).
• Fixed non-numpy-standard Docstrings. (PR/108).
• Added more metadata to package description on PyPI. (PR/108).
• Faster search_data_catalogs and extract_dataset through a faster DataCatalog.unique, date parsing and a rewrite of the ensure_correct_time logic. (PR/127).
• The search_data_catalogs function now accepts str or pathlib.Path variables (in addition to lists of either data type) for performing catalog lookups. (PR/121).
• produce_horizons now supports fixed fields (PR/139).
• Rewrite of unstack_dates for better performance with dask arrays. (PR/144).

v0.4.0

v0.4.0 (2022-09-28)

Contributors to this version: Gabriel Rondeau-Genesse (@RondeauG), Juliette Lavoie (@juliettelavoie), Trevor James Smith (@Zeitsperre) and Pascal Bourgault (@aulemahal).

New features and enhancements

• New functions diagnostics.properties_and_measures, diagnostics.measures_heatmap and diagnostics.measures_improvement. (GH/5, PR/54).
• Add argument resample_methods to xs.extract.resample. (GH/57, PR/57)
• Added a ReadTheDocs configuration to expose public documentation. (GH/65, PR/66).
• xs.utils.stack_drop_nans/ xs.utils.unstack_fill_nan will now format the to_file/coords string to add the domain and the shape. (GH/59, PR/67).
• New unstack_dates function to "extract" seasons or months from a timeseries. (PR/68).
• Better spatial_mean for cases using xESMF and a shapefile with multiple polygons. (PR/68).
• Yet more changes to parse_directory: (PR/68).
  • Better parallelization by merging the finding and name-parsing step in the same dask tree.
  • Allow cvs for the variable columns.
  • Fix parsing the variable names from datasets.
  • Sort the variables in the tuples (for a more consistent output)
• In extract_dataset, add option ensure_correct_time to ensure the time coordinate matches the expected freq. Ex: monthly values given on the 15th day are moved to the 1st, as expected when asking for "MS". (:issue: 53).
• In regrid_dataset: (PR/68).
  • Allow passing skipna to the regridder kwargs.
  • Do not fail for any grid mapping problem, includin if a grid_mapping attribute mentions a variable that doesn't exist.
• Default email sent to the local user. (PR/68).
• Special accelerated pathway for parsing catalogs with all dates within the datetime64[ns] range. (PR/75).
• New functions reduce_ensemble and build_reduction_data to support kkz and kmeans clustering. (GH/4, PR/63).
• ensemble_stats can now loop through multiple statistics, support functions located in xclim.ensembles._robustness, and supports weighted realizations. (PR/63).
• New function ensemble_stats.generate_weights that estimates weights based on simulation metadata. (PR/63).
• New function catalog.unstack_id to reverse-engineer IDs. (PR/63).
• generate_id now accepts Datasets. (PR/63).
• Add rechunk option to properties_and_measures (PR/76).
• Add create argument to ProjectCatalog (GH/11, PR/77).
• Add percentage deltas to compute_deltas (GH/82, PR/90).

Breaking changes

• statistics / stats_kwargs have been changed/eliminated in ensemble_stats, respectively. (PR/63).

Bug fixes

• Add a missing dependencies to the env (pyarrow, for faster string handling in catalogs). (PR/68).
• Allow passing compute=False to save_to_zarr. (PR/68).

Internal changes

• Small bugfixes in aggregate.py. (PR/55, PR/56).
• Default method of xs.extract.resample now depends on frequency. (GH/57, PR/58).
• Bugfix for _restrict_by_resolution with CMIP6 datasets (PR/71).
• More complete check of coverage in _subset_file_coverage. (GH/70, PR/72)
• The code that performs common_attrs_only in ensemble_stats has been moved to clean_up. (PR/63).
• Removed the default to_level in clean_up. (PR/63).
• xscen now has an official logo. (PR/69).
• Use numpy max and min in properties_and_measures (PR/76).
• Cast catalog date_start and date_end to "%4Y-%m-%d %H:00" when writing to disk. (GH/83, PR/79)
• Skip test of coverage on the sum if the list of select files is empty. (PR/79)
• Added missing CMIP variable names in conversions.yml and added the ability to provide a custom file instead (GH/86, PR/88)
• Changed 'allow_conversion' and 'allow_resample' default to False in search_data_catalogs (GH/86, PR/88)