All notable changes to this project will be documented in this file.
The format is based on Keep a Changelog, and this project adheres to Semantic Versioning.
Please add your functional changes to the appropriate section in the PR. Keep it human-readable, your future self will thank you!
0.5.42 (2026-08-13)
- Clarify-timezone-datetime (#723) (05262c2)
- Cut dependency of anemoi-registy on anemoi-datasets (#696) (be8b64b)
- cutout: User-provided pre-computed cutout masks (#627) (448a96c)
- Fix timezone corner case (#720) (87771c9)
0.5.41 (2026-07-27)
- Add consistency check for tabular (#687) (2b4db63)
- Broadcast-to-trajectories-in-open-dataset (#717) (3a14296)
- Create and ship py.typed (#714) (63dba5f)
- Deprecation warning for group by instead of silent ignore (#682) (f997ea3)
- Do not mutate source_config in bufr (#683) (9815ca5)
- Improve memory for trajectories dataset building (#693) (5ecabf0)
- Keep compatibility with zarr 3 (#694) (e2f5f68)
- Remove validation of dates by default (#691) (09d1903)
- Use group_by 1 for trajectories by default (#676) (6f6e964)
- create: Height coord standard name (#707) (943672d)
- Interpolate_frequency now preserves dtype (#710) (9e5218a)
- Negative index support (#703) (c1bbaa1)
- Pass options as kwarg in usage/_open (#690) (eff502e)
- Release rolling average assumption. (#678) (a86df41)
0.5.40 (2026-06-22)
- Add missing dependency, psutil (#670) (e913047)
- Add window and cycle date column option to ODB source (#662) (d388399)
- Extract constants and climatologies from a dataset for use in inference (#628) (55af3ce)
- metadata: Store GRIB paramId in
variables_metadata(#664) (f283884) - Not resolution check for tabular (#672) (8956a2a)
- Option when checking variables compatibility (#669) (1345ba9)
- Support empty data for bufr (#673) (ca78ec0)
- Synthetic gridded dataset (#637) (8cbfc01)
0.5.39 (2026-06-11)
- Analyse-recipe command for workflow automation (#659) (836b6ba)
- Trajectories missing features 3 (#655) (ecd9d85)
- Trajectory select file (#656) (5d507a1)
0.5.38 (2026-06-09)
- Add
create-test-recipeworkflow for test dataset creation and stabilize missing-date CI tests (#639) (8b82da8) - Missing features for trajectories (#646) (62bdbb5)
- Renaming a dataset (#640) (2ae6c89)
- Trajectories datasets (#618) (46fe4b9)
- Trajectories missing features 2 (#652) (139ef75)
- Inspect failure when dataset is not finalised (#648) (1d73c01)
- Missing-dates-efficiency (#649) (a32e7d8)
- Trajectory metadata (#654) (fa92481)
0.5.37 (2026-05-27)
- Add BUFR source (#635) (07cbdd2)
- Add units to datasets (#577) (2545ad9)
- Datasets tendencies overwrite (#633) (460bee4)
- Export json schema (#630) (f40ad9f)
- Fetch multiple chunks in parallel when reading from s3 (#617) (88e8c08)
- Keep compatibility with zarr 3 (#623) (7597d85)
- Pydantic recipes (#575) (0ded992)
0.5.36 (2026-04-22)
- Add CycleIntervalProvider and set_start_step_to_zero patch (#564) (2c8824c)
- Add planetary-computer-multipart source, tests, and docs (#610) (42117db)
- create: Add workaround for missing data at step zero (#565) (9fd4733)
- Fetch files from ecfs if path starts with ec: or ectmp: (#585) (9fb443a)
- Fix issue 569 (#574) (7f4e40a)
- Fix typo with duplicates (#580) (f33333e)
- Make anemoi-datasets agnostic to Zarr version (Optional support Zarr3) (#220) (ab8cd71)
- Observations feature branch (#480) (92d5ac9)
- Open datasets analytics (#576) (561dbd2)
- Remove https test (#608) (048e419)
- create: Repeated-dates (#572) (b73d533)
- Example accumulations section to user current accumulate API (#601) (9434007)
- Fix corner cases (#594) (bdd31ff)
- Fix race condition during build (#593) (66e2070)
- Fix read ahead while building (#611) (6d18e5e)
- Fix weatherbench test (#609) (f434a15)
- grib-index: Support querying float values (#520) (b089cd2)
- Improve MARS request handling for forecast datasets (#562) (f9efe39)
- Make dataset naming function public (#579) (b089bb0)
- Netcdf date/time metadata type should be int (#555) (9937fbe)
- Propagate resolution metadata when using anemoi_dataset source (#614) (784695c)
- Remove duplicate code (#590) (8e54420)
- Remove empty accumulators from accumulation computation (#561) (3bc087d)
- Replace pydantic class Config with ConfigDict (#592) (ce6b2ff)
- Rolling average regression (#587) (04f5b0b)
0.5.35 (2026-03-09)
- Accept only copy from *.zarr to *.zarr, no '/' are allowed (#548) (d584e6e)
- create: Support FDB as base source for accumulate (#550) (506e397)
- Fix corner case when using cutout (#558) (a2594ab)
- Explain option
allow_nans(#547) (00396bb) - Fix accumulation availability specification (#551) (c26a9d8)
0.5.34 (2026-02-17)
- Bump anemoi-transform version (#541) (09de1b2)
- Bump anemoi-transform version to 0.1.25 (#543) (f66c269)
- Utils provenance update (#533) (5a453bd)
0.5.33 (2026-02-09)
0.5.32 (2026-02-06)
- Filter out timespan as key for accumulator (#516) (6fb7f2b)
- max_distance_km: Add new max_distance_km argument to cutout (#530) (a94f3bc)
0.5.31 (2026-02-05)
0.5.30 (2026-02-03)
0.5.29 (2026-01-28)
- Abstracting accumulation (#326) (57a2895)
- Add id and location for point data (#500) (8cf8a10)
- Add xarray patches for changing time dimension and summing variables (#494) (7c44d99)
- apply mask: Apply a boolean mask when opening a dataset (#496) (820eb39)
- '--test' changes grid/number (#386) (71238c6)
- Complement with max distance (#469) (a8666fa)
- Coordinate (multiple) indexing - check indices before comparison (#506) (859ed19)
- Grids for masked datasets, field_shape thinning (#504) (e7b8d58)
- Issue #477 (#478) (42b9e27)
- Pin min version of ruamel-yaml (#514) (e6037b0)
- Update README to reflect project maturity status (#465) (ec8a15e)
- Document the option to specify types when selecting a GRIB message (#466) (8940839)
- Fix interpolate to interpolation key in combining page (#461) (990a5c3)
0.5.28 (2025-10-30)
- Added rolling average on read (#448) (5cef720)
- Fix repeat dates (#458) (16b5c83)
- Refactor dataset create (#379) (c5408da)
- Allow grib source to be used for constant data (#428) (c1d7ff1)
- Broken source imports (#450) (#454) (d6ed174)
- Ensure missing value indicator in not a valid value in accumulat… (#316) (3541a1f)
- S3 access #453 (#459) (2d810e3)
- Xarray-zarr filters (#457) (5f2973f)
0.5.27 (2025-10-06)
- Add 'model_level_number' to coordinate guesser (ecmwf/anemoi-datasets#402) (#403) (fa0b7e8)
- Extend xarray Variable.sel to handle scalar variables (#381) (1306d62)
- Improve conftest types (#395) (bd8143d)
- Remove python 3.9 from pyproject.toml. (#392) (7f3a250)
- sources: Add FDB source (#405) (9222d59)
- Validate dataset (#330) (34d06df)
- Allow multiple options to be provided to anemoi-datasets inspect (#372) (dea680e)
- Anemoi inspect wrong message (#429) (984ccef)
- Better handling of xarray metadata (#322) (4304cb4)
- Cleaning _build when building datasets (#413) (046a7eb)
- Define the missing forwards_subclass_metadata_specific method in the ZipBase class (#377) (2816314)
- Remove config_path in Config (#431) (2bf674f)
- Remove-obsolete-code-and-rename (#423) (ab3da59)
- Save_dataset (#383) (05b8f51)
- Support for sources with already opened datasets (#417) (9bc2d17)
0.5.26 (2025-07-11)
- Add planetary planetary source (#368) (c85337f)
- Extend xarray coordinate guesser to include pressure as a long_name (#335) (bcbffaa)
- Extented complement nearest functionality to support k>1 (#371) (a2c5f92)
- Point data (#358) (78c4f85)
- Relaxed compatibility check for horizontal grid. (#339) (a7dc859)
- Sort xarray coordinate with patch (#356) (638cbdb)
- Use pytest-xdist to run tests in parallel. (#346) (1d656ee)
- Deprecated filters in anemoi datasets (#350) (d08c6f5)
- grib-index: Match against filename, not full path (#374) (e71a952)
- Missing method in xarray flavour (#369) (d0b9588)
- Remove trailing slash from zarr paths. (#362) (9a9bfe5)
0.5.25 (2025-06-11)
- Incremental dataset build tasks called regardless of presence of debug flag in CLI code (#294) (37afc0d)
- Regression in accumulations #354 (#355) (f9769d7)
- Remove 2 layers of build function (#348) (7a904c4)
0.5.24 (2025-05-23)
0.5.23 (2025-05-07)
0.5.22 (2025-05-05)
- add command line tool to check naming conventions (#287) (38a81e2)
- copy datasets from http(s) sources (#307) (cedd9db)
- downloading zip files if needed (#288) (3024099)
- optionally search for zarr files at various location if not found (#281) (8bd174c)
- reset accumulations (#300) (b12666a)
- work on check (#311) (c6a8db9)
- bug in complement (#296) (4e57f7c)
- bugs affecting creation of datasets from xarray-zarr (#299) (09fcb51)
- fix copy (#289) (a79759f)
- Proper indent for parallel additions (#303) (0643fb8)
0.5.21 (2025-04-08)
- grib-index command default (#275) (04a37d8)
- pin version of numcodecs (#285) (d86f317)
- update version of transform (#284) (e72cdde)
- some cleanup (#276) (496b82a)
- time interpolation example (#274) (9d19bd6)
- update documentation (#271) (85ea386)
0.5.20 (2025-03-31)
- interpolate nearest spatial (#260) (e6c9af4)
- new data sources (#258) (708c816)
- save opened anemoi_dataset (#259) (4759dd9)
0.5.19 (2025-03-27)
- broken references in recipes (#269) (bbdf339)
- fix typo (#268) (a12f58f)
- swap meshgrid dimension ordering in xarray grid creation (#249) (938f3c9)
0.5.18 (2025-03-25)
0.5.17 (2025-03-24)
- add coordinate standard_name for rotated pole grid (#192) (5464347)
- better logging for group of dates (#235) (d7fc616)
- fix to_index (#225) (baefd70)
- plugin support (#241) (376ef1c)
- support sub-hourly steps (#188) (7f219e9)
- update test breaking due to noaa external data change (#204) (4b99ea6)
- allow xarrays with single value dimensions (#230) (ee9fbf8)
- create forcings based on xarray-zarr template (again) (#244) (84cb312)
- fix for bug #237 (#247) (de3cab8)
- Use set in computing statistics, faster and use less memory (#209) (e93dbc1)
- add animation (#208) (2af2fd6)
- Docathon 2025 (#234) (fb68b95)
- fix readthedocs (#223) (ce1b44e)
- update doc with eccc-fstd, cdsapi and regrid (#201) (57a53fc)
- update project name (#246) (5ddd4d1)
- use new logo (#211) (76ecf15)
- chore: synced file(s) with ecmwf-actions/reusable-workflows by @DeployDuck in #186
- @DeployDuck made their first contribution in #186
Full Changelog: https://github.com/ecmwf/anemoi-datasets/compare/0.5.15...0.5.16
- Fix metadata serialization handling of numpy.integer (#140)
- Fix negative variance for constant variables (#148)
- Fix cutout slicing of grid dimension (#145)
- Use cKDTree instead of KDTree
- Implement 'complement' feature
- Add ability to patch xarrays (#160)
- Add support of ECCC file formats (fstd)
- Add
use_cdsapi_datasetoption to mars and accumulations
- Call filters from anemoi-transform
- Make test optional when adls is not installed Pull request #110
- Add wz_to_w, orog_to_z, and sum filters (#149)
0.5.8 - 2024-10-26
- Bugfix in
auto_adjust - Fixed precommit CI errors
- Improve tests
- More verbosity
- Add anemoi-transform link to documentation
- Various bug fixes
- Control compatibility check in xy/zip
- Add
mergefeature - Add support for storing
supporting_arraysin checkpoint files - Allow naming of datasets components
- Contributors file (#105)
- Upload with ssh (experimental)
- Remove upstream dependencies from downstream-ci workflow (temporary) (#83)
- ci: pin python versions to 3.9 ... 3.12 for checks (#93)
- Fix
__version__import in init
0.5.7 - 2024-10-09
- Add support to fill missing dates
Allow for unknown CF coordinates - 2024-10-04
- Add
variables_metadataentry in the dataset metadata - Update documentation
- Add
variables_metadataentry in the dataset metadata
- Add
variables_metadataentry in the dataset metadata
0.5.5 - 2024-10-04
- Allow for unknown coordinates when parsing CF input
Add support for (semi-)constant fields - 2024-10-03
- Fix failing zarr lat/lon tests
Bug fixes - 2024-10-01
- Adding the user recipe in the dataset PR #59.
- Add
repeated_datesaction in create.
- Bug fix in create/rename
0.5.0 - Incremental builds and Rescaling - 2024-09-25
- New
rescalekeyword inopen_datasetto change units of variables #36 - Add support for constant fields when creating datasets
- Simplify imports
- Added incremental building of datasets
- Add missing dependency for documentation building
- Fix failing test due to previous merge
- Bug fix when creating dataset from zarr
- Bug fix with area selection in cutout operation
- add paths-ignore to ci workflow
- call provenance less often
- pytest for notebooks
- New
interpolate_frequencykeyword inopen_dataset - CI workflow to update the changelog on release
- adds the reusable cd pypi workflow
- merge strategy for changelog in .gitattributes #25
- adds ci hpc config (#43)
- update CI to reusable workflows for PRs and releases
- Support sub-hourly datasets.
- Change negative variance detection to make it less restrictive
- Fix cutout bug that left some global grid points in the lam part
- Fix bug in computing missing dates in cutout option
0.4.4 Bug fixes
- Bug fix in accumulations() that did not work with webmars
- Bug fix when using hindcasts input (missing dates on leap years)
0.4.0 Minor Release
- earthkit-data replaces climetlab
- climetlab
0.3.0 Minor Release
- hindcast source
- updated documentation
0.2.0 Minor Release
- statistics tendencies
- CubesFilter
0.1.0 Initial Release
- Documentation
- Initial code release for anemoi-datasets: create datasets for data-driven weather-models
- open datasets
- combine datasets