1.13.1.0

2026-04-09

• Read Kerchunk datasets with cfdm.read (#385)
• Read open file handle datasets with cfdm.read (#401)
• Write UGRID datasets with cfdm.write (#271)
• New keyword to cfdm.read: filesystem (#397)
• New keyword parameter to cfdm.Data.compute: persist (#389)
• New function to control the persistence of computed data: cfdm.persist_data (#389)
• Support for HEALPix grids (#370)
• New default backend for netCDF-4 in cfdm.write: h5netcdf-h5py, that allows control of the internal file metadata via the new h5py_options parameter (#386)
• New default backend for netCDF-4 in cfdm.read that allows parallel reading: h5netcdf-pyfive (#375)
• New optional backend for netCDF-3 in cfdm.read that allows parallel reading: netcdf_file (#375)
• Fix bug in cfdm.netcdf_indexer that sometimes caused a failure with a np.newaxis index (#395)
• Fix bug in cfdm.read that wouldn't read non-Zarr and Zarr datasets from the same directory (#391)
• Fix bug in cfdm.write to correctly create unlimited aggregation dimensions that correspond to unlimited normal dimensions (#387)
• Fix bug in cfdm.write when writing identical coordinates that have different formula_terms (#380).
• New dependency: pyfive>=1.1.1
• New dependency: fsspec>=2026.2.0
• Changed dependency: h5netcdf>=1.8.0
• Removed dependency: s3fs

What's Changed

• [Superseded by #383] Fix bug in cfdm.write when writing identical coordinates that have different "formula_terms" by @davidhassell in #381
• Fix bug in cfdm.write when writing identical coordinates that have different "formula_terms" by @davidhassell in #383
• Fix aggregation unlimited netCDF dimensions by @davidhassell in #387
• Enable parallel netCDF dataset reads by @davidhassell in #384
• New 'h5netcdf-h5py' netCDF backend in cfdm.write, and 'h5py_options' parameter by @davidhassell in #388
• Implement HEALPix grids by @davidhassell in #371
• Persist data after computation by @davidhassell in #390
• Fix: Zarr and non-Zarr datasets can't be read from the same directory by @davidhassell in #392
• Add filesystem keyword to cfdm.read for pre-authenticated fsspec backends (part 1) by @Copilot in #393
• Add filesystem keyword to cfdm.read for pre-authenticated fsspec backends (part 2) by @davidhassell in #398
• Fix bug in cfdm.netcdf_indexer that sometimes caused a failure with a np.newaxis index by @davidhassell in #400
• Implementation of UGRID (part 2 of 2): Write by @davidhassell in #372
• Read Kerchunk and file-like object datasets by @davidhassell in #402

New Contributors

• @Copilot made their first contribution in #393

Full Changelog: v1.13.0.0...v1.13.1.0

1.13.0.0

• Write Zarr v3 datasets with cfdm.write, and allow the reading of grouped Zarr v2 and v3 datasets with cfdm.read (#354)
• Read Zarr v2 and v3 datasets that contain a group hierarchy with cfdm.read (#355)
• New function cfdm.dataset_flatten that replaces the deprecated cfdm.netcdf_flatten (#355)
• Raise IndexError for out-of-range indices in a value-setting operation on a data array (#377)
• New function to control the creation of cached elements during data display: cfdm.display_data (#363)
• New methods: cfdm.Data.get_cached_elements, cfdm.Data.cache_elements (#363)
• Set cached elements during cfdm.Data.__init__ (#363)
• Removed the cfdm.constants.CONSTANTS dictionary, replacing it with cfdm.ConstantAccess.constants (#361)
• Reduce the time taken to import cfdm (#361)
• New optional dependency: zarr>=3.1.3
• Removed dependency (now optional): zarr>=3.0.8

What's Changed

• Reduce the time taken to import cfdm by @davidhassell in #362
• Zarr write + Zarr groups by @davidhassell in #368
• Control setting and display of cached data elements by @davidhassell in #376
• Detect out-of-range integer indices in __setitem__ operations by @sadielbartholomew in #378

Full Changelog: v1.12.3.1...v1.13.0.0

1.12.3.1

2025-10-15

• Python 3.9 support removed (#353)
• When changing array data type, retain cached data elements cast to the new type (#358)
• Fix bug that prevented cfdm.read from reading a file with # characters in its file name (#356)
• Changed dependency: Python>=3.10.0

What's Changed

• Allow # in file names by @davidhassell in #357
• Retain cached data array values when changing dtype by @davidhassell in #359
• Remove support for Python 3.9 (EOL) by @davidhassell in #360

Full Changelog: v1.12.3.0...v1.12.3.1

1.12.3.0

2025-08-18

• Fix cfdm.Data.reshape when the underlying data originate on disk (#348)
• New keyword parameter to cfdm.Field.dump: data (#345)
• New dependency: distributed>=2025.5.1

1.12.2.0

2025-06-05

• Implement lossy compression via quantization (#330)
• New quantization classes: cfdm.Quantization, cfdm.mixin.QuantizationMixin (#330)
• New quantization methods: cfdm.Field.get_quantization, cfdm.Field.get_quantize_on_write, cfdm.Field.set_quantize_on_write, cfdm.Field.del_quantize_on_write (#330)
• Improve performance in cfdm.read by caching any array values retrieved from disk (#313)
• New keyword parameter to cfdm.write: chunk_cache (#328)
• Read Zarr datasets with cfdm.read (#335)
• Read multiple datasets simultaneously with cfdm.read (#336)
• New keyword parameters to cfdm.read: recursive, followlinks (#336)
• New keyword parameters to cfdm.read: cdl_string (#337)
• Update CF aggregation keywords (#341)
• Set new minimum version of dask: 2025.5.1 (#339)
• New dependency: zarr>=3.0.8
• Changed dependency: dask>=2025.5.1

1.12.1.0

2025-04-01

• Set new method cfdm.Array.astype (#331)
• Set new maximum version of dask: 2025.3.0 (#332)
• Changed dependency: dask>=2025.2.0,<=2025.3.0

What's Changed

• Add astype method on Array subclasses. Pin Dask. by @davidhassell in #333

Full Changelog: v1.12.0.0...v1.12.1.0

1.12.0.0

2025-03-18

• Set new minimum version of dask: 2025.2.0 (#321)
• Set new minimum version of numpy: 2.0.0 (#318)
• Introduction of reading and writing of aggregation datasets (for CF-1.13) (#319)
• Changed dependency: Python>=3.9.0
• Changed dependency: numpy>=2.0.0
• Changed dependency: netCDF4>=1.7.2
• Changed dependency: cftime>=1.6.4
• Changed dependency: dask>=2025.2.0

What's Changed

• Prevent read error for datasets including badly-encoded UGRID meshes by @sadielbartholomew in #316
• Introduction of aggregation datasets by @davidhassell in #320
• Add missing docstrings by @davidhassell in #322
• Move to numpy version 2 by @davidhassell in #323
• Allow dask>2024.7.1 by @davidhassell in #326
• Add missing docs by @davidhassell in #327

Full Changelog: v1.11.2.1...v1.12.0.0

1.11.2.1

2025-02-26

• Re-introduction of cfdm.Data.datetime_as_string (#324)

What's Changed

• Reinstate cfdm.Data.datetime_as_string by @davidhassell in #324

Full Changelog: v1.11.2.0...v1.11.2.1

1.11.2.0

2025-01-28

• Introduction of dask for all data manipulations (#317)
• Fix bug that returned incorrect results when an invalid identifer is provided to cf.Field.cell_methods (#299)
• Upgrades to allow cfdm to work with Python 3.12 (#302)
• Extension to the HDF5 chunks API (#309)
• New function cfdm.netcdf_flattener that replaces the import of netcdf_flattener (#286)
• New function cfdm.netcdf_indexer that applies netCDF masking and unpacking to arbitrary arrays (#285)
• Allow access to netCDF-4 files in S3 object stores (#285)
• Allow a choice of netCDF engines (#285)
• Fix bug that caused cfdm.write to fail when a parametric Z dimension coordinate did not have a computed_standard_name attribute (#303)
• New class cfdm.H5netcdfArray
• New class cfdm.NetCDF4Array
• Changed dependency: numpy>=1.15,<2.0
• New dependency: h5netcdf>=1.3.0
• New dependency: h5py>=3.10.0
• New dependency: s3fs>=2024.6.0
• New dependency: dask>=2024.6.0,<=2024.7.1
• Removed dependency: netcdf_flattener

What's Changed

• Python 3.12 compatability by @davidhassell in #292
• Update linting package versions for Python 3.12 by @davidhassell in #306
• Fix cfdm.write when a parametric Z dimension coordinate does not have a compute_standard_name attribute by @davidhassell in #304
• h5netcdf read by @davidhassell in #307
• Extension to the HDF5 chunks API by @davidhassell in #310
• Fix bug that returned incorrect results when an invalid identifer is provided to cf.Field.cell_methods by @davidhassell in #300
• Introduction of Dask for all data manipulations by @davidhassell in #312

Full Changelog: v1.11.1.0...v1.11.2.0

1.11.1.0

2024-03-01

• New keyword parameter to cfdm.Field.insert_dimension: constructs (#287)
• New example field 11: discrete sampling geometry trajectory features (#289)