CMOR 3.15.2

Features

• Added CMIP7 Fortran examples for regular grids, pressure levels, scalar height coordinates, basin axes, hybrid sigma levels, curvilinear grids, and fixed fields. (#977)
  • The Fortran examples include a shared helper module, a run script, and nightly-build coverage.
• Added CMIP7 Python examples matching the new Fortran example coverage. (#978)
• Added example CMIP7 cmorizing notebooks using real model-data excerpts, including regular lat-lon data, pressure levels, hybrid sigma coordinates, basin axes, land-use fractions, and non-lat-lon sea-ice data. (#976)

Bugfixes

• CMOR now checks CV JSON strings and keys against the 1023-character limit before copying them into fixed-size strings, preventing crashes and reporting the offending value in the error message. (#985)
• The Fortran cmor_dataset_json wrapper now returns a reliable status code. (#977)

Changes

• Concatenated variable comment attributes now identify the source table by file name instead of using activity_id. (#984)
  • Example output now uses text like CMIP6_Amon.json comment: ....
• Error and warning color formatting is now only emitted for terminal output, avoiding ANSI color characters in log files. (#985)

Issues / Pull Requests

• Handling of grids other than lat-lon (#832)
• preparing a CMOR demo with a coordinate having character string entries (#837)
• CMORizing notebook (#878)
• CMIP7 Fortran examples (#885)
• Example CMIP7 cmorizing notebooks (#976)
• CMIP7 Fortran examples (#977)
• Add Python examples (#978)
• Use of activity id in concatenated variable comments (#979)
• No error or warning for CV string values exceeding 1023 character limit (#982)
• Error message coloring format changes (#983)
• Use table file name instead of activity_id in variable comment (#984)
• Check the CV file for strings that exceed CMOR's character limit (#985)

GitHub

CMOR 3.15.1

Bugfixes

• The table_info global attribute will now contain the table's file name. (#970)
  • Example: table_info = "Name: CMIP7_atmos.json; Creation Date: ....
• The default history global attribute text was updated to remove the redundant "CF standards" wording. (#969)
  • CMOR now writes history entries in the form CMOR rewrote data to be consistent with <Conventions> and <mip_era> data requirements.
• Time-interval validation now runs for time1 axes even when bounds are not supplied in cmor_axis. (#972)

Changes

• Climatology time axes are excluded from interval checking in cmor_write. (#972)

CI / Tests

• CMIP7 Python tests now use the latest CV file from cmip7-cmor-tables instead of the stale test CV. (#973)
  • Removed outdated CMIP7 JSON files from TestTables/.
  • Test cleanup now removes temporary input and CV files created during runs.

Issues / Pull Requests

• approx_interval (#943)
• Use the table file name in table_info (#964)
• Edit the history global attribute text (#966)
• Remove redundant "CF standards" from history template. (#969)
• Add the table's file name to the table_info string. (#970)
• Ignore interval checks for climatology time axes, and fix interval checks being ignored for time1 axes (#972)
• Python test updates (#973)

GitHub

CMOR 3.15.0

Features

• Added official Python wheel builds for CMOR releases, including Linux wheels, full Python test coverage in the wheel pipeline, and release assets attached on GitHub. (#932)
• CMIP7 controlled vocabularies can now define drs_specs and tracking_prefix as root-level strings. (#961)
  • A root-level drs_specs string is now written to the output file as a global attribute.
  • A root-level tracking_prefix string is now used when generating tracking_id.

Bugfixes

• User-provided Conventions values in the dataset JSON are now preserved and written to the output file instead of always being replaced by the table default. (#960)
• The default table-header product value is now model-output instead of model_output, aligning generated metadata with current DRS terminology. (#960)

CI / Tests

• Fixed test-cleanup failures in make test_python and make test_cmip6_cv that could raise OSError: [Errno 39] Directory not empty during teardown. (#955)
  • CMOR now removes the generated NetCDF files without trying to remove shared output directories that may still contain files from other tests.

Changes

• Nightly Linux builds that were previously run on CircleCI now run on GitHub Actions. (#958)
• Nightly package versions are now kept in sync. (#959)

Issues / Pull Requests

• Wheel build test with cibuildwheel pipeline (#932)
• Migrate the Linux nightly build pipeline from CircleCI to GitHub Actions (#952)
• testing errors in make test_python and make_test_cmip6cv using new QUICKSTART instructions (#953)
• Fix issues in #953 when running Python and CMIP6 CV tests (#955)
• Add Linux pipeline to GitHub nightly builds; remove CircleCI config (#958)
• Keep package version updated in nightly builds (#959)
• "product" and "Conventions" (#935)
• Modify CMOR user input json file (#936)
• Allow user override of Conventions attribute (#960)
• Allow CV to set tracking_prefix and drs_specs if they are single string values. (#961)

GitHub

CMOR 3.14.3

Bugfixes

• Fixed file names for CMIP7 climatology datasets. (#950)
  • CMIP7 climatology outputs now use the expected monthly time-range format in file names.
  • Added regression coverage in Test/test_cmor_CMIP7.py for the output filename, frequency attribute, and climatology metadata.

Changes

• Removed unused filename-validation and suffix-tracking code from the C core, simplifying internal output-name handling. (#949)

Documentation

• Added QUICKSTART.md to the main CMOR repository with concise instructions for cloning, building, installing, and testing CMOR in a Conda/Mamba environment. (#951)
• Updated the user-input documentation in this repository to clarify how version affects output paths and global attributes. (#192)
  • The guide now explains that setting version adds a netCDF global attribute.
  • It also shows how to use an internal attribute such as _dir_version with <dir_version> in output_path_template when only the directory name should change.

Issues / Pull Requests

• Error in creating date range in filename (#939)
• Remove unused code (#948)
• Remove unused code (#949)
• Fix file names for CMIP7 climatology datasets (#950)
• Quickstart instructions (#951)
• Update guidance on output path version (#192)

GitHub

CMOR 3.14.2

Changes

• Refreshed CMIP7-oriented user guidance in cmor_CV.c for invalid source_id and institution_id values, including current WCRP-CMIP issue links and the CMIP7 CV registration guide. (#942)
  • Updated source_id and institution_id CV error messages to point users to current CMIP7 registration issue templates and guidance instead of outdated links.
• CMIP7 datasets now write parent and branch attributes only when the selected experiment defines parent experiments in the controlled vocabulary. (#940)
  • This is done when _cmip7_option is enabled in the user input JSON.
  • When a CMIP7 experiment has no parent experiments, CMOR now warns about stray parent attributes and removes them before writing the output file.
  • When a CMIP7 experiment requires a parent, CMOR now errors if parent_experiment_id is missing, not allowed by the controlled vocabulary, or paired with an invalid parent_activity_id.

CI / Tests

• Added Test/test_cmor_parent_attrs.py coverage for CMIP7 parent-attribute handling across no-parent experiments, required-parent experiments, warning-and-removal behavior, and invalid parent metadata. (#940)
  • Updated existing CMIP7 Python tests to reflect the new parent-attribute rules and to clean up temporary test inputs created during test runs.

Issues / Pull Requests

• urgent release requested (#937)
• update links and suggestions after CV error for source_id, institution_id (#938)
• Only write parent attributes when they are required for parent experiments in CMIP7 datasets (#940)
• chore: Update cmor_CV.c with up-to-date CMIP7-oriented links and advice (#942)

GitHub

CMOR 3.14.1

Bugfixes

• Improved error messages for time interval validation to provide clearer guidance to users. (#915)
  • When a dataset has a defined frequency attribute, the error message will include both the frequency name and its corresponding approximate time interval.
• Added a clearer error message when Zstandard compression is requested but the ZSTD filter is unavailable. (#925)
  • This error instructs users to install hdf5plugin and set the HDF5_PLUGIN_PATH environment variable.

Changes

• CMIP7 datasets: ignore the branch_method global attribute and do not add table_id to output files. (#926)

CI / Tests

• CI build migrated from conda-build to rattler-build, improving build performance. (#922)
• CI test environment now installs hdf5plugin and configures HDF5_PLUGIN_PATH for the NetCDF ZSTD filter. (#922)
• Updated CMIP7 test file templates. (#921)
• Updated CMIP7 dataset unit tests to match latest required/optional global attributes. (#926)
  • Now requires license_id, adds drs_specs, no longer requires experiment, and removes grid and source_type.

Closed Issues

• Updates to dataset frequency and time interval errors (#913)
• output_file_template leads to a filename with _variant_idtime_range_ - bug or user error? (#918)
• Migrate CI build to conda-rattler (#919)
• Update error message for zstandard compression to warn users to set up HDF5_PLUGIN_PATH (#923)
• CMIP7: branch_method global attribute dropped for CMIP7 still required by CMOR (#924)

GitHub

CMOR 3.14.0

Features

• New chunking strategy to support CMIP7 repacking requirements.
  • Chunking dimensions of the time and time bounds axes will be set to the length of the time axis, plus a dimension of size 2 for the bounds, if it is known when the time axis is created. Otherwise, the time and time bounds chunking dimensions will be set to 512 by default.
  • Chunking dimensions for data variables will be determined by whether chunks can be at least 4 MB in size.
    • If multiple time slices of data can fit within 4 MB, then chunking dimensions will have multiple time slices and full spatial dimensions of the data variable.
    • If a single time slice is greater than 4 MB, then chunking dimensions will have one time slice with spatial chunking dimensions determined by the NetCDF library.
• The function cmor_set_chunking has been added to the API, allowing users to define NetCDF variable chunking dimensions.
  • Setting the chunking dimensions with cmor_set_chunking will override the automatic chunking strategy for CMIP7 repacking.

Bugfixes

• Fixed bug where using cmor.write with the store_with parameter will throw an error when time values are not passed.

Changes

• Removed legacy Python 2 code from Python interface and unit tests.
• cmor_setup in the Fortran API will now allow the use of NetCDF4 functionality when setting the parameter netcdf_file_action with either "preserve_4", "append_4", or "replace_4".

Closed Issues

• Repacking (#880)
• Allow users to manually define variable chunking through the API (#895)
• Error using store_with option in cmor_write (cmor 3.13) (#902)

GitHub

CMOR 3.13.2

Changes

• The function cmor_set_variable_attribute will now allow the variable attribute long_name to be modified.

Closed Issues

• Allow users of CMOR to override long_name from tables (#891)

GitHub

CMOR 3.13.1

Changes

• The CMIP7 CMOR tables repo will now be a submodule of the CMOR repo. CMIP7 unit tests for CMOR have been modified to use these tables.
• The attribute branded_variable will be added to a dataset's global attributes if the variable table entry name has a branding_suffix.

Closed Issues

• Update CMIP7 tests to use cmip7-cmor-tables (#879)
• branded_variable global attribute missing from output files generated with CMOR v3.13 (#886)

GitHub

CMOR 3.13.0

Features

• The global attributes realization_index, initialization_index, physics_index and forcing_index can now have string values when either _use_strings_for_indexes or _cmip7_option are present in the user input parameters.

Bugfixes

• Resolved an issue where running multiple unit tests within a unittest class that used variable tables with branding suffixes would corrupt the branding suffixes, labels, and out names in subsequent tests within the unittest class.

Changes

• Added support for Python 3.14.
• PrePARE has been removed from CMOR.
• CMIP7 tests have been updated to confirm that secondary realms can be applied to variables by having multiple realms listed for modeling_realm in variable entries. See test here.

Closed Issues

• Handling of secondary realm for CMIP7 output (#861)
• TBC: Allow *_index values to be strings (#864)
• Remove PrePARE code from main branch (#868)
• Processing multiple variable tables with branding suffixes corrupts labels and out names (#870)

GitHub