Developer guide

Guidance for contributing to the package and for maintaining its releases. If you only want to process a submission, you do not need any of this – see Getting started.

The repository is laid out like this:

ismip7_scalars/scalars.py

the processing pipeline and the ismip7-scalars entry point.

ismip7_scalars/ensemble.py

the batch driver and the ismip7-scalars-ensemble entry point.

ismip7_scalars/naming.py

the ISMIP7 filename conventions, as pure functions.

ismip7_scalars/writers.py

the NetCDF and CSV output.

ismip7_scalars/params.py

the params.nc writer and the ismip7-scalars-set-params entry point.

ismip7_scalars/slc/

the three sea-level methods, plus slc_G2020_publ as a cross-check.

ismip7_scalars/paths.py, ismip7_scalars/variables.py

where the data files come from, and the ISMIP7 data request lookup. The data request is read from isschecker, not copied; see Where the conventions come from.

ismip7_scalars/data/

the CORE experiment table — this package’s own, unlike the data request.

tests/

the test suite, including synthetic.py, which writes the miniature submissions the integration tests run over.

matlab/scalars.m

the MATLAB implementation of the same processing. See The MATLAB implementation.

manual-tests/

what cannot run in CI – the Python/MATLAB comparison – and the MINI cases. See The MINI cases.

test-data/

the committed MINI inputs.

docs/

these pages.

Contents

• Installing from source
  • Dependency ranges
• Running the tests
  • What is where
  • The synthetic submission
  • What CI runs
  • What is not tested here
• The MINI cases
  • The cases
  • How MINI differs from the real processing
  • Regenerating the input files
• The MATLAB implementation
  • Where the two differ
  • Comparing them
• Building the documentation
  • Before opening a pull request
  • Publishing
  • Writing
• Releasing
  • The relationship with isschecker
  • Cutting a release
  • The first release
  • Confirming what was published
  • Maintaining the feedstock