- workaround for newer (post Nov '22: https://github.com/numba/numba/commit/87b44afc89ff71ac6a569a898953786287dd8eb8) Numba versions which err if a jitclass has a `__call__` method #402 (in PyMPDATA used for post-iter and post-step hooks) - moved `PyMPDATA-examples` repo content into `PyMPDATA` (retaining separate packages on PyPI.org) - switched to OIDC for PyPI publishing - introduced externally sourced `devops_tests` - excluded notebooks from examples package wheels - use `extras_require` for test dependencies (`pip install pympdata[tests]`)
thanks Delcior and abulenok for brainstorming together on the repo-merge workflow
1.0.11
- optional bc data buffer allocation in Traversals (for use, e.g., in MPI) - kudos Delcior - proper math markdown in README - tip release fix (workflow was still referring to `master` and not `main` branch) - Github Actions workflow refactors
1.0.10
- redesign vector-field halo-filling code enabling separate code for normal and parallel logic (thanks Delcior) - extended docstrings and test coverage for boundary conditions (periodic, extrapolated)
1.0.9
- redesign of halo-filling architecture enabling exchanges of domain slabs instead of per-grid-cell logic (kudos Delcior)
1.0.8
- update links in README to reflect GitHub org move to open-atmos - fix JSON syntax in .zenodo.json and add a CI workflow for checking it
1.0.7
- added an option to `Stepper::__init__` to select halo-filling traversal order (left first or right first) + major code refactor around halo-filling logic (kudos piotrbartman!)