Skip to content

Various

Running tests

Prefer uv so the project .venv and lockfile are used:

uv sync
uv run pytest                  # full suite (default marker filters apply)
uv run pytest -m essential     # fast smoke — same subset as CI essential job

Default addopts in pyproject.toml deselects several markers (slowtest, smoketest, unimportant, unfinished, onlylocal). Opt in explicitly when you need those, e.g. uv run pytest -m onlylocal.

Benchmarks are separate (pytest -m benchmark or the benchmarks/ tree) — see GitHub workflows below.

More detail on golden fixtures and suite layout: tests/README.md.

The essential marker

@pytest.mark.essential marks tests that every PR must keep green. Use it for:

  • read → step table → summary pipeline smoke
  • cellpy / cellpy-core parity contract
  • golden-fixture oracles under tests/data/goldens/
  • other regressions you cannot leave to the weekly scheduled run

Keep the set small so the essential job stays fast. Platform-specific paths (Windows ODBC, macOS-only skips) usually stay in the full suite unless you add a targeted essential job for that platform.

Using pytest fixtures

Retrieve constants during tests

There is a fixture in conftest.py named parameters that exposes variables defined in tests/fdv.py. Add new test constants by editing fdv.py.

Other

Check tests/conftest.py for other fixtures.

Example

from cellpy import prms

# using the ``parameters`` and the ``cellpy_data_instance`` fixtures.

def test_set_instrument_selecting_default(cellpy_data_instance, parameters):
    prms.Instruments.custom_instrument_definitions_file = parameters.custom_instrument_definitions_file
    cellpy_data_instance.set_instrument(instrument="custom")

GitHub workflows

Workflow When What
.github/workflows/ci.yml PRs and pushes to master essential — uv run pytest -m essential, then cellpy setup / info --check. full — uv run pytest (default filters) with batch extras. Doc-only path changes under docs/, *.md, .issueflows/, etc. are skipped.
.github/workflows/ci-scheduled.yml Weekly + manual Run workflow Conda / pip platform matrix, notebook smoke, conda-forge install check. Does not block merges.
.github/workflows/release.yml GitHub release published Tag validation → pytest -m essential → PyPI publish.
.github/workflows/benchmarks.yml Manual / separate from merge gate Performance benchmarks vs baselines.
.github/workflows/docs.yml Docs changes Zensical build + link/issue checks.

On a PR, treat essential and full as the linux/uv merge signals. Run the scheduled workflow manually before a release if the weekly run is stale.

Adding another config parameter

Session config lives in cellpy.config (pydantic models + layered TOML). cellpy.parameters.prms is a deprecated shim — do not add new fields there, and do not hand-edit .cellpy_prms_default.conf as the source of defaults.

  1. Add the field (with type and default) on the right section model in cellpy/config/models.py.
  2. Read it as cellpy.config.<section>.<field> (or via get_config()).
  3. Regenerate the user-facing reference:
uv run python -m cellpy.config.reference
  1. If the new key is a path that setup/CLI must treat as OtherPath, also add it to OTHERPATHS in cellpy/parameters/internal_settings.py. That file is for headers, meta helpers, and path-key lists — not general session config.

During the v2 window, cellpy setup may still dual-write legacy YAML and cellpy.toml. When a user TOML is present, it wins over .cellpy_prms_*.conf.

See also Setup and configuration, Configuration reference, and Coming from cellpy 1.x.

Installing pyodbc on Mac without conda

If you do not want to use conda, you might miss a couple of libraries.

The easiest fix is to install unixodbc using brew as explained in Stack Overflow #54302793.