Contributing¶
Structure¶
| Path | Responsibility |
|---|---|
src/testuary/ |
Generation, repair, phase orchestration, measurement, reduction, and result persistence |
tests/ |
Package regression tests, with separate real-measurement tests |
The package contains the four phases (the starter phase, the coverage phase, the mutation phase and Finishing), with repair, the choice of a suite at each phase end and suite reduction. Code organization follows one run through the modules. How a language plugs in is described in Make a language ready for Testuary.
Setup¶
For work on the source code, check out the LangAdapter repository next to this one and install both in editable mode in a virtual environment with Python 3.10 or newer:
python -m pip install -e ../langadapter
python -m pip install -e ".[dev]"
Tests¶
python -m pytest -q
python -m pytest tests/integration -m slow -q
The first runs every test not marked slow: the unit tests, integration tests
and web tests. The unit tests check single parts in isolation. Tests marked
slow run real measurements in subprocesses and need the measurement tools
installed. When you measure run times, run nothing else heavy on the machine at
that time.
Documentation¶
The site is built from docs/ with MkDocs. Install the docs extra and build
it strictly:
python -m pip install -e ".[docs]"
mkdocs build --strict
The command-line and settings reference pages are generated from the
argument parsers and the packaged config.ini. After changing an option, a
help text, a default, a setting or an exit status, rewrite them with
python tools/reference_pages.py. It also writes the exit statuses in the
guides. The level
tables of Make a language ready for Testuary come
from Testuary's profile: after changing the profile, rewrite them with
python tools/language_page.py. A test fails while either is out of date.
Public text is ASCII only, with no dashes, ellipses or emoji, no semicolons in prose, and no internal labels such as review codes, role names, commit hashes or decision dates. This holds for the README, the documentation pages, the shipped examples and the package's docstrings and comments. Comments say why the code does what it does.