Added reporting of test code coverage.

Code coverage by the test suite is measured via pyTest and its
pytest-cov plug-in. As pyTest cannot run the entire test suite
in one go (as the Java VM is started multiple times, by separate
test scripts), the script `coverage.py` is provided in the `deploy`
folder to automate the runs of inidividual test scripts and combine
the reports into a unified file.

The script also creates the coverage badge, unless that file already
exists. In other words, delete the badge to update the coverage
percentage.

Test coverage is reported for Windows only. Code that targets other
platforms is therefore exluded from the reporting.

.gitignore

@@ -1,12 +1,14 @@
 /develop/
 /deploy/docs/
 /deploy/dist/
+/deploy/coverage/
+/deploy/coverage.sqlite
 /dist/
 **/__pycache__
 **/.pytest_cache
 flake8.txt
+.spyproject
+.idea
 *.lnk
 .DS_Store
 Thumbs.db
-.spyproject
-.idea

ReadMe.md

@@ -30,4 +30,5 @@ Find the full [documentation on Read-the-Docs][docs].
 [![license](https://img.shields.io/badge/License-MIT-green.svg)](https://opensource.org/licenses/MIT)
 [![release](https://img.shields.io/pypi/v/mph.svg)](https://pypi.python.org/pypi/mph)
 [![downloads](https://pepy.tech/badge/mph)](https://pepy.tech/project/mph)
+![coverage](tests/coverage.svg?raw=true)
 [![documentation](https://readthedocs.org/projects/mph/badge/?version=latest)](https://mph.readthedocs.io/en/latest)

deploy/clean.py

@@ -12,5 +12,5 @@
 for folder in root.rglob('.pytest_cache'):
     rmtree(folder)
 
-for folder in (root/'deploy'/'dist', root/'deploy'/'docs'):
-    rmtree(folder, ignore_errors=True)
+for folder in ('docs', 'dist', 'coverage'):
+    rmtree(root/'deploy'/folder, ignore_errors=True)

deploy/coverage.py

@@ -0,0 +1,25 @@
+"""Measures code coverage by test suite."""
+__license__ = 'MIT'
+
+from subprocess import run
+from pathlib import Path
+
+here = Path(__file__).resolve().parent
+root = here.parent
+
+file = here/'coverage.sqlite'
+if file.is_file():
+    print('Rendering coverage report.')
+    folder = (here/'coverage').relative_to(root)
+    run(['coverage', 'html', f'--directory={folder}'], cwd=root)
+    badge = root/'tests'/file.with_suffix('.svg').name
+    if badge.exists():
+        print('Coverage badge already exists.')
+    else:
+        print('Rendering coverage badge.')
+        run(['coverage-badge', '-f', '-o', str(badge)], cwd=root)
+else:
+    command = ['pytest', '--cov', '--cov-append']
+    tests = ['discovery', 'server', 'session', 'node', 'model', 'processes']
+    for test in tests:
+        run(command + [f'tests/test_{test}.py'], cwd=root)

docs/demonstrations.md

@@ -381,9 +381,6 @@ block of ice
 ├─ tables
 ├─ plots
 └─ exports
-
->>> mph.tree(model)
-block of ice
 ```
 
 The parameter group, model component, default view and coordinate

docs/limitations.md

@@ -16,18 +16,11 @@ to be started as separate Python subprocesses. This is a feasible
 work-around, but a limitation nonetheless. Refer to section  ["Multiple
 processes"](demonstrations.md#multiple-processes) for a demonstration.
 
-For the same reason, the [test suite][tests] that comes with MPh fails
-when collected and run via the testing framework [pyTest][pytest].
-The tests must be run directly from the command line. Since they cannot
-be used for continuous integration testing anyway, given that they
-depend on Comsol being installed on the machine, this is but a minor
-inconvenience.
-
-More importantly, there are some known, but unresolved issues with
-JPype's shutdown of the Java virtual machine. Notably, pressing
-<kbd>Ctrl+C</kbd> to interrupt an ongoing operation will usually crash
-the Python session. So do not rely on catching [`KeyboardInterrupt`][kbint]
-exceptions in application code.
+Additionally, there are some known, but unresolved issues with JPype's
+shutdown of the Java virtual machine. Notably, pressing <kbd>Ctrl+C</kbd>
+to interrupt an ongoing operation will usually crash the Python session.
+So do not rely on catching [`KeyboardInterrupt`][kbint] exceptions in
+application code.
 
 (There is an alternative Java bridge, [pyJNIus][jnius], which is
 not limited to one virtual machine, but then fails in another regard:
@@ -112,8 +105,6 @@ where this works reliably.
 
 [jpype]:  https://github.com/jpype-project/jpype
 [japi]:  https://comsol.com/documentation/COMSOL_ProgrammingReferenceManual.pdf
-[tests]:  https://github.com/MPh-py/mph/tree/main/tests
-[pytest]: https://docs.pytest.org
 [kbint]:  https://docs.python.org/3/library/exceptions.html#KeyboardInterrupt
 [jnius]:  https://pyjnius.readthedocs.io
 [issue8]: https://github.com/MPh-py/MPh/issues/8

pyproject.toml

@@ -25,6 +25,7 @@ classifiers = [
 
 [tool.flit.metadata.requires-extra]
 doc = ['Sphinx', 'Furo', 'MyST-parser', 'CommonMark']
+test = ['pyTest', 'pyTest-cov', 'coverage-badge']
 
 [tool.flit.metadata.urls]
 Documentation = 'https://mph.readthedocs.io'

tox.ini

@@ -17,3 +17,20 @@ ignore =
     W503,
 per-file-ignores =
     mph/__init__.py:F401
+
+[pytest]
+addopts = --verbose
+testpaths =
+    tests
+filterwarnings =
+    ignore::UserWarning
+
+[coverage:run]
+source = mph/
+omit = mph/discovery.py
+data_file = deploy/coverage.sqlite
+
+[coverage:report]
+exclude_lines =
+    pragma: no cover
+    def check_environment