Tests

The tests are plain Python scripts, without pytest, fixtures or configuration. Each script prints one line per check, starting with OK or NO, then the number of checks and of failures. It exits with a non-zero status if any check failed. The descriptions of the checks are in Italian.

python3 check_rules.py       # 147 checks: the Radiopaedia linter rules
python3 check_structure.py   #  24 checks: the article structures
python3 check_search.py      # 159 checks: query building, ISSG, strategies, lists
python3 check_journals.py    #  85 checks: SCImago, matching, Unpaywall, backups
python3 check_library.py     #  76 checks: PDF library, highlights, the Radiopaedia list
python3 check_bench.py       #  87 checks: the Screening page and its local server
python3 check_app.py         # 148 checks: the interface, driven without a browser

That is 726 checks. None of them needs the network, and none reads or changes a real archive. The five scripts that use the database set RADIOPAEDIA_DB to a temporary file before they import db. check_library.py, check_bench.py and check_app.py also set RADIOWRITER_HOME to a temporary folder, because the PDF library is created there.

check_bench.py starts the Screening server on a free port of 127.0.0.1 and sends it real HTTP requests. It checks who may call the server, what the article pages contain, and every action. It does not run the JavaScript of web/screening.html. scripts/shots.py opens that page in a browser.

check_library.py creates its PDFs with PyMuPDF, with highlights, and calls the reader's ReaderApi directly. It does not open any window. web/viewer.html has no automated test. Test it by hand when you change it.

The test that needs the network

python3 check_mesh_live.py

It asks PubMed whether every MeSH descriptor and subheading used in strategies.py and modalities.py exists. Run it when you change either file, and each January, when MeSH is updated. It takes about a minute without an NCBI API key and about 20 seconds with one.

It reads the NCBI email address and API key from the settings of your archive, so the email address must be set in the app.

check_app.py

The other scripts test the modules below the interface. check_app.py tests the interface itself, including how it calls those modules.

streamlit.testing.v1.AppTest runs the app in memory and operates its widgets by key, so the test runs the same app.py that is installed:

at = AppTest.from_file(APP, default_timeout=60)
at.run()
at = at.button(key="recent_open_btn").click().run()
is_("the button opens the panel with the editorial groups",
    at.selectbox(key="recent_group").options, "['CNS']")

Two kinds of error are found only by this script:

• a value that Streamlit cannot copy passed as a widget option, for example a sqlite3.Row, which makes the whole page fail;
• a write to session_state for a widget that already exists.

The output of st.caption is in at.caption, and is not included in at.markdown.

What CI runs

.github/workflows/test.yml runs on every push to main and on every pull request, on macOS, Linux and Windows, with Python 3.11 and 3.13.

It runs the seven scripts. It then runs radiowriter --version and radiowriter --where, starts the installed command, and requests /_stcore/health until the server answers. This last step is the only test of the radiowriter entry point as it is installed.