beffroi-tools/rundowns/README.md

32 lines
2.1 KiB
Markdown

# Supplied rundown fixtures
These five XLSX files are committed at the repository owner's request for offline
regression testing. Only filenames changed: workbook bytes, sheet names, source
cells, formulas and formatting are intact. `sources.json` records the original
filenames and SHA-256 hashes. The files contain original production information;
inclusion does not assert a blanket third-party redistribution license.
| Workbook | Regression coverage |
| --- | --- |
| `estivales-2025-plenieres.xlsx` | Two sheets; merged timing blocks, missing final duration, retaining every closing-sheet source row after classifying rows without timing as notes. |
| `pleniere-2.xlsx` | Competing TIMER columns C and D must block automatic mapping. |
| `gifav-2026-auditorium.xlsx` | Two sheets; missing title mapping, explicit parent ranges, three gaps per sheet, successful Thursday export and the Friday overlap. |
| `moebius-matin-pleniere-sante-environnement.xlsx` | French Heure/Action/Temps headers, Excel durations, merged cells and final-row validation error. |
| `moebius-apres-midi-sante-mentale.xlsx` | Same French headers and typed times, final-row validation error and overlap. |
Run `cargo test --locked -p rundown-converter --test supplied_corpus`.
The full workspace tests also run these checks. Tests do not modify the sources
or require a network connection. Expected errors are useful regression cases;
these fixtures are not all ready-to-run schedules.
Filenames use lowercase ASCII words separated by hyphens. Dates include a year
only when known. Local exports keep `.ontime.json` / `.ontime.xlsx` suffixes;
sheet-specific GIFAV exports include `2026-11-05` or `2026-11-06`. Existing export
contents were not regenerated during renaming.
The local PDFs were renamed to `beffroi-montage-04-octobre.pdf` and
`beffroi-exploitation-demontage-05-octobre.pdf`. They remain untracked: the reader
rejects PDFs before inspecting their content, so including those binaries would
not add meaningful regression coverage. Other local inputs and generated exports
are ignored; `.gitignore` explicitly allows only the five tested workbooks.