UCT ← Back to the explorer's documentation page  ·  source: extractors/README.md

Extractors

One package per faculty (com, ebe, fhs, hum, law, sci) plus fees for the student fees handbook. All seven are built for 2021–2026.

Faculty handbooks that follow the university's common publisher template (COM, EBE, LAW) are parsed by the shared engine common/handbook_parser.py; the faculty package then contains only a FacultyConfig (plan-code grammar, degree parser, heading variants, page classification, department map) plus any parse-time overrides. Faculties whose books don't fit the template (FHS, SCI, HUM) have bespoke parsers in their packages that reuse the shared grammar and the engine's catalogue parser — see REPLICATION §7–§9. In SCI and HUM the curriculum unit is the major (SB001…/HB001… plan codes), not the specialisation.

Each extractor package provides:

• extract.py — entry point (python -m extractors.<fac>.extract --year Y): a FacultyConfig + run_extractor(...) call, or a bespoke parser with the same outputs. Faculty extractors also run the rules-layer extractor (common/degree_rules.py → degree_rules.csv) after the main parse.
• overrides.py (or .csv) — explicit, reviewable corrections for parse artifacts (never hand-edit outputs; print errors go to resolutions/).
• Hazard notes belong in docs/REPLICATION.md.

Shared tables are written merge-by-year AND merge-by-faculty-within-year: re-running one extractor replaces only its own faculty's rows for that year.