mirror of
https://github.com/MadsLorentzen/ai-job-search.git
synced 2026-09-17 00:26:26 +00:00
Ayobami Adegoke
968fb1bd7e
The pairing loop required a non-empty derived category name on both
sides, but a header with no category word - "Count" + "Index", Danish
"Antal" + "Lønindeks" - strips to an empty name, so the simplest layout
the README advertises ("auto-pairs count/index columns") came out as two
unrelated standalone categories:
{"count": {"count": 500}, "index": {"index": 108.5}}
salary_lookup then rendered a "Count 500 N/A*" row above an
"Index - 108.5" row, and its footnote read the N/A* as "too few
employees to publish (privacy)" - a false statement about a company whose
headcount is in the file, shown during /apply's salary step. Any suffix
("Antal alle") made pairing work, which is why the shipped tests, all
suffixed, never saw it.
Pair on equal derived names, empty included, and give the nameless pair
the README's top-level category name (all_employees). A bare "Antal" with
no bare index column still stays a standalone count; named pairs
alongside are untouched.
Four new cases in test_convert_salary_excel.py, one of them rendering the
converter's output through salary_lookup.format_entry; all four fail
against the old pairing rule.