Hwaro::Content::Taxonomies
Class methods
The candidate files (term_metadata_path?) that would publish at the
URL (NFC) of a term page this build writes: the directory is the
term's slug in the file's language — content/tags/rust-lang/_index.md
for /tags/rust-lang/, content/tags/c-2/_index.md for a
disambiguated /tags/c-2/, _index.ko.md for /ko/tags/<slug>/ —
and no path or [permalinks] rule moved it elsewhere. Keyed by
section path. A matched file describes that term page instead of
publishing a page of its own (Builder#mark_term_metadata); every other
candidate stays an ordinary section, exactly as before. When several
files match one page (_index.md beside _index.markdown, or
_index.ko.md beside _index.ko.markdown), a plain _index.md wins,
then a .md one, else the first path; the others are reported and
not published.
Sort every term's page list per its taxonomy's sort_by/reverse (date-desc for taxonomies not in the config). ONE implementation serves both the generator (populate_taxonomies) and the render phase's rebuild_taxonomies (transform.cr) — per-term ordering drifting between the written pages and get_taxonomy is exactly the bug class this shared helper prevents.
True for content/<taxonomy>/<dir>/_index[.<lang>].md with
<taxonomy> a configured taxonomy: a file that MAY describe a term
page (gh#891). Only match_term_metadata decides whether it does;
doctor, which builds nothing, uses this alone to stay quiet about
such files. content/<taxonomy>/_index.md (the taxonomy index) is
never a candidate.