Sources

Where each programming-language record comes from. A PL can be in multiple sources; cross-presence is what makes attribution credible.

Rosters

Curated name lists. Cross-presence across rosters is what makes a PL claim credible.

SourcePLsWhat it is
Llm2,439Curated by an LLM agent in this repo (has at least one example program).
Pldb5,288Programming Language DataBase — a curated community DB.
Linguist878GitHub's Linguist — the file-type detector used to render "% of repo" stats on GitHub.
Pygments645Pygments — the syntax highlighter; entry means there's a hand-written lexer.
Wikipedia769Wikipedia article in the Programming Language category. Per-PL infobox facts (paradigms, typing, designer, year, influenced-by, license, …) are pulled separately via the wikimedia/structured-wikipedia dataset and rendered on each /l/<slug>/ page — see the Enrichment sources panel below.
Esolang6,797esolangs.org — the catalog of esoteric languages.
Hyperpolyglot670hyperpolyglot.org — side-by-side language comparison tables.
Rosettacode640Rosetta Code — task implementations across languages.
Manual_add2

Enrichment sources

Not rosters of names — structured data graphs we pull per-PL facts and cross-system identifiers out of. Both are date-pinned snapshots under data/raw/; the provenance for every cell rendered on a PL page is in data/derived/pl_taxonomy/pl_fact.csv.

SourcePLsWhat it is
Wikidata 2,013 Cross-system identifier graph. We pull two pinned snapshots from WDQS: (a) every item carrying P1195 (filename extension), (b) every item whose instance of (P31) is in the 175-QID PL-types closure (programming language, markup language, query language, logic, esoteric). The union closes gaps where Wikidata has the article + types but no P1195 — R was the canonical example. Each pl.csv row carries a wikidata_qid when matched; the rest of the site cross-links via that.
Wikipedia infoboxes 1,220 Pre-parsed Wikipedia infobox trees from the wikimedia/structured-wikipedia HF dataset. One Parquet pass over enwiki extracts both filename extensions (composited with the legacy mwparserfromhell snapshot for the ~36 multi-section infoboxes upstream flattens) and the "Programming language" infobox facts — paradigms, typing discipline, designer, first appeared, influenced by, license, implementation languages, homepage. Those land in data/raw/wikipedia_pl_facts.<date>.jsonl and get spliced into pl.csv when PLDB is empty (PLDB-primary rule); the long-table view lives at data/derived/pl_taxonomy/pl_fact.csv.