Plangs aims to become a community-oriented programming languages site: trends, tools, blog posts, and learning resources for programming languages and their ecosystems.

v3 note. The site was rebuilt in 2026 on Astro + React + a YAML/graphology data layer. The previous Bun/Preact/esbuild implementation is preserved at the final-plangs-2 tag; the migration plan it followed is in git history (git show 008e073c:PLAN.md).

Requirements

• Node 22.12+
• pnpm 10+ (install via mise/asdf, or corepack enable)

That's it — no Bun, no Ruby/Docker, no Overmind/entr.

Layout

Path	What
packages/schema	Single source of truth: node kinds, the 52 edges, Zod schemas, key helpers, inference rules.
packages/data	The dataset. One YAML file per node (nodes/{kind}/{slug}.yaml) plus logos (assets/{kind}/{slug}.webp). No code.
packages/graph	Loads the YAML into graphology; integrity checks, materialization, typed queries, exporters.
packages/pipeline	Importers (Linguist, Languish, Wikidata, …) and AI enrichment.
apps/site	The Astro 7 site: routes, layouts, React islands.

Run it locally

pnpm install
pnpm -F @plangs/site dev     # http://localhost:4321, hot reload

Astro 7's dev server runs detached: pnpm exec astro dev stop|status|logs (from apps/site).

To build everything and serve the real static output instead:

pnpm build                    # typecheck every package + build the site
pnpm -F @plangs/site preview  # serves apps/site/dist

Command Line

• pnpm build — typecheck the packages and build the static site into apps/site/dist.
• pnpm test — run the test suite (includes the migration round-trip gate).
• pnpm check / pnpm fmt — lint/format with Biome.
• pnpm url-parity — assert every v2 URL is still served (needs a build first).
• pnpm data:fmt — rewrite every node YAML in canonical order (minimal diffs).
• pnpm pipeline run --source=<id> [--dry-run] — refresh data from an upstream source.

See ROADMAP.md for what's done and what's left, and CLAUDE.md before changing the code — it records which oddities are load-bearing (and why), the verification loop, and the environment facts.

Editing data

Every node is a YAML file. To add or edit a language, edit packages/data/nodes/plang/{slug}.yaml and open a pull request — each node page links straight to its file on GitHub. A logo is a 128x128 webp at packages/data/assets/{kind}/{slug}.webp.

name: Nim
shortDesc: Statically typed, compiled language…
created: "2008"
extensions: [".nim"]
rels:
  compilesTo: [pl/c, pl/javascript]
  paradigms: [para/compiled, para/functional]
  licenses: [lic/mit]

CI validates the data: every edge target must exist, every edge must be legal between its two kinds, and every node must satisfy its Zod schema.

Node keys and prefixes

Keys are prefix/slug (e.g. pl/nim). The v2 form was pl+nim.

Prefix	Description
app/	Software Application.
author/	Author for Blog Posts.
bun/	Bundle of tools.
comm/	Community.
learn/	Learning Resource (book, video, course, etc).
lib/	Software Library.
lic/	Software License.
para/	Language paradigm.
pl/	Programming Language.
plat/	Platform (Operating System, WASM, CPU, etc).
sys/	Subsystem: databases, queues, and other services.
tag/	Generic Tag.
tool/	Software Tool.
tsys/	Type System.

For LLMs and agents

• Every page has a markdown twin — append .md to any URL (e.g. /nim.md).
• /llms.txt describes the dataset and the filter-URL grammar, so an agent can construct queries like /plangs?paradigms=functional&platforms=web&mode=all.
• /data/plangs.json is the full graph: {vertices: {kind: {key: data}}, edges: {edgeName: {fromKey: [toKey]}}}, keyed by prefix/slug (pl/nim, para/oop) like every other data route.

Deploying

The site is pure static output deployed to Cloudflare Workers static assets:

pnpm -F @plangs/site build && npx wrangler deploy

The v3 cutover has not happened yet — plangs.page still serves v2. Read ROADMAP.md "Deploy / cutover" before the first deploy (or any push): the legacy Cloudflare Pages project may still have a git integration.

Acknowledgements

Some sources only appear in git history — for example, a bulk script used just once. To credit those sources too, we list them here.

• @llaisdy's list of logic programming languages: llaisdy/PrologInfo.

Graph inference notes

Some relationships can be used to infer others. These notes are no longer just notes: they are encoded as declarative rules in packages/schema/src/inference.ts and applied by the materialization pass in packages/graph/src/materialize.ts, with the negative cases enforced by tests.

In Bundle -> Tool -> Plang implies Bundle -> Plang, meaning a bundle supports a plang if any of its tools supports it.

The example B/T/P demonstrates a transitive relationship. There are other relationships that are possible but should not be used for inference.

In Plang -> App -> License, inferring Plang -> License would be wrong, since an App built with Plang doesn't necessarily share the License.

There are some relationships that are simpler to infer.

In A -i-> B -i-> C (i = influenced), A influenced both B and C.

The process of adding a relationship to the data that could be inferred is called "materializing". Materialized edges are marked inferred: true, live only in the built graph (never written back to YAML), and are what the facets and queries run against — so searches like "languages transitively influenced by ML that compile to JS" work.