feat(tooling): port the doc/lint/grammar checks to Ludic (no Python)
All checks were successful
bootstrap / cfree-fixpoint (push) Successful in 13s
ci / build-and-test (push) Successful in 52s
commit-lint / conventional-commits (push) Successful in 3s
docs / build-and-deploy (push) Successful in 2s

Replace the Python doc/lint/vocabulary guards with Ludic equivalents that
run through the `x` task runner, so the checks need no Python interpreter:

  x check-docs         every ```ludic doc fence parses (or is marked)
  x check-impl         every implemented feature has a docs/language page
  x check-vocabulary   vocabulary in sync across grammar / lexer / header / parser
  x lint-asset <file>  validate one editor .json / .xml asset

New fragments: tools/x/json.ludic (a small JSON reader — objects/arrays/
strings with \uXXXX + surrogates/numbers/literals, used by the vocabulary
check's grammar navigation and the asset validator) and tools/x/checks.ludic
(the checks + string helpers + a minimal XML well-formedness validator).

`x test-tools` now runs the vocabulary + docs-coverage checks and the
JSON/XML asset validation through Ludic instead of python3; ci.yml's
docs-coverage step calls `x check-impl` / `x check-vocabulary`. Each port was
verified against its former Python script for exact verdict parity on the
clean tree and on injected drift (a removed keyword, a broken grammar
alternation, an undocumented method).

Deletes the superseded scripts: tools/check-vocabulary.py, tools/check-docs.py,
tools/docgen/check-impl.py, tools/docgen/validate.py. The docgen site
generator (gen.py/check.py/palette.py) and the LSP protocol driver
(test-lsp.py) remain and are tracked separately.

Toolchain unchanged (seed byte-identical); `x test` (56) and `x test-tools`
(29) stay green.

Part of #31

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
This commit is contained in:
Orkun ÇAKILKAYA 2026-08-31 01:06:38 +03:00
parent c040fff8c8
commit e65e862244
11 changed files with 922 additions and 486 deletions

View file

@ -141,7 +141,7 @@ output. The formatter cannot change what a program means.
The vocabulary is written down in five places that cannot include each other —
the compiler's two tables, `ludic_syntax.h`, the TextMate grammar (JSON), and the
JetBrains lexer (Kotlin). Adding a builtin and forgetting the rest is silent
failure, so `tools/check-vocabulary.py` compares all five and
failure, so `bin/x check-vocabulary` (written in Ludic) compares all five, and
`bin/x test-tools` runs it.
When you add a keyword or builtin: put it in `ludic_syntax.h`, then run