Skip to content

v0.5.2

Choose a tag to compare

@KaiErikNiermann KaiErikNiermann released this 18 Sep 06:28
· 492 commits to main since this release

What's new

Fixes

  • Prose inside Typst content blocks is checked; #columns(2)[...], #align(center)[...], #figure(caption: [...]) and #emph[...] bodies were dropped whole (9b16779, #87)
  • The first and last word of an inline content block keep their diagnostics (a664cbb)
  • reStructuredText directive bodies are checked; every .. note::, .. warning:: and .. only:: went unread, and figure captions were skipped by type (73b2ced)
  • Org quote and verse blocks, table cells, footnote definitions and #+TITLE: values are checked (700f843)
  • Markdown table cells are checked (afa7f2d)
  • \verb|x|, \lstinline!x! and \mintinline{}|x| arguments are no longer spell-checked as prose (9abfb7a)
  • LaTeX no longer emits overlapping exclusion ranges for adjacent display math (3e9eed7)

New lang ergonomics

  • One gap syntax per language replaces the two scanners that could drift apart. Adding a language is now a single function (3e9eed7)
  • The five per-grammar FFI modules collapse into one vendored_grammars! macro (9c92cdb)
  • Gap scanners match on slice patterns instead of hand-written bounds checks (9149db6, 95f90e4)
  • A composition fuzzer shuffles prose and non-prose fragments into documents and checks what reaches the checker, and that exclusion ranges stay sorted and disjoint (4a1536a, 5db092d)
  • Duplicate helpers collapsed into shared ones, each with a semgrep rule guarding it (670c605, 04cf98d, 14a296a, 2022334, 285d16d)

Documentation

  • The add-a-language guide matches the code again; its TinyLang snippet is the source verbatim (24d24df)
  • Per-language extraction details, the README structure tree and CONTRIBUTING paths corrected (029873e, 76dce96)

Maintenance

  • rand updated past RUSTSEC-2026-0097 (7f52831)
  • just docs bootstraps its venv instead of assuming one exists (a2e8462)

Full Changelog: v0.5.1...v0.5.2