@codelaw/rules-spell

codelaw's spell checker — one of the first-party rule packs (SR-RULE-PACKS), on npm and PyPI (codelaw-rules-spell). It reads the prose of every L3 language from the tree the engine parsed: comments and docstrings, prose strings, Markdown and HTML text, the string values of JSON, YAML and TOML, and the names a file declares — identifiers, keys, classes and ids. It is codelaw's own code and codelaw's own word data, generated from SCOWL (SR-SPELL); it does not wrap CSpell.

codelaw init never adds it: a team installs it on purpose.

# codelaw.toml
extends = ["codelaw:spell/recommended"]
preview = true   # every rule is `preview` until a corpus run promotes it

npm i -D @codelaw/rules-spell      # or
pip install codelaw-rules-spell

Presets

| preset | rules | ported from | |---|---|---| | spell/recommended | 3 | codelaw's own: known typos from a list generated from SCOWL, fixed unattended | | spell/strict | 6 | spell/recommended plus the dictionary check, one spelling per word across the project and the [spell] words check |

Rules

| rule | fix | flags | |---|---|---| | spell/known-typo | auto | a word on the known-typo list in a comment, a doc, a prose string or Markdown | | spell/known-typo-identifier | auto | a known typo in a JS, TS or Python name where it is declared; renamed only when the name is provably file-local | | spell/known-typo-key | manual | a known typo in a JSON, YAML or TOML key, a CSS or SCSS class, id or custom property, or an HTML id | | spell/unknown-word | suggest | a word of prose in no word list; the nearest dictionary word as the edit, up to three named (spell/strict) | | spell/locale-variant | auto | the other locale's spelling (colour under en-US), while [spell] locale is set (in no preset) | | spell/inconsistent-term | suggest | one word spelled two ways across the project (canceled / cancelled); the spelling the project uses most as the edit (spell/strict, a project rule) | | spell/suspicious-word | manual | a [spell] words entry that is a known typo or one edit from a common word (spell/strict) |

Config

# codelaw.toml (the root config only)
[spell]
locale = "en-US"            # or "en-GB"; spell/locale-variant reports the other one
words = ["kubectl", "Acme"] # accepted by every spell/* rule; spell/suspicious-word checks them

[rules]
"spell/locale-variant" = "warn"

The word data

data/spell/ is written by scripts/spell-data.py and committed; the same sources write the same bytes: typos.json (188,138 generated known typos), words.json (92,758 words: SCOWL's US and British English to size 60, and the software terms) and locale.json (VarCon's one-to-one US and British pairs). Sources and their notices: NOTICE.