Files
overleaf-cep/libraries/eslint-plugin/french-typography-in-locales.js
T
Antoine ClausseandClaude Opus 4.7 9dde7d3d71 [web] Add ESLint rules for locale JSON (#34221)
* Remove double spaces

* Enforce curly apostrophes

* Enforce French typographic spaces in fr.json

Add vitest test enforcing French spacing rules per
https://typographisme.net/post/Les-espaces-typographiques-et-le-web:
  - NNBSP (U+202F) before ? ! ;
  - NBSP (U+00A0) before :
  - NBSP (U+00A0) around guillemets « »

Fix 38 fr.json values that were using NBSP (or no space) before ? ! ;.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Extend French typography test to numbers and percent signs

Add rules per https://typographisme.net/post/Les-espaces-typographiques-et-le-web:
  - NNBSP (U+202F) between digit groups (thousands separator)
  - NBSP (U+00A0) between a number and a following word/unit
  - NBSP (U+00A0) before %

Fix 9 fr.json values where a regular space was used instead of NBSP between
a number and a following word (e.g. "1 résultat", "10 jetons", "7 jours").

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Use point médian for inclusive writing in fr.json

Replace parenthetical gender/plural markers like "sûr(e)", "abonné(e)",
"licence(s)" with the point médian "·" (e.g. "sûr·e", "abonné·e",
"licence·s") per modern French écriture inclusive.

Add a vitest rule rejecting "word(suffix)" patterns where suffix is 1-3
letters and there is no space before the opening parenthesis.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Simplify inclusive-writing check to "(e)" substring match

The previous regex (matchAll without the /g flag) would have thrown at
runtime. Drop matchAll entirely and use includes('(e)').

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Add point médian to participles in fr.json

Apply inclusive writing (·e) to past participles agreeing with the
addressed user ("affilié·e", "invité·e", "connecté·e", "abonné·e", etc.).

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Use inclusive forms in fr.json

* Change "de Overleaf" to "d’Overleaf"

* Fix tests

* Check for consecutive spaces, not just double spaces

* Update comment about guillemets

* Add ESLint rules for locale JSON conventions

Add three rules to @overleaf/eslint-plugin that run against locale JSON
files via @eslint/json:

- no-consecutive-spaces-in-locales
- no-straight-apostrophes-in-locales
- french-typography-in-locales

Wire them up in services/web/eslint.config.mjs against `locales/*.json`
(and `locales/fr.json` for the French-typography rule).

* Add `@eslint/json` so JSON files can be parsed

* Scope JS/TS-specific config to JS/TS files only so that built-in rules aren't applied to JSON files

* Disable `no-irregular-whitespace` in french-typography-in-locales.js

* Make the rules fixable

* Improve regexes: user `\s` instead of custom matchers

* Add sorted-keys-in-locales ESLint rule

Mirrors scripts/translations/sort.js as a fixable ESLint rule:
require top-level keys in locale JSON files to be sorted alphabetically.
The autofix re-serializes the file with sorted keys.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Add no-unused-locale-keys ESLint rule

Mirrors scripts/translations/cleanupUnusedLocales.js as a fixable ESLint
rule applied only to en.json. Shells out to `find` + `cat` once per
process (cached) to grep all candidate source files for key references,
then reports any en.json key that isn't found. The autofix re-serializes
en.json without the unused keys.

Hardcoded ALWAYS_USED set covers keys built at runtime (month names,
third-party-datastore notifications) that the regex scan can't see.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Add locale-variables-match-en ESLint rule

Mirrors scripts/translations/checkVariables.js as an ESLint rule applied
to non-en locale JSON files. Reads en.json (cached) and reports any
member whose `__name__` placeholders differ from the base.

GLOBALS (e.g. __appName__) are allowed as extras; IGNORE_NESTING_FOR
preserves the script's per-key exceptions.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Replace some double linebreak (@overleaf/no-consecutive-spaces-in-locales)

* Add no-orphan-locale-keys ESLint rule

Detect translation keys in non-en locale files that aren't present in
en.json. Mirrors the `--sync-non-en` flag of cleanupUnusedLocales.js.
The autofix re-serializes the locale file without the orphan keys,
sorted alphabetically.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Add tests for the locale JSON ESLint rules

Cover the four single-file rules:
  - no-consecutive-spaces-in-locales
  - no-straight-apostrophes-in-locales
  - sorted-keys-in-locales
  - french-typography-in-locales (including the inclusive-writing
    suggestion for "(e)")

The three cross-file rules (no-unused-locale-keys, no-orphan-locale-keys,
locale-variables-match-en) read from disk and aren't covered here; they
need integration tests via fixtures.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Drop bin/lint_locales — checks are now covered by ESLint rules

The four checks in bin/lint_locales are duplicated by the new
@overleaf/eslint-plugin rules:

  - sort.js --check                  → sorted-keys-in-locales
  - cleanupUnusedLocales.js --check  → no-unused-locale-keys
                                     + no-orphan-locale-keys
  - checkVariables.js                → locale-variables-match-en
  - single-quote grep                → no-straight-apostrophes-in-locales

`yarn lint` runs the ESLint rules against locales/*.json, so the
lint_locales script and Makefile target are redundant.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Remove Locales.test.mjs

* Delete scripts replaced by @overleaf/eslint-plugin rules

Remove the three locale scripts now superseded by ESLint rules:

  - scripts/translations/sort.js                   → sorted-keys-in-locales
  - scripts/translations/cleanupUnusedLocales.js   → no-unused-locale-keys
                                                   + no-orphan-locale-keys
  - scripts/translations/checkVariables.js         → locale-variables-match-en

Also drop the matching Makefile targets (sort_locales,
cleanup_unused_locales), update AGENTS.md / copilot-instructions.md to
point at `yarn lint --fix` instead, and prune the deleted entries from
the esm-check-migration allowlist.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Delete redundant scripts/translations/sort-locale-file.js

Single-file sort utility — superseded by `yarn lint --fix path/to/file.json`
via the sorted-keys-in-locales ESLint rule. No callers in repo.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Update comments and messages

* Add test file pattern to ESLint configuration

* Reorder "Backend specific rules" so `no-extraneous-dependencies` doesn't fail on test files

* Make french-typography fixes idempotent and catch edge cases

Apply Copilot's review feedback on PR #34221:

  - autoFix() now uses `\\s*`/`\\s+` instead of `\\s?`/`\\s` so it
    collapses any run of mixed whitespace into the correct NBSP/NNBSP,
    making the fix idempotent and order-independent.
  - The `:` lint regex now captures optional whitespace and reports
    when it isn't NBSP (catches `"Liste :"` with a regular space).
  - The `%` lint regex now matches the missing-space case (`"10%"`)
    via `\\s*` instead of requiring one char between digit and `%`.

Co-Authored-By: Claude Opus 4.7 <noreply@anthropic.com>

* Remove sort-locale-file.js from pre-commit

* Revert no-unused-locale-keys ESLint rule; Restore lint_locales scripts

---------

Co-authored-by: Claude Opus 4.7 <noreply@anthropic.com>
GitOrigin-RevId: 8351fab074e6e7f77920b70237df1cd1c781e053
2026-06-10 08:06:44 +00:00

161 lines
4.7 KiB
JavaScript
Raw Blame History

This file contains invisible Unicode characters
This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
// French typographic spacing rules for JSON string values.
// Per https://typographisme.net/post/Les-espaces-typographiques-et-le-web
// - NNBSP before ? ! ;
// - NBSP before :
// - NBSP inside guillemets (after «, before ») — keeps the
// guillemet attached to its content so it cannot be orphaned at a
// line break
// - NNBSP between digit groups (thousands separator)
// - NBSP between a number and a following word/unit (50 Mo, 5 jours)
// - NBSP before %
// - point médian "·" instead of parenthetical (e) for inclusive writing
const NBSP = ' '
const NNBSP = ' '
function autoFix(value) {
let v = value
// Normalize whitespace before ? ! ; to NNBSP
v = v.replace(/\s*([?!;])/g, `${NNBSP}$1`)
v = v.replace(/([^\s])([?!;])/g, `$1${NNBSP}$2`)
// Normalize whitespace before : (after a letter, not before /) to NBSP
v = v.replace(/(\p{L})\s*:(?!\/)/gu, `$1${NBSP}:`)
// NBSP after «
v = v.replace(/«\s*/g, `«${NBSP}`)
// NBSP before »
v = v.replace(/\s*»/g, `${NBSP}»`)
// NNBSP between digits (thousands separator)
v = v.replace(/(\d)\s+(\d)/g, `$1${NNBSP}$2`)
// NBSP between digit and following letter (number + unit/word)
v = v.replace(/(\d)\s+(\p{L})/gu, `$1${NBSP}$2`)
// NBSP before %
v = v.replace(/(\d)\s*%/g, `$1${NBSP}%`)
return v
}
function lint(value) {
const violations = []
for (const m of value.matchAll(/(.)([?!;])/g)) {
if (m[1] !== NNBSP) {
violations.push({
message: `expected NNBSP before "${m[2]}", got ${JSON.stringify(m[1])}`,
fixable: true,
})
}
}
for (const m of value.matchAll(/(\p{L})(\s*):(?!\/)/gu)) {
if (m[2] !== NBSP) {
violations.push({
message: `expected NBSP before ":" after letter ${JSON.stringify(m[1])}, got ${JSON.stringify(m[2])}`,
fixable: true,
})
}
}
for (const m of value.matchAll(/«(.)/g)) {
if (m[1] !== NBSP) {
violations.push({
message: `expected NBSP after "«", got ${JSON.stringify(m[1])}`,
fixable: true,
})
}
}
for (const m of value.matchAll(/(.)»/g)) {
if (m[1] !== NBSP) {
violations.push({
message: `expected NBSP before "»", got ${JSON.stringify(m[1])}`,
fixable: true,
})
}
}
for (const m of value.matchAll(/(\d)(\s)(\d)/g)) {
if (m[2] !== NNBSP) {
violations.push({
message: `expected NNBSP between digits "${m[1]}${m[3]}", got ${JSON.stringify(m[2])}`,
fixable: true,
})
}
}
for (const m of value.matchAll(/(\d)(\s)(\p{L})/gu)) {
if (m[2] !== NBSP) {
violations.push({
message: `expected NBSP between digit "${m[1]}" and "${m[3]}", got ${JSON.stringify(m[2])}`,
fixable: true,
})
}
}
for (const m of value.matchAll(/(\d)(\s*)%/g)) {
if (m[2] !== NBSP) {
violations.push({
message: `expected NBSP before "%" after "${m[1]}", got ${JSON.stringify(m[2])}`,
fixable: true,
})
}
}
if (value.includes('(e)')) {
violations.push({
message:
'expected point médian "·" instead of "(e)" for inclusive writing',
fixable: false, // suggestion-only
})
}
return violations
}
module.exports = {
meta: {
type: 'problem',
fixable: 'code',
hasSuggestions: true,
docs: {
description:
'Enforce French typographic spacing in JSON string values (NNBSP before ? ! ;, NBSP before : and inside « », etc.).',
},
schema: [],
},
create(context) {
return {
Member(node) {
if (node.value.type !== 'String') return
const original = node.value.value
const violations = lint(original)
if (violations.length === 0) return
const fixed = autoFix(original)
const replacement = JSON.stringify(fixed)
const inclusiveSuggestion = original.includes('(e)')
? JSON.stringify(original.replace(/\(e\)/g, '·e'))
: null
for (const v of violations) {
context.report({
node: node.value,
message: v.message,
...(v.fixable && fixed !== original
? {
fix(fixer) {
return fixer.replaceText(node.value, replacement)
},
}
: {}),
...(!v.fixable && inclusiveSuggestion
? {
suggest: [
{
desc: 'Replace "(e)" with "·e"',
fix(fixer) {
return fixer.replaceText(
node.value,
inclusiveSuggestion
)
},
},
],
}
: {}),
})
}
},
}
},
}