[renovate]

This PR contains the following updates:

Package	Change	Age	Confidence
i18next-cli	1.74.1 → 1.80.4

---

Release Notes

i18next/i18next-cli (i18next-cli)

v1.80.4

Compare Source

• fix(rename-key): calls with a string default value are renamed in the source. t('key', 'Default') and
  t('key', 'Default', { … }) were left untouched while the translation files were renamed, so the next extract
  brought the old key back. Moving such a key to another namespace keeps the default value:
  t('key', 'Default', { ns: 'ns2' }).
• fix(extract): plural forms of a key stored flat stay flat in secondary locales. With "Loading..._one" in
  ru/translation.json, extract also wrote a nested copy, { "Loading": { "": { "": { "_one": … } } } }; the same
  happened for generated forms (_few, _many), preserved _zero forms, context variants and preservePatterns.
  A file that already has these nested copies loses them on the next extract.

v1.80.3

Compare Source

• fix(rename-key): keys the extractor stores flat are renamed in the translation files. Renaming "Loading..." to
  "Saving..." left the old key in place and wrote { "Saving": { "": { "": { "": "Loading..." } } } }; renaming
  common.ok to "Okay, thanks." wrote { "Okay, thanks": { "": "OK" } }. The new key is now written where the next
  extract puts it (flat for "Loading...", natural-language and <Trans> keys), and the old key is removed however
  the file stores it. #​311

v1.80.2

Compare Source

• fix(sync): flat keys that contain the key separator keep their translations. The extractor writes keys like
  "Loading..." or "Hello world. Bye" as one flat key, but sync split them on the separator, so a secondary file
  with { "Loading...": "Lädt..." } became { "Loading": { "": { "": { "": "" } } } } and the translation was lost.
  sync now writes them flat, and status, localize and rename-key find their values too. A secondary file
  already broken this way gets its flat key back (with an empty value) on the next sync.
  #​310
• fix(types): placeholders with spaces on one side only ({{ name}}) are trimmed, and enableSelector no longer
  throws on plural keys whose base name is constructor, toString, … or also an object
  (i18next-resources-for-ts 2.1.2).

v1.80.1

Compare Source

• fix(extract): a <Trans> interpolation variable named format is extracted. <Trans>Export as {{ format }}</Trans>
  was extracted as Export as , because react-i18next read format as the formatter name and dropped the variable;
  it is now Export as {{format}}, as react-i18next 17.1.8 renders it. Requires react-i18next ^17.1.8.
• chore(deps): update dependencies within their ranges and neostandard to 0.14.

v1.80.0

Compare Source

• feat(lint): in the i18next format, the translation files of every locale are compared with the primary language:
  interpolations by name ({{name}}, {{- name}}, {{val, number}}), nested $t(key) references and tags (<0>,
  <1/>, <bold>), except the basic HTML react-i18next keeps (transKeepBasicHtmlNodesFor). A translation that
  renamed or dropped a placeholder is reported with file, key and line, e.g.
  greeting: missing {{name}}, unexpected {{nom}} (compared with "en"). Plural forms are compared as a group, so
  {{count}} may be left out of _one or _zero. A warning by default, so existing CI runs keep passing;
  lint.checkTranslationPlaceholders: 'error' fails the run, 'off' disables it. Until now only ICU and MF2 projects
  had this check.
• feat(status): values identical to the primary language (English copied into de) are marked
  ≈ (identical to en) and counted in the summary. status.checkIdentical: 'warn' (default) still counts them as
  translated and keeps the exit code, 'error' counts them as untranslated, 'off' disables it. Locales of the
  primary's language (en-GB with en), values without letters, URLs, e-mail addresses and short all-caps words
  (OK, FAQ) are skipped; other values that stay the same go into status.ignoreKeys.
• feat(lint): lint --update-baseline writes the current issues to i18next-lint-baseline.json (lint.baseline for
  another path). While the file exists, lint reports and fails only on new issues, so a codebase with existing
  hardcoded strings can run lint in CI. Issues are matched by file, type and text, not by line.
• fix(lint): interpolations are read like i18next does. The unescaped {{-name}} was read as a parameter -name, so
  t('Hi {{-name}}', { name }) reported two false interpolation issues; {{- name}} and names outside [A-Za-z0-9_.-]
  ({{größe}}) were not read at all.
• docs(types): transKeepBasicHtmlNodesFor defaults to react-i18next's ['br', 'strong', 'i', 'p'].
• docs(skill, localize): after translating by hand, run lint to check the placeholders.

v1.79.1

Compare Source

• fix(init, migrate-config): the default source glob includes .mjs (src/**/*.{js,jsx,mjs,ts,tsx}), in init, in the
  config detected for a project without one, and in migrate-config when the old config has no input. Existing
  configs do not change. #​309

v1.79.0

Compare Source

• feat(extract, instrument, lint): a .mjs file with JSX is parsed with the same JSX fallback as a .js file,
  including the constants pre-scan of extract. Before, it failed with a syntax error and extract exited with 1.
  #​309

v1.78.0

Compare Source

• feat(lint): with messageFormat: 'mf2', lint warns when whitespace at the start or end of a pattern differs from
  the primary language. MF2 keeps a space inside {{ }} as text, so {{ {$count} files}} shows it in the app. A
  warning does not fail the run.
• fix(lint): a language code with _ (ru_RU, pt_BR) is read as ru-RU, pt-BR for the plural categories. Before,
  Intl rejected it and only other was required, so a ru_RU plural with only one and other passed.
• fix(lint): the required plural categories are other plus every category a whole number below one million selects
  (was 0 to 1000), the same rule as Locize; exact selectors count for a category only when it selects a few small
  numbers. No locale's result changes with today's CLDR data.
• docs(skill): the ICU and MF2 reference names the common MF2 plural mistakes: other instead of *, {{ {$count}
  instead of {{{$count}, and spaces inside {{ }}.

v1.77.0

Compare Source

• feat(types): with extract.messageFormat: 'icu' or 'mf2', a newly generated i18next.d.ts sets
  parseInterpolation: 'icu' or 'mf2' when the project's i18next is 26.5.0 or later, so t() options are type-checked
  against the arguments of each message. With i18next 26.2.0 to 26.4.x it still sets parseInterpolation: false. The
  version is read from node_modules/i18next/package.json, from the working directory upwards (hoisted monorepo
  installs count); without an installed i18next it stays false. The warning for an existing i18next.d.ts without
  parseInterpolation suggests the value for your i18next version, and an existing file that still sets false
  although your i18next supports the typed value gets a hint.
• docs(skill): the ICU and MF2 reference describes parseInterpolation: 'icu' / 'mf2'.

v1.76.0

Compare Source

• feat: new extract.messageFormat option, 'i18next' (default), 'icu' (i18next-icu)
  or 'mf2' (i18next-mf2), for the syntax of your translation values.
• feat(extract): with messageFormat: 'icu' or 'mf2', extract writes one key per message. t('photos', '{count, plural, ...}', { count }) gave photos_one and photos_other (plus _few and _many for languages like Polish),
  keys i18next-icu and i18next-mf2 never read, and context gave _male keys; now it gives photos. $t() in a value
  is no longer read as nesting, and the children of <Trans> are written with {name} (ICU) or {$name} (MF2)
  instead of {{name}}, which ICU and MF2 rendered as text.
• feat(extract): IcuTrans and IcuTransWithoutContext are extracted, also when imported under another name (key from
  i18nKey, or defaultTranslation when there is none), and so are the Trans, Plural, Select and SelectOrdinal
  components of react-i18next/icu.macro, with the message the macro builds (<Trans>Welcome, <strong>{name}</strong>!</Trans> gives Welcome, <0>{name}</0>!). Before, IcuTrans was
  not extracted, and the macro <Trans> was read as react-i18next Trans and written as
  Welcome, <strong>{{name}}</strong>!, which the compiled IcuTrans rendered as text.
• feat(status): with messageFormat: 'icu' or 'mf2', a plural or select message counts as one key, so status no
  longer reports missing _few and _many keys.
• feat(instrument): with messageFormat: 'icu' or 'mf2', instrument writes {name} / {$name} placeholders, a
  count === 1 ? … : … ternary becomes one plural message instead of defaultValue_one / defaultValue_other, and
  the i18n init file it creates registers i18next-icu (with intl-messageformat in the install hint) or i18next-mf2.
• feat(extract, instrument): with messageFormat: 'icu' or 'mf2', the text around converted placeholders (Trans values, instrument) is
  escaped: File '{{name}}' not found becomes File ''{name}' not found, because an apostrophe before { starts quoted
  text in ICU and '{name}' would render as the text {name}; literal braces are quoted (ICU) or escaped (MF2).
• feat(lint): with messageFormat: 'icu' or 'mf2', lint parses every value in the translation files of all
  locales and reports, with file, key and line (JSON files):
  • syntax errors with the parser's message (greeting: ICU syntax error: EXPECT_ARGUMENT_CLOSING_BRACE at column 7),
    with a hint when the value contains i18next's {{
  • argument names that differ from the primary language, so a renamed or missing {name} is caught (# is not an argument)
  • plural categories that do not fit the locale, for plural and selectordinal (MF2: .match on :number or
    :integer). Allowed are the locale's CLDR categories, required are other plus every category a whole number from
    0 to 1000 selects: ru needs one, few and many, while fr, es, it and pt may leave out many (only from
    1,000,000) and cs its many (fractions only), and zero in en is reported as never used. Exact selectors like
    =0 are allowed and count for a category when they cover every whole number it is used for (=1 covers one in
    en, not in ru, where one is also 21, 31, ...). For a locale Intl does not know only other is required.
• fix(lint): with messageFormat: 'icu' or 'mf2', the t() interpolation check reads the arguments with the same
  parser. ICU braces like male {{name}} are no longer taken for {{name}} interpolation (which reported the gender
  passed to a select as unused), and t('greeting', 'Hello {name}', { nmae }) is reported.
• feat(types): with messageFormat: 'icu' or 'mf2', a newly generated i18next.d.ts sets parseInterpolation: false
  (i18next 26.2.0+), so ICU and MF2 plurals no longer break the type-safe t. For an existing file without it, types
  prints a warning.
• docs(skill): the i18next-localization skill has an ICU and MF2 section: which format to choose, setup, what silently
  stops working, IcuTrans, parseInterpolation: false and extract.messageFormat.
• deps: @formatjs/icu-messageformat-parser (the parser i18next-icu uses through intl-messageformat) and messageformat
  (the MF2 implementation i18next-mf2 uses). They are only loaded when messageFormat is 'icu' or 'mf2'.

v1.75.3

Compare Source

• fix(extract): options passed in a variable are read like inline ones. With const params = { count },
  t('key', params) now extracts the plural keys (and reads context, ns, defaultValue the same way),
  where it used to extract only key. A variable a file declares also no longer resolves to a same-named
  variable of another file: t('key', overwrittenParams) took y=true as its default value from an
  unrelated function-local overwrittenParams array elsewhere. Fixes
  #​307.

v1.75.2

Compare Source

• fix(extract): declarations with the same name in different files no longer overwrite each other.
  Two generated OpenAPI clients that both export components resolved to whichever file was read first,
  so import { components as githubComponents } from './github-api' and
  import { components as digitalOceanComponents } from './digital-ocean-api' extracted the keys of only one
  of them, and moving files around changed which. A name now resolves to the module the file imports it
  from, for renamed and plain imports, namespace imports (GH.components), .js import specifiers and
  tsconfig paths aliases. Imports that cannot be traced to a scanned file (packages, re-exporting barrels)
  still resolve by name as before. Fixes #​306.

v1.75.1

Compare Source

• fix(extract): types and values read through a namespace import resolve like named imports. With
  import * as Types from './types', a qualified type ((color: Types.Color) => …, x as Types.Color,
  FC<Types.Props>, declare const s: Types.Status) and a namespace member (Types.SIZES.map(…),
  Types.Status.Done, t(Types.LABELS.save), Object.keys(Types.LABELS)) now extract their keys, also
  for a namespace re-exported from a barrel (export * as Shared from './types'). Fixes
  #​305.
• fix(extract): an exported as const object resolves in the files that import it: t(LABELS.save) and
  Object.keys(LABELS) only worked in the declaring file before.
• fix(extract): a value typed keyof typeof MAP resolves to the map's keys, so
  t(`label.${key}`) with declare const key: keyof typeof LABELS extracts label.save and
  label.cancel. This branch never matched before.

v1.75.0

Compare Source

• feat(extract): new extract.maxTypeDepth option sets how many levels of nested object members are
  followed when a dynamic key reads a typed value (default: 5). Higher values cost time and memory on
  large recursive types such as OpenAPI schemas. Requested in
  #​302.
• fix(extract): the depth now counts from the type an indexed access picks instead of from the declared
  wrapper type. type Pet = components['schemas']['Pet'] gets the full depth below Pet, where the
  components/schemas wrapper used to take two of the five levels.
• fix(extract): numeric keys resolve in object types and indexed access types, so openapi-typescript
  response types work:
  operations['getPet']['responses'][200]['content']['application/json'] and
  op.responses[200].content['application/json'].status resolved nothing before.

v1.74.8

Compare Source

• fix(extract): a commented template literal no longer extracts its placeholder as a key.
  // t(`QUARTER.${q}`) extracted QUARTER.${q} and // t(`status.${deepObject.status}`) extracted
  status.${deepObject.status}. Such a comment now keeps the existing keys it matches (QUARTER.1,
  QUARTER.2, …) even with removeUnusedKeys, the way a commented static key keeps its key, while code can
  still add new matching keys. A pattern without any static part (`${a}.${b}`) is ignored. Fixes
  #​304.

v1.74.7

Compare Source

• fix(extract): a selector i18nKey on <Trans t={t}> now routes into the hook's namespaces the same way
  t() does, matching react-i18next 17.0.16. With const { t } = useTranslation(['ns1', 'ns2']),
  <Trans t={t} i18nKey={$ => $.ns2.message} /> was filed as ns1:ns2.message; it is now filed as
  ns2:message, like t($ => $.ns2.message). Under types.enableSelector: 'strict' a leading primary
  namespace is stripped too (useTranslation('only') with $ => $.only.key files only:key). Reported in
  react-i18next#1933.
• fix(extract): the props of a component typed React.FC<Props> (also FC, FunctionComponent),
  forwardRef<Ref, Props>(…) or memo<Props>(…) are typed by Props, and iterating a typed member binds
  the element. props.colors.map((color) => t(\Color.${color}`))withcolors: Color[]extracted nothing before unlesscolorwas asserted withas Color. Arrays of objects work the same way: props.items.map(({ kind }) => …), for (const item of props.items)and a destructured({ items }) => items.map(…)`. Fixes #​303.

v1.74.6

Compare Source

• fix(extract): iterating an inline array literal binds the element like iterating a named constant
  does. ([1, 2, 3, 4] as const).map((q) => t(\QUARTER.${q}`)), for (const unit of ['day', 'hour'] as const)`
  and inline arrays of objects extracted nothing before. Fixes
  #​301.
• fix(extract): members of nested object types resolve. t(\barStatus.${deepObject.Bar.status}`)andt(`fooBarStatus.${deepObject.Foo.Bar.status}`) now extract the nested union's values, also through references to other types, optional chaining (pet.category?.kind), destructuring (const { kind } = pet.category) and indexed access types (DeepObject['Bar']['status'], components['schemas']['Pet']), so types generated by openapi-typescript work. Nested members are followed up to five levels below the declared type, as in components.schemas.Pet.category.kind`.
  Fixes #​302.

v1.74.5

Compare Source

• fix(extract): a value typed as an array of a string union (Color[], readonly Color[],
  Array<Color> or an alias of one) now resolves to the union's values, so
  colors.map((color) => t(\color.${color}`))andfor (const color of colors)extract every key instead of nothing. Before, this only worked withcolor as Color`. Fixes
  #​300.

v1.74.4

Compare Source

• fix(extract): a conditional namespace argument resolves to the namespaces in its branches.
  useTranslation(confirmEmail ? ['reset-email-page', 'reset-page'] : 'reset-page') filed
  t('form.title') under defaultNS, so status reported it as absent and extract removed the
  existing translations from reset-page and reset-email-page. The key is now extracted into the
  primary namespace of every branch (here reset-email-page and reset-page), also for
  <Trans t={t}>. Fixes #​299.

v1.74.3

Compare Source

• fix(extract): dynamic keys typed by a declaration in a file without any translation call
  resolve again when that declaration builds on one from a file that sorts after it. In 1.74.2,
  type Kind = Base | 'extra' in api/types.ts with Base declared in shared/base.ts expanded
  t(`kind.${kind}`) to kind.extra only, and the same happened to interface members,
  function return types and functions returning an enum member declared further down the list.
  Such declarations are now resolved again once every file has been scanned, which also removes
  the file-order dependence this had before 1.74.2.

• feat(extract): more implicitly typed values expand to their finite set of keys. Fixes
  #​298:

  • members of an object returned by a function with an object return type:
    const animal = getAnimal(); t(`animal.${animal.type}`), also through
    const { type } = getAnimal(), const kind = animal.type or getAnimal().type.
  • unions and intersections of object types (type Animal = Duck | Dog gives animal.type the
    values of both) and generic object types (type AnimalBase<T extends AnimalType> = { type: T }
    used as AnimalBase<'DUCK'>, or bare, which falls back to the parameter's default or
    constraint).
  • (typeof COLOR)[keyof typeof COLOR] for an as const object or an enum.
  • elements of an array, colors[i], without an annotation on the variable they are assigned to.

  Narrowing is not followed: after if (color !== 'BLUE'), color.BLUE is still extracted.

• fix(extract): shorthand options resolve like their long form. t('k', { ns }) filed the key
  under the default namespace, { defaultValue } was ignored and
  useTranslation('ns', { keyPrefix }) lost its prefix. Thanks to
  @​aaronshaf for reporting it in
  #​291.

v1.74.2

Compare Source

• fix(extract): the output no longer depends on the order in which glob happens to return the
  input files. When two call sites give one key different defaults (t('k', 'Adding...') in one
  file, t('k', 'Importing...') in another), the first file visited wins, so an unchanged source
  tree could produce a different catalog from run to run and flip CI checks that compare against a
  committed catalog. Files are now visited in sorted path order, which makes the winner stable. If
  your sources contain such conflicts, the kept default may change once with this release. To
  hear about them instead of silently keeping one default, set
  extract.warnOnConflicts to 'warn' or 'error'. Fixes
  #​296.
• perf(extract): files that contain no translation call, hook, TFunction, getFixedT or Trans
  component (as configured via functions, useTranslationNames and transComponents) are no
  longer parsed and walked a second time for keys. The constants pre-scan still reads every file,
  so types and as const values declared in such files keep resolving dynamic keys elsewhere.
  Plugins with an onVisitNode hook still see every file. On large trees where most files hold no
  translations this cuts extraction time considerably. Fixes
  #​297.
• chore: update dependencies

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • Between 12:00 AM and 03:59 AM, only on Monday (* 0-3 * * 1)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[github-actions]

🎉 This pull request is included in Gitify v8.0.0! 🎉

Downloads for macOS, Windows, and Linux are available on the release page.

Thanks for helping make Gitify better!