every translation in jsx incorrectly flagged as unused

[kitsunekyo]

i18next-react project setup:

// ./i18n-unused.config.cjs
module.exports = {
  localesPath: 'src/analysis/_locales',
  srcPath: 'src',
};

translation files

./src/analysis/_locales/de.json
./src/analysis/_locales/en.json

(reduced for brevity)

{
  "tab": {
    "progress": "Progress",
  }
}

example file that uses translations

// ./src/analysis/views/TestView.tsx
import { useTranslation } from 'react-i18next';

export const TestView = () => {
  const { t } = useTranslation('analysis');
  return <div>{t('tab.progress')}</div>;
};

i have installed i18n-unused as dev dependency and run yarn i18n-unused display-unused
this prints almost all translations as unused. what i noticed is that it works with translations that are in javascript (like useEffect, or other pure javascript contexts), but it does not seem to work in jsx at all.

[DonatienD]

Experiencing the same issue in my project.
This is also true when you implement the collectUnusedTranslations script (see below as implemented on my side).

const { collectUnusedTranslations } = require('i18n-unused');
const { sync: globSync } = require('glob');

const dictionaries = globSync('./src/i18n/dictionaries/*.json', {realpath: true});
const sources = globSync('./src/**/*.+(tsx|ts|jsx|js|es6)', {realpath: true});

const handleTranslations = async () => {
  const unusedTranslations = await collectUnusedTranslations(
    dictionaries,
    sources,
    {}
  );
  return unusedTranslations;
}

handleTranslations().then(result => { console.log('unused:', result) });

[TheNemus]

Same to me but with .tsx

[halcyonshift]

I had this issue and mostly fixed it with the translationKeyMatcher config:

translationKeyMatcher: /['"](translation|prefixes|like|tab|here)\.([\w\.]+)['"]/gi

[beamery-tomht]

Seems like the default value for translationKeyMatcher, which is /(?:[$ .](_|t|tc|i18nKey))\(.*?\)/gi, isn't working - it gives no results

It's looking for $, space or . characters preceding the t function which isn't always present.

I tried the one provided in the comment above - but, it struggles with several cases:

When there isn't a closing bracket on the same line, they won't get recognised

t('myKey', {
  someToken: 123
})

If the key used is dynamic, they aren't recognised

t(somePredicate ? 'keyTrue' : 'keyFalse')

If keys are substrings of each other, they won't get flagged

For example, given:

{
  "keyA": "x",
  "keyAB": "x",
}

And code of:

t('keyAB')

The tool won't flag that keyA is unused because it's a substring of keyAB.

If other functions match the same structure, they are also captured

e.g.

matchViewport('someString') // matcher sees `t('someString')` here and marks it as a used key

---

I've added a custom translationKeyMatcher that is sort of working....

translationKeyMatcher: /(?:[ ={:]t\(|i18nKey=)'\w+'[,)]?/gi,

Some gotchas to note as this is very basic...

• It assumes only t('key' or i18next='key' are the only ways keys are consumed (as we only use this in our codebase, it optionally looks for either a closing bracket or comma after the key string)
• We only use single quotes so didn't add support for double quotes or backticks
• it doesn't support nested keys, we use a flat key structure
• when keys are substrings, it marks them as false positive
• can't handle dynamic keys still

I've also had to refactor some of my keys to support these gotchas:

-  t(predicate ? 'key1' : 'key2')
+  predicate ? t('key1') : t('key2') 

Making sure my keys aren't substrings

- "myControl": "value",
+ "myControlLabel": "value",
  "myControlTooltip": "tooltip",

[manuelpoelzl]

Having the exactly same issue in a React+TS project

[fredrivett]

after a good chat with my friend chatGPT, we (they) came up with this:

translationKeyMatcher:
  // this regex was generated by chatGPT, and has the following features:
  // * only match for `t()` and `tToDocLanguage()` calls
  // * preceded by a space, opening parenthesis, or opening curly brace
  // * also match for `i18nKey` prop in JSX
  // * captures double quotes, single quotes, and backticks
  // * works with optional chaining e.g. `t?.()`
  // * works with multiline strings e.g. `t(\n"key")`
  /(?<=[\s{(])(?:t|tToDocLanguage)\??\.?\(\s*["'`]?([\s\S]+?)["'`]?\s*\)|i18nKey="([\s\S]+?)"/g,

we have a custom function used in the codebase, tToDocLanguage which we wanted to match on too, but if you don't you can remove the |tToDocLanguage part.

works really well for us, only downside is it doesn't match on dynamic strings (e.g. t(`foo.${bar}`)), so we just have a list of those in our excludeKey array.

here's some test code to verify it works as expected for your use case:

const regex = /(?<=[\s{(])(?:t|tToDocLanguage)\??\.?\(\s*["'`]?([\s\S]+?)["'`]?\s*\)|i18nKey="([\s\S]+?)"/g;
const text = `
  t("1.basic.doublequotes.string")
  t?.("2.basic.doublequotes.string.with.optional.chaining")
  t('3.basic.singlequotes.string')
  t?.('4.basic.singlequotes.string.with.optional.chaining')
  t(\`5.basic.backtick.string\`)
  t?.(\`6.basic.backtick.string.with.optional.chaining\`)
  const multilineString = \`abc \${t(
    "7.multiline.string",
  )}\`;
  tToDocLanguage("8.custom.func.beginning.with.t")
  <Element i18nKey="9.i18nKey.string">
`;

const matches = [...text.matchAll(regex)];
const keys = matches.map(match => match[1] || match[2]);

// output each key on a new line
keys.forEach(key => console.log(key));

// outputs:
//   '1.basic.doublequotes.string'
//   '2.basic.doublequotes.string.with.optional.chaining'
//   '3.basic.singlequotes.string'
//   '4.basic.singlequotes.string.with.optional.chaining'
//   '5.basic.backtick.string'
//   '6.basic.backtick.string.with.optional.chaining'
//   '7.multiline.string",'
//   '8.custom.func.beginning.with.t'
//   '9.i18nKey.string'

hope that helps some folks!

[mhyassin]

I've adjusted the regex a bit to also include t methods where you use variables,

/t\(\s*["'`]?([\s\S]+?)["'`]?\s*(?:\)|,)|i18nKey="([\s\S]+?)"/gi

Now this would be matched correctly as well

t(`translation.key`, { variable: x })