Working with LinguiJS CLI¶

@lingui/cli provides the lingui command for extracting, merging and compiling message catalogs. Follow setup instructions to install required packages.

Note

This tutorial assumes you use yarn to run commands. If you use npm, type npm run <command> instead of yarn <command>.

Extracting messages¶

We’re going to use an app we built in a React tutorial. The extract command looks for messages in the source files and extracts them:

yarn extract

Extracting messages from source files…
Collecting all messages…
Writing message catalogs…
Messages extracted!

Catalog statistics:
┌──────────┬─────────────┬─────────┐
│ Language │ Total count │ Missing │
├──────────┼─────────────┼─────────┤
│ cs       │     40      │   40    │
│ en       │     40      │   40    │
└──────────┴─────────────┴─────────┘

(use "yarn extract" to update catalogs with new messages)
(use "yarn compile" to compile catalogs for production)

The message catalog will look like this:

{
  "Message Inbox": "",
  "See all <0>unread messages</0> or <1>mark them</1> as read.": "",
  "{messagesCount, plural, one {There's {messagesCount} message in your inbox.} other {There're {messagesCount} messages in your inbox.}}": "",
  "Last login on {lastLogin,date}.": "",
}

It’s in a JSON dictionary, where ‘key’ is message ID and value is an object with some relevant information: translation, defaults and origin for the message.

This catalog is ready for translation. Let’s translate it into Czech by filling the translation fields:

{
  "Message Inbox": "Přijaté zprávy",
  "See all <0>unread messages</0> or <1>mark them</1> as read.": "Zobrazit všechny <0>nepřečtené zprávy</0> nebo je <1>označit</1> jako přečtené.",
  "{messagesCount, plural, one {There's {messagesCount} message in your inbox.} other {There're {messagesCount} messages in your inbox.}}": "{messagesCount, plural, one {V příchozí poště je {messagesCount} zpráva.} few {V příchozí poště jsou {messagesCount} zprávy. } other {V příchozí poště je {messagesCount} zpráv.}}",
  "Last login on {lastLogin,date}.": "Poslední přihlášení {lastLogin,date}",
}

If we run the extract command again, we can see in the stats that all messages are translated:

Catalog statistics:
┌──────────┬─────────────┬─────────┐
│ Language │ Total count │ Missing │
├──────────┼─────────────┼─────────┤
│ cs       │      4      │    0    │
│ en       │      4      │    4    │
└──────────┴─────────────┴─────────┘

Messages extracted!

(use "yarn extract" to update catalogs with new messages)
(use "yarn compile" to compile catalogs for production)

extract merges all translations with new messages, so you can run this command any time without worrying about losing any translations.

Preparing catalogs for production¶

Once we have all catalogs ready and translated, we can compile the JSON into a minified JS file with the compile command. This command parses the messages in MessageFormat and compiles them into simple functions. It also adds plural rules to a production ready catalog:

yarn compile

Compiling message catalogs…
Done!

The locale directory now contains the source catalogs (messages.json) and the compiled ones (messages.js).

It is also possible to merge the translated catalogs into a single file per locale by specifying catalogsMergePath. For example if catalogsMergePath is assigned locales/{locale} then catalogs will be compiled to /locales/cs.js and /locales/en.js.

Cleaning up obsolete messages¶

By default, the extract command merges messages extracted from source files with the existing message catalogs. This is safe as we won’t accidentally lose translated messages.

However, sooner or later some messages will be removed from the source. We can use the --clean option to clean up our message catalogs:

yarn extract --clean

Validation of message catalogs¶

It might be useful to check if all messages were translated (e.g: in a continous integration runner). The compile command has a --strict option, which does exactly that.

The example output might look like this:

yarn compile --strict

Compiling message catalogs…
Error: Failed to compile catalog for locale en!
Missing 42 translation(s)

Configuring source locale¶

We see that checking for missing translations has one drawback – English message catalog doesn’t require any translations because we’re using English in our source code!

Let’s fix it by setting sourceLocale in package.json:

{
   "lingui": {
      "sourceLocale": "en"
   }
}

Running lingui extract again shows the correct statistics:

Catalog statistics:
┌─────────────┬─────────────┬─────────┐
│ Language    │ Total count │ Missing │
├─────────────┼─────────────┼─────────┤
│ cs          │      4      │    0    │
│ en (source) │      4      │    -    │
└─────────────┴─────────────┴─────────┘

And compilation in strict mode no longer throws an error:

yarn compile --strict

Compiling message catalogs…
Done!

If you use natural language for message IDs (that’s the default), set sourceLocale. You shouldn’t use this config if you’re using custom IDs (e.g: Component.title).

Pseudolocalization¶

There is built in support for pseudolocalization <https://en.wikipedia.org/wiki/Pseudolocalization>. Pseudolocalization is a method for testing the internationalization aspects of your application by replacing your strings with altered versions and maintaining string readability. It also makes hard coded strings and improperly concatenated strings easy to spot so that they can be properly localized.

Example: Ţĥĩś ţēxţ ĩś ƥśēũďōĺōćàĺĩźēď

To setup pseudolocalization add pseudoLocale in package.json:

{
   "lingui": {
      "pseudoLocale": "pseudo-LOCALE"
   }
}

pseudoLocale option can be any string examples: en-PL, pseudo-LOCALE, pseudolocalization or en-UK

PseudoLocale folder is automatically created based on configuration when running yarn extract command. Pseudolocalized text is created on yarn compile command. The pseudolocalization is automatically created from default messages. It can also be changed by setting translation in message.json into non-pseudolocalized text.

How to switch your browser into specified pseudoLocale¶

We can use browsers settings or extensions. Extensions allow to use any locale. Browsers are usually limited into valid language tags (BCP 47). In that case, the locale for pseudolocalization has to be standard locale, which is not used in your application for example zu_ZA Zulu - SOUTH AFRICA

Chrome: a) With extension (any string) - https://chrome.google.com/webstore/detail/quick-language-switcher/pmjbhfmaphnpbehdanbjphdcniaelfie b) Without extension - chrome://settings/?search=languages

Firefox: a) With extension (any string) - https://addons.mozilla.org/en-GB/firefox/addon/quick-accept-language-switc/?src=search b) Without extension - about:preferences#general > Language

Catalogs in VCS and CI¶

The locale/_build folder and locale/*/*.js (compiled catalogs) are safe to be ignored by your VCS. What you do need to keep in VCS are the json files (locale/*/*.json) that contain the messages for translators. The JavaScript functions that return the actual translations when your app runs in production are created from those json files. See Excluding build files guide for more info.

If you’re using a CI, it is a good idea to add the yarn extract and yarn compile commands to your build process.