Radically simple gettext tokens extraction tool for:
- HTML
- Jade/Pug
- Javascript/ES7+
- Vue
- TypeScript (see Known Issues)
- Nativescript Vue with native and web shared code
files.
Also ships with a PO-to-JSON converter.
You can install the easygettext package by running
npm install --save-dev easygettext
or
yarn add --dev easygettext
Simply invoke the tool on the templates you want to extract a POT dictionary template from. The optional '--output' argument enables you to directly output to a file.
gettext-extract --output dictionary.pot foo.html bar.pug component.vue sourcefile.js
CLI usage:
gettext-extract [--attribute EXTRA-ATTRIBUTE] [--filterPrefix FILTER-PREFIX] [--output OUTFILE] [--parser auto|acorn|babel] <FILES>
It recognizes the following token flavours (currently; feel free to extend it!)
<div translate>Hello World</div>
<div translate translate-context="According to...">Hello World</div>
<div translate translate-comment="My comment...">Hello World</div>
<div translate translate-plural="Hello worlds">Hello World</div>
<div placeholder="{{ 'Hello World' | translate }}"></div>
<div placeholder="{{ scopeVariable || ('Hello World' | translate) }}"></div>
<get-text>Hello World</get-text>
<i18n>Hello World</i18n>
<translate>Hello World</translate>
<!-- The following becomes 'Broken strings are joined' -->
<span ng-bind="{{ 'Broken '
+ 'strings ' +
'are ' +
'joined' |translate}}"></span>
<span ng-bind="'Bed n\'' + ' breakfast' |translate"></span>
<!-- JavaScript expressions are parsed and compiled -->
<span ng-bind="true ? 'Always' : 'Never' |i18n "></span>
<!-- By supplying the --filterPrefix '::' parameter -->
<span>{{:: 'Something …' |translate}}</span>
<!-- The default delimiters '{{' and '}}' must be changed to empty strings to handle these examples -->
<span ng-bind=":: 'Something …' |translate"></span>
<div placeholder="'Hello World' | translate"></div>
You can combine any context, comment and plural together. Also, you can use 'i18n' instead of 'translate' as master token.
You can also provide your own master tokens:
gettext-extract --attribute v-translate --output dictionary.pot foo.html bar.jade
gettext-extract --attribute v-translate --attribute v-i18n --output dictionary.pot foo.html bar.jade
gettext-extract --startDelimiter '[#' --endDelimiter '#]' --output dictionary.pot foo.html bar.jade
gettext-extract
can also remove optional HTML whitespaces inside tags to translate (see PR 68 for more information):
gettext-extract --removeHTMLWhitespaces --output dictionary.pot foo.html
If you want to use optional-chaining, nullish-coalesce or any babel plugin, you might want to set the parameter --parser babel
.
gettext-extract --parser babel --output dictionary.pot foo.html
It can be set to:
--parser
auto|acorn|babel
More info at PR 72
The usage stays the same but with a Javascript file !
gettext-extract somefile.js
const myVar = $gettext("My fantastic msgid")
const myConcatVar = $gettext(
"My"
+ "fantastic"
+ "msgid"
)
const myTempVar = $gettext(
`My
fantastic
msgid`
)
const myContextualizedVar = $pgettext("some context", "Some other string")
const myPluralVar = $ngettext(...)
We recognize the $gettext
, $pgettext
and $ngettext
tokens as the ones from which we can extract text from.
Those tokens are frozen for now, but feel free to make a pull request and add support for variable ones :)
We currently can't extract template strings with variables though.
You can also extract the strings marked as translatable inside the <script>
and <template>
sections of Vue.js components:
gettext-extract MyComponent.vue
With a component that looks like this:
<template>
<h1>{{ greeting_message }}</h1>
<p>{{ number_of_people_here }}</p>
<h2 v-translate> Some text to be translated
</template>
<script>
export default {
name: "greetings",
computed: {
greeting_message() {
return this.$gettext("Hello there!")
},
number_of_people_here(nb_folks) {
return this.$ngettext("There is ${ n } person here.", "There are ${ n } persons here.", nb_folks)
}
}
}
</script>
The Javascript & HTML (or Pug) extraction within a Vue component works with the same rules as stated upper in this document.
gettext-extract
needs the exact file paths to work. If you want to extract gettext from all files in a folder, you can use the UNIX find command. Here is an example as a npm script:
{
//...
"scripts": {
// This is for VueJS files, please adapt for HTML or Jade/Pug templates
"extract-gettext-cli": "gettext-extract --attribute v-translate --output dictionary.pot $(find scripts/src/components -type f -name '*.vue')"
}
}
Outputs or writes to an output file, the sanitized JSON version of a PO file.
gettext-compile --output translations.json fr.po en.po de.po
If you use easygettext
to extract files from an AngularJS code base, you might find the following tips helpful.
To cover the cases (1)
<input placeholder="{{:: 'Insert name …' |translate }}">
<input placeholder="{{ 'Insert name …' |translate }}">
and (2)
<a href="#" ng-bind=":: 'Link text' |translate"></a>
<a href="#" ng-bind="'Link text' |translate"></a>
<a href="#">{{::'Link text' |translate}}</a>
<a href="#">{{'Link text' |translate}}</a>
you should run the extraction tool twice. Once with the command-line arguments
--startDelimiter '{{' --endDelimiter '}}' --filterPrefix '::'
and once with the command-line arguments
--output ${html_b} --startDelimiter '' --endDelimiter '' --filterPrefix '::'
The following Bash script shows how msgcat
might help
#!/usr/bin/env bash
input_files="$(find ./src/ -iname \*\.html)"
workdir=$(mktemp -d "${TMPDIR:-/tmp/}$(basename $0).XXXXXXXXXXXX") || exit 1
html_a=${workdir}/messages-html-interpolate.pot
html_b=${workdir}/messages-html.pot
./dist/extract-cli.js --output ${html_a} --startDelimiter '{{' --endDelimiter '}}' --filterPrefix '::' ${input_files}
./dist/extract-cli.js --output ${html_b} --startDelimiter '' --endDelimiter '' --filterPrefix '::' ${input_files}
# Extract gettext “messages” from JavaScript files here, into ${es_a} …
es_a=${workdir}/ecmascript.pot
# [...] > ${es_a}
# Merge the different catalog templates with `msgcat`:
merged_pot=${workdir}/merged.pot
msgcat ${html_a} ${html_b} ${es_a} > ${merged_pot}
# Cleanup, in case `msgcat` gave merge-conflicts in catalog header.
header=${workdir}/header.pot
sed -e '/^$/q' < ${html_a} > ${header}
body=${workdir}/body.pot
sed '1,/^$/d' < ${merged_pot} > ${body}
cat ${header} ${body} > ${output_file}
# Remove temporary directory with working files.
rm -r ${workdir}
Please note that the script needs to be modified to match your needs and environment.
Run the tests using jest:
npm test
Run:
./src/extract-cli.js --attribute v-translate --attribute v-i18n ~/output.html
angular-gettext is a very neat tool, that comes with Grunt tooling to extract translation tokens from your Pug/Jade/HTML templates and JavaScript code.
Unfortunately, this has two drawbacks:
- It isn't a simple command-line interface, and forces the usage of Grunt;
- It is angular-specific.
This library comes up with two simple CLI tools to extract and compile your HTML tokens.
Our frontend toolchain, systematic doesn't rely on Grunt/Gulp/Broccoli/... and uses a combination of simple Makefiles and CLI tools to do the job.
The toolchain being framework-agnostic, we don't want to depend on Angular to extract our HTML translation tokens. On another note, we use the standard xgettext tool to extract our JavaScript translation tokens.
Nevertheless, the way angular-gettext does it (with tokens, directly in HTML) is elegant, is used by many other libraries and will also be the way to do it in Angular2.
Also, by utilizing either acorn or babel, this tool will parse and compile typical JavaScript expressions used in translate-filter expressions. For example, exposed to a (AngularJS-based) fragment like
<span ng-bind="isNight ? 'Moon' + 'shine' : 'Day' + 'light' |translate"></span>
<span ng-bind="isC ? 'C' + (isD ? 'D' : 'd') : 'c' + (isE ? 'E' : 'e') |i18n "></span>
will produce the following strings
Moonshine
Daylight
CD
Cd
cE
ce
Which will be correctly looked up and translated during runtime, at least by angular-gettext.
TypeScript support is currently limited in that line numbers are not tracked and won't show in generated .po files. This can lead to issues with more complex translations and should be kept in mind.
Thanks a million to @rubenv for the initial ideas and implementations in angular-gettext-tools, which inspired me a lot.
Thanks to ES6 and Babel for being awesome.
MIT