A `granularity` setting of `0` will remove the line numbers from the PO files. This reduces the churn as we update the English source texts.
8.4 KiB
Translations of Comprehensive Rust 🦀
We would love to have your help with translating the course into other languages! Please see the translations page for the existing translations..
We use the Gettext system for translations. This means that you don't modify
the Markdown files directly: instead you modify .po
files in a po/
directory. The .po
files are small text-based translation databases.
Tip: You should not edit the
.po
files by hand. Instead use a PO editor, such as Poedit. There are also several online editors available. This will ensure that the file is encoded correctly.
Important: You need to run
dprint fmt
after editing the PO file. This ensures consistent formatting of the file. You need to install the Gettext tools for this, see the Preparation section below.
There is a .po
file for each language. They are named after the ISO 639
language codes: Danish would go into po/da.po
, Korean would go into
po/ko.po
, etc. The .po
files contain all the English text plus the
translations. They are initialized from a messages.pot
file (a PO template)
which contains only the English text.
We will show how to update and manipulate the .po
and .pot
files using the
GNU Gettext utilities below.
Preparation
Please make sure you can build the course. You will also
need the msgmerge
and msgcat
Gettext tool installed. Please see our
contribution guide for details.
Creating and Updating Translations
First, you need to know how to update the .pot
and .po
files.
You should never touch the auto-generated po/messages.pot
file. You should
also not never the msgid
entries in a po/xx.po
file. If you find mistakes,
you need to update the original English text instead. The fixes to the English
text will flow into the .po
files the next time the translators update them.
Tip: See our style guide for some things to keep in mind when writing the translation.
Generating the PO Template
To extract the original English text and generate a messages.pot
file, you run
mdbook
with a special renderer:
MDBOOK_OUTPUT='{"xgettext": {"pot-file": "messages.pot", "granularity": 0}}' \
mdbook build -d po
You will find the generated POT file as po/messages.pot
.
Initialize a New Translation
To start a new translation, first generate the po/messages.pot
file. Then use
msginit
to create a xx.po
file for the fictional xx
language:
msginit -i po/messages.pot -l xx -o po/xx.po
You can also simply copy po/messages.pot
to po/xx.po
. Then update the file
header (the first entry with msgid ""
) to the correct language.
Tip: You can use the
cloud-translate
tool to quickly machine-translate a new translation. Install it withcargo install cloud-translate
Untranslated entries will be sent through GCP Cloud Translate. Some of the translations will be wrong after this, so you must inspect them by hand afterwards.
Next, please update the file .github/labeler.yml
to include the new language:
+"translation/xx":
+ - changed-files:
+ - any-glob-to-any-file: po/xx.po
Refreshing an Existing Translation
As the English text changes, translations gradually become outdated. The translations contain a POT-Creation-Date header which tells you when they were last updated with new English messages.
To update the po/xx.po
file with new messages, first extract the English text
into a po/messages.pot
template file. Then run
msgmerge --update po/xx.po po/messages.pot
Notice that the POT-Creation-Date field is updated to the current time and date. This becomes the new baseline for the translation: new English text added afterwards will not show up in your translation, including completely new pages.
When running msgmerge
, unchanged messages stay intact, deleted messages are
marked as old, and updated messages are marked "fuzzy". A fuzzy entry is not
used when we publish a translation! You have to go over the fuzzy entries by
hand and verify that the translation is correct the fuzzy marker.
Note: Your PRs should either be the result of running
msgmerge
or the result of new translation work on the PO file for your language. Avoid mixing the two since it often creates a very large diff, which is hard or impossible to review.
Editing a Translation
You should install a PO editor to edit the .po
file for your language. The
files are simple text files, but it helps to use a dedicated editor since it
will take care of escaping things like "
correctly.
There are many PO editors available. Poedit is a popular cross-platform choice, but you can also find several online editors.
Formatting a Translation
If the file is not formatted correct, you will get an error on the PR. Make sure
to follow the steps to install Gettext and
dprint
and then run:
dprint fmt po/xx.po
This will automatically format the .po
file for you. Commit the formatting fix
and push to your branch. Your PR should now be error free.
Using Translations
This will show you how to use the translations to generate localized HTML output.
Note:
mdbook
will use original untranslated entries for all entries marked as "fuzzy" (visible as "Needs work" in Poedit). This is especially important when usingcloud-translate
for initial translation as all entries will be marked as "fuzzy".
Building a Translation
Make sure you have gone through the build setup at least once.
To use the po/xx.po
file for your output, run the following command:
MDBOOK_BOOK__LANGUAGE=xx mdbook build -d book/xx
This will update the book's language to xx
, it will make the mdbook-gettext
preprocessor become active and tell it to use the po/xx.po
file, and finally
it will redirect the output to book/xx
.
Serving a Translation
Like normal, you can use mdbook serve
to view your translation as you work on
it. You use the same command as with mdbook build
above:
MDBOOK_BOOK__LANGUAGE=xx mdbook serve -d book/xx
When you update the po/xx.po
file, the translated book will automatically
reload.
Reviewing Translations
When a new translation is started, we look for people who can help review it. These reviewers are often Googlers, but they don't have to be. To automatically get an email when new PRs are created for your language, please add yourself to the CODEOWNERS file.
When reviewing a translation, please keep in mind that translations are a labour of love. Someone spends their free time translating the course because they want to bring Rust to users who speak their language.
Nothing is published right away after a PR lands for a new in-progress language. It is therefore safe to merge the PR as long as the translation is reasonable. This is often better than leaving 50+ comments since this can be overwhelming for the contributor. Instead, please work with the contributor to improve things in follow-up PRs.
GitHub Suggestions
When reviewing a translation PR, please use the GitHub suggestion feature. This feature allows you to directly write how you think a line or paragraph should be phrased. Use the left-most button in the toolbar to create a suggestion.
The PR author can apply the changes with a single click afterwards, drastically reducing the number of round-trips needed in a review.
Incomplete Translations
When the first 1-2 days of the course have been translated, we can publish the translation and link it from the translations page. The idea is to celebrate the hard work, even if it is incomplete.