This is a basic sketch of the workflow needed to add translations:

Create translations/kubectl/<language>/LC_MESSAGES/k8s.po. There's no need to update translations/test/... which is only used for unit tests.

There is an example PR here which adds support for French.

Once you've added a new language, you'll need to register it in pkg/util/i18n/i18n.go by adding it to the knownTranslations map.

There is a simple script in translations/extract.py that performs simple regular expression based wrapping of strings. It can always use improvements to understand additional strings.

Once the strings are wrapped, you can extract strings from go files using the go-xgettext command which can be installed with:

go get github.com/gosexy/gettext/go-xgettext

Once that's installed you can run ./hack/update-translations.sh, which will extract and sort any new strings.

Edit the appropriate k8s.po file, poedit is a popular open source tool for translations. You can load the translations/kubectl/template.pot file to find messages that might be missing.

Once you are done with your k8s.po file, generate the corresponding k8s.mo file. poedit does this automatically on save, but you can also run ./hack/update-translations.sh to perform the po to mo translation.

We use the English translation as the msgid.

Run ./hack/generate-bindata.sh, this will turn the translation files into generated code which will in turn be packaged into the Kubernetes binaries.

There is a script in translations/extract.py that knows how to do some simple extraction. It needs a lot of work.

To use translations, you simply need to add:

import pkg/i18n
...
// Get a translated string
translated := i18n.T("Your message in english here")

// Get a translated plural string
translated := i18n.T("You had % items", items)

// Translated error
return i18n.Error("Something bad happened")

// Translated plural error
return i18n.Error("%d bad things happened")