Software labels translation is not so easy

Some developers have hardly ever touched software labels translation, some do it on a day-to-day basis. It sure helps to work in a country with more than one language – official or de facto.

Even for in the first case, it’s considered good practice to externalize labels in properties files. As for the second case, languages are in general related.

In Java, the whole label translation mechanism is handled through a hierarchy of properties files. At the top of the hierarchy lies the root file, at the second level language-specific files and finally at the bottom country-specific files (let’s forget about lower levels since I haven’t seen them used in 15 years). The translation for message strings are searched along a specific locale, starting from most specific – country, up to the root. If the translation is found at any level, the resolution mechanism stops there and the label is returned.

As a simple example, let’s take a simple use-case, displaying the number of items on a list. I’d need probably 3 labels:

This is probably the resulting messages.properties file:

Things get interesting when the customer wants to translate the software after the initial release into an un-related language. Let’s not go as far as pictograph-based languages such as Mandarin Chinese or RTL languages such as Arabic, but use Russian, a language I’m trying to learn (emphasis on try).

Russian is a language that has cases, like in Latin.

Case is a grammatical category whose value reflects the grammatical function performed by a noun or pronoun in a phrase, clause, or sentence. In some languages, nouns, pronouns, and their modifiers take different inflected forms depending on what case they are in.

So, what the fuss about it? Just translate the file and be done with it! Well, Russian is an interesting language for counting. With one, you’d use singular, from 2 to 4, you’d use plural and nominative case, but starting from 5, you use plural and genitive case – indicating quantity.

Now keys will look like the following (the messages are not very important by themselves):

Is it OK to translate, now? Not quite. Russian is derived from old Slavic and old Slavic had three grammatical numbers: singular, plural and dual. Russian has only singular and plural but there’s a remnant of that for the feminine case. In this case, you’d use две instead of два.

This requires the following keys:

And this only for the things I know. I’m afraid there might be more rules than I don’t know about.

There are a couple of lessons to learn here: