This issue is based on feedback from Priit Jõerüüt working on the Estonian translation in Weblate. Since Weblate’s comment box is too short for deeper discussion, I’m moving the full context here.
Original comment
Languages have very different word orders and use different language constructs (e.g. > declension instead of prepositions) and different approaches to singular/plural (e.g. > one|some|many).
This means, that best translations will be achieved when:
- visual sentences are not made of many translation strings, joined in code, but one > translation string for one sentence with variables is used (this allows to change word order and find ways to handle cases)
- in these difficult sentences translation strings are not reused, but each one has its own (this is also avoided if p 1. is in use)
- languages with complex approach to singular/plural would probably prefer to avoid "hour(s)"-style of translation and have full use of grammatical number as supported in Weblate
From the current approach of splitting everything, for example Estonian translation is quite unnatural. The UI uses a lot of selectboxes here and I do not know if source code language and Weblate can handle variables