Custom translation system for sentences files

[Stypox]

Problems

Currently sentences files have to be translated by copying the *.dslf files from the en folder to the new language folder, and then translating them. This is cumbersome since:

1. GitHub pull requests are not the best tool to handle translations, since even a small update requires a branch, a dev approval, a merge, ... (that's why Weblate is used for app strings)
2. The sentence and capturing group ids and the section specificity should not be translated, but there is nothing making sure that this does not happen (not even when building the app!)
3. Syntax errors are only found out about when building the app, but we can't expect translators to be able to build the app by themselves (this would be partially solved by adding a github action to pull requests that reports build errors, though the feedback would still not be instant)
4. Translating a dslf file is more difficult than just translating a string, since you have to think about all possible ways to say something. It is easy to miss a sentence and only realize about it when testing out the app (and again, we can't expect translators to build the app by themselves)

Custom translation system

Adding the dslf files to Weblate is doable, though that would only solve point 1, since it would not ensure that there are no syntax/semantic errors. I don't think Weblate has any way to add custom plugins that would do that.

Therefore I think the way to go would be to create a custom translation system with these features (italic means "difficult to implement"/"maybe later"):

• git integration like Weblate (but more basic)
• a view that shows the file currently being translated, with syntax highlighting
• a view that shows syntax errors, highlighting them in the code
• a view that shows other errors, such as mismatched sentence ids, capturing group ids or section specificity
• a view with suggestions on how to improve a sentence
• a text field that allows inserting a user input and then shows whether it matches or not, the matched capturing groups and which sentence actually matched
• when a skill marketplace will be setup, third-party skill developers should be able to get their skill sentences hosted on the same service

[MXC48-zz]

It would be very useful for me for contribute to the project