Remove the Typescript Extractor?

[nathanforce]

What do you think about removing the Typescript extractor in favor of always using Babel? Babel supports parsing TS nowadays so keeping both feels redundant. Also, the current typescript extractor reports incorrect line numbers when creating origin. This is because the code that is passed to Babel is first transpiled by the Typescript API.

[semoal]

Yeah was thinking just the same. I'm just trying to find a way of being sure that we don't brake any setup.
I've seen some issues related to the typescript extractor, like the required NODE_ENV=development in lingui extract command sometimes is required to make it work...

I don't want to release a breaking change, so any suggestion it's welcome

[nathanforce]

Maybe the best solution is to expose the extractors and allow users to choose. I see a few ways of doing this. Something similar to Jest's moduleNameMapper where you map an extension => extractor. Current behavior could be maintained for now with

extractors: {
   babel: babel.defaultExtensions,
   typescript: ['.ts', '.tsx']
}

Alternatively you could keep the current { match, extract } signature and accept something like:

extractors: ['babel', 'typescript']

Both of these options open up the path for supporting custom extractors if we modified the extractor signature to return the extracted messages rather than relying on babel-plugin-extract-messages to write the catalog. (IMO this change should happen anyhow as the code is clearer/separated concerns).