Allow tuples in untagged variants

[tsnobip]

fixes #6544

[zth]

Hmm, we need to have code that checks the length of the array at runtime as well to see if it matches the tuple, right? So I think tuple needs to be its own code path fully.

I think here is where you'll want to change things to cover checking lengths when tuple is its own thing: https://github.com/rescript-lang/rescript-compiler/pull/6383/files#diff-41883149b99918ba2a405ce00ee26ec4679a3fdf02f1026fe98782983faa6c0a

Brain dump:

• You'll need to track the length of each tuple, and make sure that there aren't multiple tuples with the same length, in addition to checking if there are arrays already
• Next level: Support "array catch all" case by verifying that all tuples only have the same types as the array case

Edit: Or maybe it's fine to not check length at this stage when only one is allowed.

[tsnobip]

@zth I think it's already useful to have the possibility to have one case of tuple.

Being able to treat tuples a bit like we do for strings or ints is nice, but I think it's an addition that would belong in a followup PR. What do you think?

[zth]

@tsnobip I agree 👍 Extending should be a follow up.

This looks good to me, but I'd like @cristianoc have a final look as he knows his way around untagged variants better than me.

[cristianoc]

With only 1 case this looks good.
For future, Distinguishing sizes also opens the question on whether tuples of size Eg 2 can be differentiated based on their types.

[zth]

Ready to merge then? @tsnobip can you make a corresponding PR to the docs as well?

[tsnobip]

Sure I'll do the PR for the documentation!