As mentioned in SR-2135, this isn't really intended to work because we don't support general destructuring in parameter lists. There was special-case behavior that only worked for a single tuple argument but that is going away in -swift-version 4.

This on the other hand is a close alternative that explicitly uses destructuring on the single tuple argument:

[t].map { (tuple) -> Int in
  let (x, (y, z)) = tuple
  return x + y + z
}

Again I'll note the bad diagnostic. We would ideally do better here.

This issue is possibly a duplicate of SR-1620, but that one is unassigned and has no recent comments...

Thanks, Pavol

Pavel, can you provide PR #?
Also, is this just improvement to diagnostic messages or a fix for general tuple decomposition?

If the first… what is the justification for having the single tuple decomposition in the for-in loop - the same that was removed from closure arguments?

It's #9425 and it's improvement to diagnostic yes.

If the first… what is the justification for having the single tuple decomposition in the for-in loop - the same that was removed from closure arguments?

I'm actually not quiet sure what is the motivation behind that but maybe @slavapestov can shed some light?

PR 9425 has landed to 4.0 branch so here is the diagnostic example for the description is going to produce:

<REPL Input>:1:15: error: closure tuple parameter does not support destructuring
[t].map { (x, (y, z)) -> Int in x + y + z }
              ^~~~~~
              arg1              let (y, z) = arg1;