Hisham Muhammad

I suspect this is not a coding bug, but a side-effect of the combination of unbound type variables and bivariant types. This may be actually tricky to solve.

@FourierTransformer yeah, this use case is not currently covered by the type system. The underlying concept you're looking for is whether the type system is _nominal_ or _structural_. In a...

@bjornbm Thank you for your input, much appreciated! > > In particular, you want union types to be structural because the point of having a type U = A |...

@fgaz Sorry for not providing proper feedback earlier (for some reason I thought I did?? but I clearly didn't...) I plan to add a way to specialize the behavior of...

The canonical solution for the original problem pointed here is indeed to add parentheses to the return tuple. I'm not entirely sold on the idea that parentheses should always be...

Record functions are pretty special because their declaration semantics are meant to accomodate some natural Lua patterns for declarations of tables and their methods. So there's a few special behaviors...

@OrfeasZ not yet, but this is something I want to implement. Something along these lines should come with the interfaces implementation, which should cover this, as well as #655 (and...

The second case (__add is not defined in rec_mt) is unlikely to be supported soon (the compiler currently does not do nil checks in general), but the first one sounds...

@svermeulen This looks like a different issue. Could you open a separate ticket? Thanks!! Edit: also, I tried your example above in `v0.15.3`, `master` and `next`, and all three branches...

@uramer Thanks for the feedback! I've pushed a pair of commits that make dot-notation for `__index` work... that's not fully equivalent to maprecords, but it's a compiler improvement nonetheless!