Multiple binders in patterns about quotedpatternmatchingproof

[liufengyun]

Multiple binders are indeed annoying, I don't know any good reference about this.

Instead, I think the mechanization can try to simplify the pattern language so that it only has single binder and try to keep the same expressiveness. This is one area, where pen & paper systems look better esthetically.

The annoying case is application. If we assume the pattern for application is always App[T] f x, where f and x are bound variables, will simplify the problem? By doing so, patterns are no longer composable, but they can nest patmat to have the same expressiveness.

[radeusgd]

This sounds like a good idea to simplify things.
However are you sure we are not limiting the expressiveness by restricting to not-nested pattern matching? I think that this is likely true, but it doesn't seem completely trivial to assume it.

As for binding two variables in the case of App, it is certainly much simpler than the arbitrary amount of variables. I can try experimenting with autosubst if that will somehow be possible.

Unfortunately, the simplest solution I tried:

| Lam (ebody : {bind term}) (typ: type)
| Lam2 (ebody : {bind {bind term}}) (typ: type)

where I hoped Lam2 to capture 2 binders instead of just one doesn't seem to work - when tested with substitutions Lam2 seems to behave exactly like Lam.

[radeusgd]

This is solved as described in #6