Short functor type syntax

[lpw25]

Add support for simple functor types of the form:

S -> T

equivalent to:

functor (_ : S) -> T

The S and T can be any module type expressions.

While not actually ambiguous there is a potential shift-reduce conflict related to with constraints:

S with type t = int -> float

vs

S with type t = int -> T

This is resolved in favour of the first one (shift), meaning that the second type must be written:

(S with type t = int) -> T

[gasche]

While I previously didn't like using functor S -> T as syntactic sugar for functor (_:S) -> T, it may be that the S -> T form actually doesn't have this problem: it makes sense to consider functor (X:S) -> T to be the dependent arrow syntax, while simply S -> T would be the syntactic sugar for the non-dependent form.

Without a convincing use-case, I haven't much of an opinion on this.

[lpw25]

Without a convincing use-case, I haven't much of an opinion on this.

I don't know how convincing a use-case it is, but every time I write things with higher-order functors I get annoyed that I can't write a non-dependent arrow like this.

[bobot]

This PR have been discussed during the last developers' meeting, where I had the pleasure to be. To sum-up:

• it seems not really useful (usually you need to bind the module argument)
• it is not clear that users can easily understand this syntax
• it could be ambiguous (in a general meaning) with a futur syntax for the types of inline record.

So currently they don't want to integrate this proposition. A compelling full example or examples in existing libraries could change that position.

[yallop]

There are a few places in ctypes where this would make the code more succinct. Here's one, from cstubs.mli:

module type BINDINGS = functor (F : TYPE) -> sig end

Since F is not used in the output, we could write

module type BINDINGS = TYPE -> sig end

[gasche]

There was a code review of this patch during the last developer meeting (I wasn't there), and the complain was the show boolean that incurs a lot of code changes. Remove this boolean (and simply do not print the name when it is _), and the patch will be ready for merging.

[lpw25]

I've removed the show boolean and rebased the patch on current trunk.

[gasche]

It's nice to see #42 finally merged in trunk! Thanks for the nice work -- and testsuite coverage.

[bluddy]

very cool