Use Proxy type for type info?

[paf31]

First off, this looks fantastic.

I just noticed you're using things like an unused m a in the Bind check, to infer the type info. Maybe a set of Proxy types would be an improvement, since you might not even be able to produce a value of type m a easily.

[garyb]

Ah, you noticed this :) I was going to open an issue asking for review anyway.

Yeah, I was thinking of something like that too. Where would be a good place for the Proxy type to live? I made one in affjax too, but having an entire library for it perhaps seems overkill? Maybe not.

I wasn't sure about the utility of the tests for Semigroupoid and Category too, seems it might be a little hard to produce values for those that are testable due to the Eq (a b e) constraint, as obviously that won't work for (->), for example.

[paf31]

The issue is that Proxy is naturally poly-kinded, which we can't do right now. Some kind of purescript-proxy library wouldn't be a bad idea though.

I was wondering if it would be possible to introduce some fake types to take the place of the a, b, c, etc. in tests. For Monad, for example, you don't care what a is, just m. So we could make an enum type with say, 20 inhabitants or something, and just use that every time.

[garyb]

Yeah, maybe I'll do that for now then, make a tiny library and include Proxy2, Proxy3 etc.

Agreed on the Monad stuff too. Could just use unit for that I guess?

[garyb]

Actually I guess unit isn't good enough, it could easily hide problems.