search

ekmett / pointed

pointed and copointed data
Other
23 stars 9 forks source link

Can number of pointed's dependencies be reduced? #6

Closed neko-kai closed 11 years ago

neko-kai commented 11 years ago

Could comonads/semigroups depend on pointed, not the other way around? We'd like to use Copointed in monad-control, but the amount of dependencies prevents adoption.

ekmett commented 11 years ago

I don't think so. This was actually flipped the other way around rather deliberately.

There are 58 reverse dependencies of semigroups, and 41 on comonad, but only 8 reverse dependencies on pointed.

Something has to sit on the outside, and pointed drew the short straw, because for the most part its the least useful abstraction: The problem with Copointed (and Pointed) in general is that the only law it offers is a free theorem, so almost any code that is written generically in terms of Copointed (Pointed) requires one-off reasoning about the particular instance -- it simply provides a convenient ad hoc overloading for an family of algebras (coalgebras).

You some times need just that, but for every one of those, someone else writes foldMap point.

Flipping this would mean that I get a whole lot more complaints wanting the exact opposite request and favor the less useful abstraction.