Cofree monad and and applicative instances are inconsistent

[prophile]

pure and return have differing results. In the case of Cofree Maybe for instance, pure yields an infinite string of its argument, where as return yields a one-element string containing its argument.

[ekmett]

Ack. Then we need to change the Applicative to line up with the Monad.

[prophile]

That does mean adding an Alternative constraint on the functor to the Applicative instance.

[ekmett]

So be it. =) The existing Applicative I think was added without much thought.

[xuwei-k]

maybe Apply (Cofree f) is inconsistent too?

https://github.com/ekmett/free/blob/b229616f6/src/Control/Comonad/Cofree.hs#L149-L155

[ekmett]

Appears to be!

On Tue, Dec 17, 2013 at 12:57 AM, kenji yoshida [email protected]:

maybe Apply (Cofree f) is inconsistent too?

https://github.com/ekmett/free/blob/b229616f6/src/Control/Comonad/Cofree.hs#L149-L155

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/36#issuecomment-30728047
.

[esoeylemez]

I'm currently working on a fix. I propose that we disconnect the comonadic and monadic branches of Cofree's logic. That implies that ComonadApply and Apply will disagree. Does anyone object to that?

[ekmett]

I very much do not like the idea of disconnecting the two sides of Cofree. I'm ambivalent about whether we rip out the "incidental" Monad instance, etc. or split it into separate types, but pretending the class hierarchy is unrelated isn't a solution I'd accept.