Why FL specifies the ChainRec typeclass when there is the trampoline monad? about fantasy-land HOT 7 CLOSED

I think it's a good question. I've always found ChainRec confusing, though, so I'm no help. ;)

from fantasy-land.

Did you answer your own question, @ivenmarquardt?

from fantasy-land.

Actually not, @davidchambers. I have continued my study of different forms of monadic recursion and am far from done, Since nobody answered I figured the question was a bit rash.

from fantasy-land.

Maybe take a look at the original issue
#151 (comment)

for example in PureScript there are instances of MonadRec for many monads https://pursuit.purescript.org/search?q=MonadRec

This example might highlight motivation:
https://github.com/purescript/purescript-tailrec/blob/master/src/Control/Monad/Rec/Class.purs#L33-L55

from fantasy-land.

OK, will look into that more closely. I figured that purescript just needs it this way, so that the compiler could apply the usual tail call elimination. Then, however, it would be a purescript specific issue.

from fantasy-land.

One of the reasons MonadRec exists is that purescript compiler can't optimize monadic recursion.
It can optimize regular tail call recursion into for loop but not monadic.

from fantasy-land.

I wonder why PS cannot optimize recursion within a monad when we can write a stack-safe monad instance for the Trampoline type in a principled way. I am not a compiler programmer though.

With Trampoline you can add stack-safety to many transfromer stacks and keep your code DRY. ChainRec on the other hand is probably more efficient. I will use the monad instance for the time being and report here if I ever feel the need for ChainRec.

from fantasy-land.