design - Why does Haskell not have an I Monad (for input only, unlike the IO monad)? -

conceptually, seems computation performs output different 1 performs input only. latter is, in 1 sense, purer.

i, one, have way separate input parts of programme ones might write out.

so, why there no input monad?

any reason why wouldn't work have monad (and o monad, combined io monad)?

edit: meant input reading files, not interacting user. use case, can assume input files not change during execution of programme (otherwise, it's fine undefined behaviour).

i disagree bdonlan's answer. it's true neither input nor output more "pure" are quite different. it's quite valid critique io single "sin bin" effects crammed together, , make ensuring properties harder. example, if have many functions know read memory locations, , never cause locations altered, nice know can reorder execution. or if have program uses forkio , mvars, nice know, based on type, isn't reading /etc/passwd.

furthermore, 1 can compose monadic effects in fashion besides stacked transformers. can't all monads (just free monads), case that's need. iospec package, example, provides pure specification of io -- doesn't seperate reading , writing, seperate them from, e.g., stm, mvars, forkio, soforth.

http://hackage.haskell.org/package/iospec

the key ideas how can combine different monads cleanly described in data types la carte paper (great reading, influential, can't recommend enough, etc.etc.).