Monad vs Monad transformers

Question

"Monads allow the programmer to build up computations using sequential building blocks" therefore it allows us to combine some computations. If this is the case, then why the following code can not be run?

import Control.Monad.Trans.State
gt :: State String String
gt = do
 name <- get
 putStrLn "HI" -- Here is the source of problem!
 put "T"
 return ("hh..." ++ name ++ "...!")
main= do
 print $ execState gt "W.."
 print $ evalState gt "W.."

• Why cannot we put different functions in a monad (like the above example)?

• Why do we need an additional layer, i.e. transformers to combine the monads?

Answer 1

Monad transformers are the mechanisms for putting different functions in a monad.

A monad only knows how to combine computations that are within the abilities of that monad. You can't do I/O in a State monad, but you can in a StateT s IO a monad. However, you will need to use liftIO on the computations that do I/O.

import Control.Monad.Trans.State
import Control.Monad.IO.Class (liftIO)
gt :: StateT String IO String
gt = do
 name <- get
 liftIO $ putStrLn "HI"
 put "T"
 return ("hh..." ++ name ++ "...!")
main = do
 print =<< execStateT gt "W.."
 print =<< evalStateT gt "W.."

• 4xx: Thanks for the answer! Could you please tell me why you used =<< in the last two lines?
• castletheperson: execStateT and evalStateT both return an IO String here, and you probably want to print the String inside them. =<< calls a function on the left using the inner value of the monad on the right.
• chi: @4xx This equivalent version might be more familiar to you: main = do x <- execStateT gt "W.." ; print x ; y <- evalStateT gt "W.." ; print y
• 4xx: Is your answer for the standard monads, like IO, State, etc or it's generic. I wondered if we can have a function (or monad) that performs series of arbitrary functions (computations) sequentially on an input?