Sometimes, it is nice to write code in MonadError to indicate that the code can throw an exception.

foo :: MonadError FooError m => Int -> m Char

We can call foo in Either FooError for a pure test, or a ExceptT FooError IO if we are writing in some monad directly. However, sometimes you can't use ExceptT. One possible reason is performance: ExceptT can add overhead. Another reason is MonadUnliftIO - ExceptT cannot have an instance for this.

You don't want to stop using MonadError, but you also can't use ExceptT anymore. What can you do?

import Control.Monad.Except.IO
main :: IO ()
main = do
    eres <- runExceptIOT (foo 3)
    case eres of
        Left FooError ->
            putStrLn "got FooError"
        Right c ->
            putChar c

You can replace ExceptT with ExceptIOT, and all exception use will switch to IO based exceptions instead of Either-based values. This allows you to use MonadUnliftIO without fear.