This is the back door into the IO monad, allowing IO computation to be performed at any time. For this to be safe, the IO computation should be free of side effects and independent of its environment.

If the I/O computation wrapped in unsafePerformIO performs side effects, then the relative order in which those side effects take place (relative to the main I/O trunk, or other calls to unsafePerformIO) is indeterminate.

However, it is less well known that unsafePerformIO is not type safe. For example:

     test :: IORef [a]
     test = unsafePerformIO $ newIORef []
     
     main = do
     	      writeIORef test [42]
     	      bang \<- readIORef test
     	      print (bang :: [Char])

This program will core dump. This problem with polymorphic references is well known in the ML community, and does not arise with normal monadic use of references. There is no easy way to make it impossible once you use unsafePerformIO. Indeed, it is possible to write coerce :: a -> b with the help of unsafePerformIO. So be careful!

A variant of throw that can be used within the IO monad.

Although ioError has a type that is an instance of the type of throw, the two functions are subtly different:

 throw e   `seq` return ()  ===> throw e
 ioError e `seq` return ()  ===> return ()

The first example will cause the exception e to be raised, whereas the second one won't. In fact, ioError will only cause an exception to be raised when it is used within the IO monad. The ioError variant should be used in preference to throw to raise an exception within the IO monad because it guarantees ordering with respect to other IO operations, whereas throw does not.