---------- Forwarded message ----------
From: Tom Hobbs <tvhobbs@googlemail.com>
Date: Tue, Jul 6, 2010 at 10:12 AM
Subject: Re: [Haskell-beginners] More Deserialization Woes
To: Stephen Tetley <stephen.tetley@gmail.com>

Hello again,

I've been reading through various tutorials and they all put IO as the outermost monad, like you suggest. However, I don't think that's what I want.

My IO operation is reading from a network stream, not a file, so I want a failure halfway through should signify a complete failure. So, I believe that the type signature for my function should be;

ping :: String -> PortNumber -> Maybe (IO [String])

because the result "Just []" is a valid one and does not signify a failure.

However, the result of "IO [Just "a", Just "b", Nothing, Nothing]" would signify that communication failed halfway through and would not make sense in my context. This is what the advice seems to be suggesting I write. But in this case, I'd prefer to return "Nothing" to signify that a problem occurred.

So my first question is; because I want to do something so different from the majority of the articles I've read; am I in a niche where my requirement makes sense, or does my requirement make no sense - a theory that is backed up by the fact that no one else seems to be doing that...

Now, I'm not sure I can get there by myself, since I'm struggling to get the right incantation of catching errors but I'll keep plugging away at that for a while.

But can someone help me with my next question also.

Is it possible, to extract the values out of the IO monad so it can be used in pure functions.

For example, once ping has returned it's Maybe [IO String], I would like to be able to create another function such as;

purePing :: String -> PortNumber -> Maybe [String]

purePing a p = removeIOMonad (ping a p)

where

removeIOMonad Nothing = Nothing

removeIOMonad [] = Just []

removeIOMonad (x:xs) = clevelDropIOMagic x : removeIOMonad xs

...or something...

Once the IO [String] has been read from the stream, no further IO is necessary, so any other function should be able to just use the list and not worry about the IO stuff.

Again my questions are, Is this an okay thing to do, or is my design/idea very, very wrong?

Thanks again for the help,

Tom

On Thu, Jul 1, 2010 at 1:53 PM, Stephen Tetley <stephen.tetley@gmail.com> wrote:

Hi Tom

This is bit that is wrong:

((UTF.toString name) : readNames (n-1) h)

You are trying to cons (:) a String onto an IO [String]:

(UTF.toString name) :: String

(readNames (n-1) h) :: IO [String]

(:) :: a -> [a] -> [a]

This is easy mistake to make, I probably still make it myself now and
again if I'm typing faster than I'm thinking.

As for ping, you are right that the common answer type will be "IO
(Maybe [String])" i.e.

ping :: String -> PortNumber -> IO (Maybe [String])

Generally you will want IO as the outermost (type-) constructor for an
expression involving IO, this is because you can't 'escape' IO. There
are some exceptions where IO isn't the outermost type constructor, the
common one I can think of is you might want to build a list of IO
actions [IO ()]. You would then commonly pass on this list to evaluate
it later with mapM_.

Neil Mitchell's "IO without (concentrating) on monads tutorial" is a
very good place to start:
http://neilmitchell.blogspot.com/2010/01/haskell-io-without-monads.html

Best wishes

Stephen