Purity of the stack execution model? (fwd)

[Koen Claessen]

Hi!

Somebody posted the following message on
comp.lang.functional. It reports an error when having a main
function which has type "IO [()]", together with using
runhugs.

This is a bug; the Haskell report says that main "must be a
computation of type IO t for some type t".

/Koen.

---------- Forwarded message ----------
Date: Wed, 14 Feb 2001 23:23:50 +0100
From: Tore Lund <tl001@online.no>
Newsgroups: comp.lang.functional
Subject: Re: Purity of the stack execution model?

Markus Mottl wrote:
>
>   actions = [putStr "foo", putStr "bar"]
>   nl = putStr "\n"
>   main = sequence actions >> nl >> mapM (>> nl) (reverse actions)

Maybe some kindly soul could tell me what is the type of main here.
Running the right-hand clause within a "do" construct works fine.
However, if I try the following:

    module Main where
    actions = [putStr "foo", putStr "bar"]
    nl = putStr "\n"
    main = sequence actions >> nl >> mapM (>> nl) (reverse actions)

I get this message from runhugs:

    Program error: fromDyn failed.  Expecting <<IO() >> found <<IO[
    Program error: {instShow_v1536_v1601 [instTypeable_v1537 (head
    (unsafePerformIO main))]}runhugs: Error occurred

Trying to equip main with various IO types just leads to more errors.
What am I missing please?

                                *

Moreover, this is admittedly a powerful example of what can be done with
Haskell and monads.  Is it possible to write such a program with
comparable ease in, say, SML or OCaml?
-- 
    Tore