My prosaic solution would be to have more stock implementations of mconcat (here balancedMconcat, though some parallelMconcat's would also be fun) that make use of the associativity guarantee.  Then use that explicitly:--

balancedMconcat . fmap Sum $ [a, b, c, 2, 3, d]

rather than asking the compiler to be magic for you.

Freddie

On 14 September 2010 13:11, Bas van Dijk <v.dijk.bas@gmail.com> wrote:
Oops, I mixed up associative with commutative.

On Tue, Sep 14, 2010 at 1:27 PM, Bas van Dijk <v.dijk.bas@gmail.com> wrote:
> On Mon, Sep 13, 2010 at 4:23 PM, Nick Bowler <nbowler@elliptictech.com> wrote:
>> ... not all Num instances have an associative (+).
>
> Indeed:
>
> $ cabal install repr    # [1]
> ...
> $ ghci
> Prelude> :m Text.Repr
> Prelude Text.Repr> show (1 + 2 :: Repr Int) == show (2 + 1 :: Repr Int)
> False
>
> because:
> show (1 + 2 :: Repr Int) == "1 + 2"
> show (2 + 1 :: Repr Int) == "2 + 1"
>
> but note:
> Prelude Text.Repr> (1 + 2 :: Repr Int) == (2 + 1 :: Repr Int)
> True
>
> Bas
>
> [1] http://hackage.haskell.org/package/repr
>
_______________________________________________
Haskell-prime mailing list
Haskell-prime@haskell.org
http://www.haskell.org/mailman/listinfo/haskell-prime