[Haskell-cafe] A practical Haskell puzzle

[Gábor Lehel]

On Mon, Feb 28, 2011 at 12:41 PM, wren ng thornton <wren at freegeek.org> wrote:
> On 2/28/11 2:43 AM, Yitzchak Gale wrote:
>>
>> You have written a large software system in Haskell. Wishing to
>> play to Haskell's strength, you have structured your system
>> as a series of composable layers. So you have data types
>>
>> Layer1, Layer2, ...
>>
>> and functions
>>
>> layer2 :: Layer1 ->  Layer2
>> layer3 :: Layer2 ->  Layer3
>> ...
>>
>> etc.
>
> Assuming you actually name them Layer1, Layer2, etc, or use any other
> regular naming scheme, you can break apart the names and use typeclasses to
> help out:
>
>    type family Layer :: * -> *
>    data Z
>    data S n
>
>    class Layerable n where
>        layer :: Layer n -> Layer (S n)
>
> Then it's just a matter of getting the right number of them, a la lifting
> through monad transformer stacks. Of course, from here it's not that hard to
> add in some type hackery to do the lifting for you, a la "Data types a la
> Carte"[1]. It's not the cleanest thing ---there's a good deal of boilerplate
> up front--- but once it's set up, it should Just Work(tm).

I was thinking something like:

class IsNat n => LayerID n where
    data Layer n :: *
    runLayer :: LayerID (S n) => Layer n -> Layer (S n)
    serialize :: Layer n -> L.ByteString
    deserialize :: L.ByteString -> Layer n

and then use a reifyIntegral-esque function (a la the reflection and
type-level packages) to reify two numbers up to the type level and run
all of the runLayers in between the two, but where I ran into trouble
was that there's no guarantee of there being a LayerID instance for
any given n; and at that point you either need the non-existent
'do-this-if-there's-an-instance-and-do-that-otherwise' construction,
or perhaps you can hack something with OverlappingInstances which
moves the whole thing from sophisticated-but-clean-oleggery to kind of
ugly. Is there any cleaner way around this?


>
>
> [1] http://www.cs.nott.ac.uk/~wss/Publications/DataTypesALaCarte.pdf
>
>
> --
> Live well,
> ~wren
>
> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe at haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>



-- 
Work is punishment for failing to procrastinate effectively.