Hi,

Yes, Michael's hack would do part of the trick, but I forgot to exemplify part of my question in the example code.

The whole idea is that I would like to be able to splice the generated Type into a TH quotation. Consider one of the world's most complicated identity functions:

{-# LANGUAGE TemplateHaskell #-}

import Data.Typeable

import Language.Haskell.Meta.Parse (parseType)

import Language.Haskell.TH

f :: Q Type -> Q Exp

f m = m >>= \t -> SigE (VarE 'id) (FunT t t)

liftT :: Typeable a => a -> Q Type

liftT x = either fail return $ parseType $ show $ typeOf x

-- separate files

splicef :: Typeable a => a -> a

splicef x = $(f (liftT x)) x

This code does not work because liftT depends on the value of x, exposing TH's state restrictions, but since we only need the type of x to evaluate liftT, we should be fine.

Thanks,

hugo

On Thu, Oct 16, 2014 at 1:22 AM, Michael Sloan <mgsloan@gmail.com> wrote:

Here's a hacky solution to this, using haskell-src-meta to parse the type:

{-# LANGUAGE TemplateHaskell #-}

import Data.Typeable
import Language.Haskell.Meta.Parse (parseType)
import Language.Haskell.TH

liftT :: Typeable a => a -> Type
liftT x = either error id $ parseType $ show $ typeOf x

This doesn't handle qualification properly, as the instance of Show
for TypeRep doesn't qualify names. A proper solution would involve
directly writing a (TypeRep -> Type) function.

On Wed, Oct 15, 2014 at 9:06 PM, Hugo Pacheco <hpacheco@gmail.com> wrote:
> Hi list,
>
> I am 99.9% sure that this is currently not possible, but I might as well
> ask:
>
> Is there a way to lift type variables into Template Haskell type splices?
>
> What I had in mind would be something like this (erroneous) code:
>
> liftT :: a -> Q Type
> liftT (_::a) = [t| $a |]
>
> I have no idea how hard it would be to implement such a feature, or if it is
> remotely doable.
> Naively, it seems to me that TH would have to delay evaluating the splice
> until the type variable is fully expanded, but all the necessary information
> would still be available at some point during compilation.
>
> Cheers,
> hugo
>
> --
> www.cs.cornell.edu/~hpacheco
>

> _______________________________________________
> Haskell-Cafe mailing list
> Haskell-Cafe@haskell.org
> http://www.haskell.org/mailman/listinfo/haskell-cafe
>

--
www.cs.cornell.edu/~hpacheco