dons@cse.unsw.edu.au (Donald Bruce Stewart) wrote:

...

Here is another portability problem in Data.ByteString.Lazy:

Constructor :*: used at 590:22-590:24 is not defined.

It's defined in Data.ByteString.Fusion: infixl 2 :*: -- |Strict pair data PairS a b = !a :*: !b deriving (Eq,Ord,Show)

I did not find it with a quick grep, but yes, there it is. So this indicates a bug in nhc98/yhc's import resolution, not in your library... Regards, Malcolm

On Thu, Nov 02, 2006 at 11:51:48AM +0000, Duncan Coutts wrote:

On Thu, 2006-11-02 at 10:37 +0000, Ross Paterson wrote:

...

Data.ByteString.Lazy defines ByteString as a type synonym, and then uses that in instances, which isn't permitted by Haskell 98. How about defining a newtype ByteString in Data.ByteString.Lazy.Base instead?

It is defined as a newtype in Data.ByteString.Lazy.Base. But that module also defines the strict variant, so within the same module they needed different names. [...] Or we could use a separate .Lazy.Base module for exposing the lower level internals of the lazy version and then not use a type alias at all.

So can we do that? What's the process?

Malcolm.Wallace:

Ross Paterson wrote:

...

...

...

Data.ByteString.Lazy defines ByteString as a type synonym, and then uses that in instances, which isn't permitted by Haskell 98.

I recently sent some patches to the fps package, to enable it to build with nhc98. They simply change these instance decls to use the original type name rather than the synonym. (The original name was even in scope, so there was no compelling reason to use the synonym in the first place.)

Btw, my patches have not yet appeared in the official fps repo. I think dons was going to be responsible for pushing them into base at the next merge point between fps and base.

The patches never arrived! I was wondering when they were going to appear. When they do, I'll apply, test, push into base. Can you resend, Malcolm? -- Don

Am Mittwoch, 22. November 2006 14:18 schrieb Ross Paterson:

On Wed, Nov 22, 2006 at 12:28:33PM +0000, Malcolm Wallace wrote:

...

Ross Paterson wrote:

...

...

...

Data.ByteString.Lazy defines ByteString as a type synonym, and then uses that in instances, which isn't permitted by Haskell 98.

I recently sent some patches to the fps package, to enable it to build with nhc98. They simply change these instance decls to use the original type name rather than the synonym. (The original name was even in scope, so there was no compelling reason to use the synonym in the first place.)

There is the minor issue of which name appears in the Haddock docs.

In the OpenGL package this is "solved" via preprocessing: http://darcs.haskell.org/packages/OpenGL/include/HsOpenGLTypes.h I haven't followed the Haskell' discussion in detail, but I hope that this restriction on instances will be dropped. Cheers, S.