Re: [Haskell] ANNOUNCE: Generic Haskell 1.80 (Emerald)
Hello, I took the liberty to move this discussion to the Haskell-Cafe mailing list. Adrian Hey wrote:
Thomas van Noort wrote:
Pleasant programming,
Hello,
This looks like good stuff. But having done all this work it seems a pity not to go the extra mm and cabalise this and make it buildable on all platforms (at least ghc supported platforms).
You are right, cabalizing the Makefiles implies that Generic Haskell can be build on any platform that is supported by GHC. However, this requires a tremendous amount of effort since the Makefiles of Generic Haskell are enormous.
The reason I'm interested is this may be useful for the GSoC project I'm mentoring..
http://code.google.com/soc/2008/haskell/about.html
But as things are, I'd be unlikely to consider introducing a dependency on this. Even if it built "out of the box" with cygwin (which it doesn't BTW) I don't really think many windows users will be keen to install cygwin and learn how to use it just so they can build GH.
As you already noticed, there is no Windows binary available for the Emerald release. However, there is one for the Coral release, available from: http://www.generic-haskell.org Although this is an old release of Generic Haskell, this release already supports generic types, which is what you need for your project probably. In the user's guide, there is a small example available which defines a generic type to represent tries. The latest release of Generic Haskell supports generic views for generic types. This allows you to define a generic type that generates efficient balanced tries, as you can read in my Master's thesis :) This requires the use of the balanced view on the type level. Unfortunately, this generic view is not implemented in the Generic Haskell compiler...
Thanks -- Adrian Hey
Regards, Thomas
Thomas van Noort wrote:
As you already noticed, there is no Windows binary available for the Emerald release. However, there is one for the Coral release, available from:
http://www.generic-haskell.org
Although this is an old release of Generic Haskell, this release already supports generic types, which is what you need for your project probably. In the user's guide, there is a small example available which defines a generic type to represent tries.
Thanks, but this won't register itself as a package with current ghc, and even I fixed that problem I guess it probably still wouldn't work :-( It's entirely up to you folks of course. I don't know if anyone at GHHQ cares enought to do anything about the buildability problem. But if not I'm afraid I can't see this being used or accepted as part of standard Haskell infrastructure. It'll just be something people look at, admire the coolness, maybe even tinker with for a while, but never really use for anything serious. Meanwhile I think the GSoC project I mentioned will have to make other arrangements :-) To be honest though, I'm not sure we'd use it anyway as I understand it generates essentially pure H98 and there are ghc extensions we'd probably want to use for performance reasons (like unboxed Ints and unboxed tuples). Regards -- Adrian Hey
On Wed, Apr 23, 2008 at 3:40 PM, Adrian Hey
Thomas van Noort wrote:
As you already noticed, there is no Windows binary available for the Emerald release. However, there is one for the Coral release, available from:
http://www.generic-haskell.org
Although this is an old release of Generic Haskell, this release already supports generic types, which is what you need for your project probably. In the user's guide, there is a small example available which defines a generic type to represent tries.
Thanks, but this won't register itself as a package with current ghc, and even I fixed that problem I guess it probably still wouldn't work :-(
It's entirely up to you folks of course. I don't know if anyone at GHHQ cares enought to do anything about the buildability problem. But if not I'm afraid I can't see this being used or accepted as part of standard Haskell infrastructure. It'll just be something people look at, admire the coolness, maybe even tinker with for a while, but never really use for anything serious.
I don't see GH moving away from Makefiles in the future. First, we need to track dependencies on AG and GH source files, which would require a fair amount of effort to implement in a Cabal-based build. Second, GH installs with no problems on the platforms we use (Linux and Mac OS X systems). As an aside, note that there are other Haskell projects that do not have Cabal based builds such as GHC and Gtk2hs. GH previously supported cygwin-based builds but it is broken due to some cygwin changes. At the moment we do not plan to fix it, but if there is enough interest we could be persuaded to do so. However, if you really need a Cabal-based build without Cygwin or MSYS dependencies, then I guess you are out of luck :).
Meanwhile I think the GSoC project I mentioned will have to make other arrangements :-)
To be honest though, I'm not sure we'd use it anyway as I understand it generates essentially pure H98 and there are ghc extensions we'd probably want to use for performance reasons (like unboxed Ints and unboxed tuples).
You can always use a generated H98 module with a non-H98 module. However, if your generic function uses non-H98 features, it could get tricky. There are solutions for that, though. Cheers, Alexey
Regards -- Adrian Hey
_______________________________________________ Haskell-Cafe mailing list Haskell-Cafe@haskell.org http://www.haskell.org/mailman/listinfo/haskell-cafe
participants (3)
-
Adrian Hey -
Alexey Rodriguez -
Thomas van Noort