| I wouldn't mind having a look at this. If you could tell me a little | about the purpose of the dictionary stuff I'd be willing to have a go | at adding this feature. Great! GHC's type checker generates a *translated program* as well as a *type*. The translated program has explicit type abstractions and applications (in the style of System F), as well as explicit dictionary abstractions and applications. The latter is part of the "dictionary-passing" or "evidence-passing" way of implementing type classes, which you can find described in any paper about type classes. Rather than attempt a half-baked explanation, you might want to talk to Don or Manuel, both of whom understand the translation idea. (At least, I know Manuel does.) In Template Haskell, the typechecker also expands splices, so the translated program has the expanded version. The idea, then, would be to print out the type-checked program, as Haskell source that could be stuffed back into GHC right at the front. The main issues I can think of are these: 1. The pretty-printer for HsSyn does not print precisely Haskell syntax. This should be easy to fix. 2. To print Haskell source, you need to remove the extra stuff inserted by the translation. I suggest this is a HsSyn->HsSyn pass, to avoid writing a second pretty-printer. Many of the translated bits are flagged by special HsSyn constructors (e.g. DictLam). I'm not certain that all are. 3. One particular construct, AbsBinds, is what a binding group is translated to. There you will find dictionary bindings and programmer bindings mixed up together. I think the former are all VarMonoBinds while the latter never are, so I think you can disentangle them fine. I'm happy to help you, but the easiest way to do so is probably to create a branch in ghc/compiler (only) so that I can see what you've done and fiddle where necessary. Manuel knows how to do that. Make sure you *tag* the tree just before you *branch* it, so that we can subsequently merge easily cd ghc/compiler cvs update cvs tag th-print-branch-point cvs tag -b th-print-branch Simon | -----Original Message----- | From: template-haskell-bounces@haskell.org [mailto:template-haskell-bounces@haskell.org] On | Behalf Of Sean Seefried | Sent: 21 May 2004 05:30 | To: Template Haskell Mailing List | Subject: Re: [Template-haskell] TH preprocessing | | | On 10/05/2004, at 6:29 PM, Simon Peyton-Jones wrote: | | > Not at present. It's actually a little trickier than it looks, because | > the post-type-checked program (with the splices expanded) has | > dictionary | > stuff floating around, and it's not entirely straightforward to convert | > it back to vanilla source Haskell. | > You're right it would be useful. I think it's feasible with a bit of | > work. I'd be willing to help anyone who wanted to undertake it. | > | | I wouldn't mind having a look at this. If you could tell me a little | about the purpose of the dictionary stuff I'd be willing to have a go | at adding this feature. | | Sean | | _______________________________________________ | template-haskell mailing list | template-haskell@haskell.org | http://www.haskell.org/mailman/listinfo/template-haskell