Discussion leading to this issue:

> > In your case (full concrete object syntax rules), it might actually be
> > very useful if Stratego could match on the types (syntactic sorts) of
> > trees. This hasn’t been done before, but in principle it is possible,
> > by just using a different implosion from the parse tree to the AST
> > that annotates every term with it’s type. Also, variables meta would
> > have to match with this annotated type.
> >
> > If this would help you a lot, then we could consider to work on this.
> > Please let us know … It’s a non trivial fix, so please consider how
> > much it would actually help you.
>
> In retrospect, it seems that this would have been very useful to have,
> since it would allow us to avoid cluttering the AST with not strictly
> necessary constructors as we do now. Things are generally working pretty
> well, and we’re finding the only hassle/trouble we currently have is
> related to this issue, which means we have to hunt through the grammars
> and either adapt our rules or introduce constructors. Having this happen
> ‘automatically’ would also have the benefit that it would impose some
> small amount of discipline on the use of metavariables – since, as you
> say, they would have to match the type of what they bind to.

Submitted on 19 June 2006 at 13:47