Pavlo Korzhyk wrote: > I play with extensive checks and static analysis at compile-time, and > probably not belong to Parrot's target group at all )) > Oh, I wouldn't say that; we will, even in Rakudo - a compiler for a very dynamic language - need to do a bunch of static analysis, if not for error checking then certainly for optimization. The difference probably just lies in what analyses we'll be doing.
I expect that doing the analysis over the PAST will be a common way to go, and I'm expecting to be able to use some kind of attribute grammar approach and be able to write the analyses/transformations in NQP, not PIR (we already do have TGE, a attribute grammar engine, implemented). Then it's just a case of inserting this extra stage into the compilation pipeline (which PCT is designed to allow you to do). I think the real issue here is that while we have the components to do such things, nobody has actually tried to introduce such a stage yet and worked out the pitfalls and things we need to fix up to make it simple. (Note - if anyone reading actually has done this, I'd love to hear about it.) Somebody mentioned they expected it to be fairly easy to do, however... Jonathan _______________________________________________ http://lists.parrot.org/mailman/listinfo/parrot-dev
