JOSEPH RYAN writes: > As bad of an idea that I think this is, I wonder if Perl6's reflection > capabilities will be powerful enough to where a module/pragma could be > written that would be able to do this? For instance, one idea was: > lexically change the current grammar to a subclass of the grammar. In > this subclass, there is a hook on the "variable" deparsing rule that > will implictly declare a variable into its outer scope if it has not > yet been declared in the current scope. Totally whacky, sure; but > doable?
I don't doubt it. I'd write it right now, but we don't know enough about the compiler interface yet. > That brings up another idea that I had just now: will it be possible > to load 2 different grammars at once if they don't conflict with each > other? For instance, say we have loaded a grammer, > Grammar::WhackyVars, that subclasses from the main Perl6 grammar but > modifies the "variable" rule somehow. However, then say I want to use > a grammar that also subclasses from the main Perl6 grammar that lets > you use "happyfunactiontime" instead of the word "class". Since the > modified rules don't conflict with each other, can I just "use > Grammar::HappyFunActionTime" and everything will work? Or will > Grammar::HappyFunActionTime overload the changes done by the > Grammar::WhackyVars? This begs the use of a particular abstraction Perl 6 has introduced with grammars. Maybe grammar munges aren't subclasses at all, but grammatical roles. The only problem I see with such roles is code generation. If we're going to make it parse in such a modular way, we ought to generate code in the same modular way. That will take some clever design (but I think it can be done). This is important insight right about now, since we're rethinking what the parse tree that rules spit out looks like. Luke
