Grammar support.

Today I added two more rules in the default grammar, in addition to <p6rule>:

rule p6namedrule
    { rule \s+ ([\w|<'::'>]+) \s* \{ <p6rule> \} ;? \s* }
rule p6grammar
    { ^ grammar \s+ ([\w|<'::'>]+); \s* <p6namedrule>* $ }

With this, the PIL^N grammar is now successfully parsed with its Syntax.grammar file.  To support that, tewk and I have implement various contructs:

• Enumerated character classes: <[abc]>
• Complemented character classes: <-[abc]>
• Literals: <'abc'>
• Begin/End anchors: ^ $ ^^ $$
• Partial parse of <p6rule> so it stops at the closing brace when used as part of <p6grammar>.

In addition to that, I've written a simple tree trasforming function for <p6grammar>, which takes a Match object and turns it into compiled rule expressions.  Instead of using TGE-style separate set of attribute grammars as decribed in the Parrot compiler tools plan, I'm associating each rules with consumers, using e.g. SYB Generics to handle post-subrule-parsing production.

However, that scheme means we needed to perform multiple redundant traversals.  I'd much rather something like this, with a conjectural :{} rule trait:

rule p6namedrule
  :{ ($0 => $<p6rule>) }
   { rule \s+ ([\w|<'::'>]+) \s* \{ <p6rule> \} ;? \s* }
rule p6grammar
  :{ Grammar.new(name => $0, rules => hash(@<p6namedrule>) }
   { ^ grammar \s+ ([\w|<'::'>]+); \s* <p6namedrule>* $ }

Some discussion on #perl6 ensued; luqui and I bounced some ideas about embedding attribute grammars into the rule declarations, and promptly went into tangents such as the evilness of keeping $! as global -- $!, $/ and $_ really should all be environmental -- expect some p6l threads about them in the near future...