Re: LR(k) Lojban Grammar

[Chris Bogart <cbogart@QUETZAL.COM>]

chris:
>>>eventually knock wood there'll
>>>be academic interest in the language and we'll want to be able to
>>>define it in a more theoretically understandable way.  So it ought to
>>>be at least a long-term goal.

>This assumes that lojban has an LR(k) grammar...

Well, if ly. has a non-LR(k) grammar, that would be slow to parse using a
generic yacc-like tool for whatever class of grammar it has, then we don't
want to use that parser day-in and day-out.  But we'll still want to do that
work for theoretical reasons.

>I can't find the
>reference right now, but the last I remember was that the grammar
>was not actually completely context-free -- there were some shift-reduce
>conflicts

Do shift-reduce conflicts mean the language is not context-free?  Or do they
mean the language is not LR(1)?

>...that YACC resolved using precedence of rules/productions.
>If S-R or R-R conflicts exist, the language is not LR(1), regardless
>of whether or not YACC can parse it; syntactic ambiguity exists,
>it's just hidden by the mechanics of the parser.

If the parser is the official language definition, then the mechanics of the
parser are part of the definition as well.  So the parser does its job of
providing an unambiguous definition of syntax (after all, it parses a
sentence the same way every time, regardless of conflicts) and also does its
job of proving the language can be machine-parsed.  It just does it in a
opaque way :-(

BTW the conflicts may not be actual ambiguities; they could be places where
a more sophisticated parser would have no problem.

                     ____
 Chris Bogart        \  /  http://www.quetzal.com
 Boulder, CO          \/   cbogart@quetzal.com