Re: LALR (was Re: Working Group Proposal)

Chip Salzenberg <[email protected]> writes:
>According to Dan Sugalski:
>> Might as well go all the way, then--separate out the lexer/parser,
>> the bit that turns the parsed syntax into an optree, the optimizer,
>> and the backend that actually executes perl. Which, thinking about
>> it, isn't that bad an idea [...]
>
>This isn't new.  It's pretty much pluggable right now in Perl 5 --
>just make an XS that does what toke.c and perly.y do, maybe create a
>hook for yyparse, and you're off.
>
>The fact that people haven't used this opportunity may tell us
>something.  Or maybe not.

As one of those well able to do the plugging, parsing and lexing what
scared me off was the black magic of the op "tree" - which is neither
a parse tree in traditional sense, nor threaded code as in FORTH but 
a weird hybrid.  My cache friendly data structures idea might result
in having a tree (for parse, symbols introspection) and threaded code
for execution.

-- 
Nick Ing-Simmons