Just be warned, I spent time trying to make the grammar correct and I
just ran into shift/reduce conflicts. I don't think D's grammar
(especially with nested things) can be properly understood by flex and
bison.
I may be wrong, though, I never liked .y and .l files, so it may just be
that I was bad at them.
What I wanted to do, eventually, was try to reuse more of DMD's existing
code. I mean, it's only a few key routines that aren't newline
friendly, and if a proper plan was laid out Walter might even be
friendly to changing them, I'd guess.
-[Unknown]
Arild Boes wrote:
Thanks.
I had not really thought about efficiency of the scanner with regards to
realtime usage, and you are proberbly right about scanning/lexing on a
line by line basis. Maybe the right thing to do, is to write a custom
scanner that accomodates that. I was trying to avoid this however,
because changes in the language, means changing the lexer obviously.
An viable alternative would be to use some scanner generator, that suits
my needs. Now only the grammar needs to be updated.
The DCoder thing sounds interesting.
/ab
Unknown W. Brackets skrev:
I went and rewrote the bulk of the main parser/lexer in a wa
y that was
more VSIP-friendly, which worked out for me. Unfortunately, it's just
a separate thing, and I need to update it badly at this point.
One problem is, you really want to be able to scan/lex by line. If
you don't do some caching/optimization based on that in Visual Studio,
you'll be eating dirt. The design of the parser and lexer in D
doesn't really lend itself to that, as-is.
But, I never went much farther than colored tokens and brace matching,
and it's been a few months since I've had time to mess with it.
You may also be interested in Dave Sieber's DCoder, which was on
dsource iirc, although it was built trying to use yacc/etc. for D,
which didn't really work out well at the time. It was for 2003 iirc.
-[Unknown]
