[Prev][Next][Index][Thread]
Re: A Dylan parser-generator.
"Michael T. Richter" wrote:
> I'm already aware of ANTLR which is a Java program and which could in
> principle have a Dylan back-end plugged into it. Lately, however, ANTLR 2.n
> seems to have been developing some lethal weaknesses which make it less than
> ideal for the kinds of things I'd like. On the other hand its pred-LL(k)
> grammars make programming simpler. It's a toss-up.
FYI, I started last year a port of ANTLR 2.x (5,6?) to Dylan. It did not
go too far (about 10hours of work total), but already I was seeing parts
of the object modeling where Dylan was better suited than Java.
It was mostly direct (stupid) translation of Java to Dylan, as I planned
to do a second pass to "dylanize" the code after this.
I had given some thoughts to modifying ANTLR to generate Dylan, but
decided against it. I believe that doing the full port is not that
much more work compared to only adding Dylan generation capabilities,
and offer more possibilities in terms of tighter integration with
Dylan (esp. if/when functional macros appear). Anyway most of the work
is in understanding how ANTLR operates, and in providing the necessary runtime
elements for Dylan (like a <bit-vector>, class you can find in the guts of FD,
undocumented, but which would need to be implemented for GD).
I do not have that much spare time to work on Dylan right now, so
I may not get back to the port until summer. I could certainly
give a bit of my time if someone wants to start a port before.
Regards - Eric
--
Eric Gouriou eric_gouriou@hp.com
Follow-Ups:
References: