Hello
Thanks for the input! (And to Alex too).
There is much that I want to add to this project, but first I need to
work on a few others (the one called Skeleton in particular, which is a
code generator for creating code generators...)
I'm not sure the problem with the option grouping - it's not a feature
many people would need, but it serves a purpose that makes sense to me.
I have to apologize for saying the algorithm uses rule rewriting - this
is perhaps not a good way of explaining the process (it doesn't really
rewrite anything). I will try and have an informal writeup of the method
in a week or two and a more formal one later. I think there is a very
nice proof of the method's robustness (though I may be wrong - gut
feeling is it should be all good).
The basic left-recursion handling is similar to the approaches I've seen
described before.
thanks again,
Matthew
On 13/08/13 11:50, Orlando Hill wrote:
Looks great, Matthew!
I would encourage you to keep refining things. Option grouping
shouldn't be needed for specifying precedence levels, and the
difference in generated code when using a multi-option pattern is
going to confuse users.
You mention "rewriting the pattern resolution on the right", so I
imagine the way you're handling left-recursion is similar to the
method I suggested back in 2007. I've already failed twice to describe
that method on this mailing list, so I guess I'll have to get help in
turning it into a published work that people will actually believe.
Anyway, keep up the good work,
Orlando
On Mon, Aug 12, 2013 at 5:28 PM, Matthew Goode
<matt...@scratchy.org.nz <mailto:matt...@scratchy.org.nz>> wrote:
Hello, I'm pretty much ready to release a parser generator I've
been working on for a while. It is a packrat inspired parser
(though can avoid memorising), that handles left recursion
indirectly and directly, and takes into account option order with
left recursion. So, the following
e = { e "+" e } | {e "*" e } | NUMBER
will parse "1+2*3+4" as (1+((2*3)+4) (if specified correctly - see
documentation). This seems to work okay with indirect recursion as
well, and with or without memorisation. It does use a tokenisation
step, because I liked that better, but that may change in future
versions. It also has extensions to PEG that allow it to match
things like Fibonacci sequences...
It can be found at http://scratchy.org.nz/austen.php
(It is still a work in a progress, but for the most part is a well
matured tool - I've been working on and using it for years).
Matthew
_______________________________________________
PEG mailing list
PEG@lists.csail.mit.edu <mailto:PEG@lists.csail.mit.edu>
https://lists.csail.mit.edu/mailman/listinfo/peg
_______________________________________________
PEG mailing list
PEG@lists.csail.mit.edu
https://lists.csail.mit.edu/mailman/listinfo/peg
