Re: [lojban] Testing Lua/LPeg version of the Lojban PEG

[Robin Lee Powell <rlpowell@digitalkingdom.org>]

On Thu, Jun 21, 2012 at 11:08:21AM +0300, Veijo Vilva wrote:
> I,ve now got a preliminary version of the full parser running.
> 
> Basically the parser is just the re-formatted PEG, the rest is
> about 150 lines of quite ordinary Lua code for glue between the
> stages, some very small  help functions and pretty printing. The
> source files (The driver program, the morphology PEG and the
> grammar PEG) presently total 2560 lines (incl. the comments and
> the empty separator lines), about 78 kbytes. There is no binary
> for the parser, which is compiled for each run. The compilation
> time is about 100 ms on any decent PC, which has helped a lot
> during the testing and refinement stage.

Nice!  Go you.

> I've omitted the erasure handling rules as they seemed to cause
> too much slowdown and  rewritten some rules to speed up the
> parsing process. 
>
> I've also added  rules to bracket sumti tcita and
> zei  lujvo. I had to add rules to the morphology PEG in order to
> keep any quoted non-Lojban text intact - now the quoted text is
> sent as a single non-L word to the grammar PEG.

For all your changes that you believe do not change the language,
can you comment on them at
http://www.lojban.org/tiki/BPFK+Section%3A+Formal+Grammar ?

> My present, very simple pretty printer is quite flexible. It can
> produce either the full parse tree, which is probably required
> only for checking the parser, or omit the numbered sub-rules
> (sumti-1,...) or omit any user-defined set of intermediate levels
> from the tree. It would be trivial to add glosses for cmavo and
> gismu to the output. I've also given some thought to passing the
> lujvo split from the morphology PEG.

FWIW, the trick that I've used for programmatic tree pruning, that
works very well, is to prune anything that has only one child.
That, IIRC, is the entire difference between camxes and camxes -v.

> I'll have to do some more testing before releasing the program for
> general consumption. 

You might as well throw it up on github for people to play with,
no?

-Robin

-- 
You received this message because you are subscribed to the Google Groups "lojban" group.
To post to this group, send email to lojban@googlegroups.com.
To unsubscribe from this group, send email to lojban+unsubscribe@googlegroups.com.
For more options, visit this group at http://groups.google.com/group/lojban?hl=en.