[developers] MRS in PET SVN with selective packing

Stephan Oepen oe at csli.Stanford.EDU
Sat Oct 28 17:21:10 CEST 2006


hi eric,

> I gave the new PET SVN with selective packing a try and am having
> trouble producing MRS using the latest LKB and ITSDB releases from
> the Stanford FTP server (I tried both build 2006/10/19 and "latest"
> which currently points to build 2006/09/05). 
>
> When I run cheap with MRS output enabled, I get a cryptic "No MRS"
> message as follows. My memory is fuzzy from all of the builds and
> retesting, but I think thinks were working when I compiled using
> older LKB/TSDB code from around May 2006. Is there a preferred branch
> or version of LKB for building cheap?

the no-MRS problem is unrelated to selective unpacking (in fact i had
noticed it several months ago, but not needing to extract MRSs in PET, 
i did not find time to look into it until today).  both ann and i have
made non-trivial changes to the MRS Lisp library, and unfortunately we
do not always have the time or opportunity to test against PET.

the problem you encountered is vaguely interesting.  we had introduced
a call to vsym() where its argument was a symbol (`ING') rather than a
string.  the LKB implementation of vsym() is robust to such variation,
but the PET-side implementation was not.  lacking a full specification
of the interface, neither side is to blame :-).

i just checked in a few changes in `pet-interface.lisp' (making vsym()
more accomodating) and `mrsoutput.lisp' (converting `ING' to a string),
so once you update from CVS or fetch the latest `test' build, i expect
the problem should disappear.

at the same time, PET will then support a new output format: `mrs=mrx'
should yield MRS output in XML (as provided by ann).  finally, i added
some debugging code in `pet-interface.lisp': once the ECL MRS library
is compiled with the :debug feature on, it will write a trace of calls
back from ECL into the PET core as the file `/tmp/mrs.debug'.

                                                     all best  -  oe

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
+++ Universitetet i Oslo (IFI); Boks 1080 Blindern; 0316 Oslo; (+47) 2284 0125
+++     CSLI Stanford; Ventura Hall; Stanford, CA 94305; (+1 650) 723 0515
+++       --- oe at csli.stanford.edu; oe at ifi.uio.no; stephan at oepen.net ---
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++



More information about the developers mailing list