[developers] Follow up: [incr tsdb()] question

[Michael Wayne Goodman]

I haven't chimed in because I didn't think I really had anything to
add, but since I updated the Matrix regression tests to the new format
I may have some useful experience on the matter. I've not had any
problems running up-to-date profiles with an up-to-date LOGON tree. I
did, however, run into problems when mixing up-to-date with
"down-from-date" (i.e. old) profiles, with error messages much like
those described in this thread.

While it seems [incr tsdb()] can read old profiles, it cannot write
them. So there should be no trouble merely comparing, say, and old
gold profile with and updated profile. However, if I am using [incr
tsdb()] to first create the new profile from a skeleton, then the
skeleton files (item and relations) should be up-to-date. I reckon
this is the problem, or something related in the processing pipeline.

I created a script that updates profiles. It's the itsdb.py file in
the python-delphin directory in the project at
http://code.google.com/p/delph-in/. Give it a profile and a new
relations file and it will convert the profile to the format described
in the provided relations file. If you are interested and want help
using it, let me know.

Otherwise, I'd be happy to take a look at your setup if I have access to it.

On Wed, Dec 7, 2011 at 1:00 PM, Emily M. Bender <ebender at uw.edu> wrote:
> Good to know we're not alone :) We haven't found a work around
> at UW ... we've just been busy with other things.
>
> Emily
>
>
> On Wed, Dec 7, 2011 at 12:49 PM, Francis Bond <bond at ieee.org> wrote:
>>
>> G'day,
>>
>> I am getting what is I think a related problem.   On a not quite
>> up-to-date logon tree (due to issues with the new pet), I start logon
>> and load jacy.  I then create  a profile (mrs) which has a fully
>> up-to-date relations file.   I parse it with the lkb.  I then try to
>> treebank it:
>> ------------
>> read_tuple(): arity mismatch (31 vs. 32) for `parse' (11).
>>
>> create-cache(): write-through mode for `jacy/090705/mrs/11-12-08/lkb'.
>> install-gc-strategy(): disabling tenure; global garbage collection ...
>> [04:40:44] gc-after-hook(): {L#66 N=8.6K O=179k E=91%} [S=845m R=552m].
>>  done.
>> flush-cache(): flushing `jacy/090705/mrs/11-12-08/lkb' cache ... done.
>> close-connection(): `jacy/090705/mrs/11-12-08/lkb' expiry.
>> -------
>>
>> So it looks as though itsdb is not creating the parse file correctly,
>> even without cheap in the loop.
>>
>> On 7 December 2011 20:23, Stephan Oepen <oe at ifi.uio.no> wrote:
>> > hmm, i had to check in SVN, but assuming you are using the
>> > LOGON 'trunk', the 32-bit binaries were re-generated not too
>> > long ago.  i cannot easily imagine that 32- vs. 64-bit should
>> > make a difference here.
>> >
>> > 0 oe at mv (~/src/logon/lingo/lkb) 11 $ svn log linux.x86.32/logon.dxl
>> > |head
>> > ------------------------------------------------------------------------
>> > r10216 | oe | 2011-10-27 17:28:23 +0200 (to., 27 okt. 2011) | 1 line
>> >
>> > new 32-bit LOGON run-time
>> >
>> > the 'parse' relation was indeed extended by two fields a short
>> > while ago, so it would seem that the installation that varya has
>> > ends up writing the old format.
>> >
>> > does the 'relations' file in the broken profile contain the new
>> > fields (p-input and p-tokens)?
>> >
>> > since you have been silent about this problem for the past few
>> > weeks, has there been some progress in the meantime?  or
>> > just patient despair?
>> >
>> > cheers, oe
>> >
>> >
>> > On Tue, Nov 15, 2011 at 04:30, Emily M. Bender <ebender at uw.edu> wrote:
>> >> Hi all,
>> >>
>> >> Following up on the question that Varya and I had before:
>> >> The ^M are gone, but we still are not getting to working profiles
>> >> on Varya's system (Ubuntu 11.10).  I checked (svn update)
>> >> and logon is up to date.
>> >>
>> >> However, when we run the attached grammar over its associated
>> >> test suite on her system, we get a broken profile.  Specifically, the
>> >> number of fields in the parse file appears to be incorrect.
>> >> (There are two fewer @ signs per line in
>> >> tsdb/home/[...]/Varya-lkb/parse,
>> >> as opposed to Emily-lkb/parse, done on my system.)
>> >>
>> >> Looking at the run file, it looks like I'm using the 64-bit software
>> >> while Varya is not.  Is there perhaps some change that didn't get
>> >> made in the 32-bit version?  Or are we barking up the wrong tree
>> >> here?
>> >>
>> >> Thanks,
>> >> Emily
>> >>
>> >> --
>> >> Emily M. Bender
>> >> Associate Professor
>> >> Department of Linguistics
>> >> Check out CLMS on facebook! http://www.facebook.com/uwclma
>> >
>>
>>
>>
>> --
>> Francis Bond <http://www3.ntu.edu.sg/home/fcbond/>
>> Division of Linguistics and Multilingual Studies
>> Nanyang Technological University
>
>
>
>
> --
> Emily M. Bender
> Associate Professor
> Department of Linguistics
> Check out CLMS on facebook! http://www.facebook.com/uwclma
>



-- 
-Michael Wayne Goodman