[Brahms-dev-l] new brat version and global tracking stuff

Hello,

I have updated brat to version 2.17.8, tag BRAT-2-17-8.

I basically added an option to BrModuleMatchTrack, namely the possibility 
to evaluate matching offsets at Finish(), save them to an ascii file and 
read the latter at Init in a second pass.

Technically:

1st pass
BrModuleMatchTrack::SetFitOffset(kTRUE); // false by default
BrModuleMatchTrack::SetOffsetFile("<some file">); 


2nd pass
BrModuleMatchTrack::SetReadOffset(kTRUE); // false by default
BrModuleMatchTrack::SetOffsetFile("<some file">); 

An exemple of how this can be used (and hopefully in an official way) can 
be found in 

~ouerdane/ana/run04/gtr

I have there two bratmain scripts (TrackOffset.C and GlobalTracking.C) 
and a bash script for LSFjob submission.

Now, the neat thing is the following :

This bash script goes in two step:
step 1: matching offsets by group of 2 sequences
step 2: global tracking using tmp offset files from step1

trick: all jobs are launched together (offset and global tracking) BUT the
generated glb track scripts check every 5s that the offset file exist
before bratmain is launched. That's neat because it does not depend on the
whole bunch of sequences but only on the specific sequence that is
processed!


In clear: when the offset job is done for e.g.  seq 0-1 (I grouped offset
jobs by two sequences), the corresponding offset file is produced. The
global track jobs (for seq 0 and independently 1) are then launched
automatically right after the file has been produced. The advantage is 
that if some parameters fluctuate a bit within the same run, this would 
catch it :)

Now, there will be no more mess about which offset files were used since
offsets are generated quasi "on the fly". The other advantage is to get
rid of some ugly hack, namely the old BrMatchingOffsetModule...


One drawback: if an offset job crashes for a reason (who knows ?), the 
corresponding global track job will wait forever since the tmp offset file 
could not be produced. But this should only happen if we're "unlucky" :)
and bkill 0 is easy to type :)

The last thing: the bash script respects the file catalog convention, 
namely : picks up the latest version of the local track files and produces 
the highest  gtr file version number possible (0 if no other version 
exists already in the official dir - data07 for those who don't know yet).

One can also specify the production number, user's decision (p = 1 by 
default). By the way, it would be nice to know more about the file catalog 
soon.


Oh yes: in my bratmain scripts, I used Flemming's modified geometry. 
Hopefully, this will make its way to the geom DB.

That's it.

Enjoy
Djam



-- 
Djamel Ouerdane ------------------------------------------o
|  Niels Bohr Institute      |  Home:                     |
|  Blegdamsvej 17, DK-2100 Ø |  Jagtvej 141 2D,           |
|  Fax: +45 35 32 50 16      |  DK-2200 Copenhagen N      |
|  Tel: +45 35 32 52 69      |  +45 35 86 19 74           |
|                  http://www.nbi.dk/~ouerdane            |
|                  ouerdane@nbi.dk                        |
o---------------------------------------------------------o


_______________________________________________
Brahms-dev-l mailing list
Brahms-dev-l@lists.bnl.gov
http://lists.bnl.gov/mailman/listinfo/brahms-dev-l