[Tpe] Off Line Software

[Mac Mestayer]

Hello Larry and all;

I agree with you that our software is more complicated than it
needs to be to do its job.  In my opinion, this comes about partly
from duplication of code (RECSIS <-> a1c, seb <-> pid, redundant
run index tables, trk_mon <-> trk_mon_lite and probably others),
but more importantly from duplication of effort, in this case,
throwing grad student after grad student into the calibration morass
when a few good post-docs and dedicated grad students assigned to the
various sub-system experts could bring stability to the system.

I do think you chose bad examples to illustrate this, and that this
will confuse the real issues; hence my reply.

First, trk_mon: trk_mon (short for track monitor) was initially
just a simple program to turn the tbla bank of drift chamber
information sorted on a layer-by-laywer basis into an ntuple.
The drift chamber calibration program, dc3, used the information
in this ntuple.  Over time, people began adding so many new
applications ("creeping featurism") to trk_mon that it became
unstable.  Trk_mon_lite was produced by simply stripping
unneccessary (non-DC) code out of trk_mon.  If in fact the
meaning of the proton selection cut actually flipped, then this
is a simple bug and must be fixed.  This is the first I've heard
of this even though trk_mon_lite has been around for five years or so.
Why didn't you or your students talk to me about this?

Second, on the issue of different track reconstruction efficiency
for in- and out-benders: this is probably not a software issue at
all but a subtle analysis issue.  Last August I heard about the issue
and suggested you look at the sector dependence: a sizeable sector
dependence probably indicates a hardware problem due to dead areas
in the DC and TOF while a sector-indepence might indicate a more
general issue.  An example of a general issue is the focussing in
phi for inbenders compared to the de-focussing for outbenders.
This can be studied by varying the fiducial cuts in phi.  At any
rate, it is unlikely that this is purely a software problem.

Finally, and the reason I chose to respond, is to point out that
we should fix ou software problems now, for our present and near-term
analyses.

I'd be glad to be part of a team to understand the dependence
of efficiency on track charge.

Pawel Nadel-Turonski is my go-to guy for software issues.
Pawel, could you contact Larry for particular details about the
proton selection cut in trk_mon and trk_mon_lite?

 					- Mac

"mestayer at jlab.org", (757)-269-7252

On Wed, 16 Feb 2011, Larry Weinstein wrote:

> Dear Folks,
>
> As I watch the TPE people struggle with the calibration and
> reconstruction software, I am amazed at how needlessly complicated it is.
>
> Let me give one very small but telling example.  There are two programs
> used to calibrate the drift chambers, trk_mon and trk_mon_lite.  The
> user should specify that he wants to calibrate using protons.  However,
> the '-p' flag includes protons in one program and excludes protons in
> the other.
>
> A much larger example is that the reconstruction efficiency differs by
> about 10% for inbending and outbending tracks. We do not understand why.
>
> In general, we appear to lack a coherent, well understood, offline
> software package.  This needlessly delays and complicates analysis and
> hence publication of our results.
>
> It is too late to fix this for CLAS6, but I think that it is crucial to
> remedy this for CLAS12.
>
> --
> 				Sincerely,
> 				Larry
>
> -----------------------------------------------------------
> Lawrence Weinstein
> University Professor
> Physics Department
> Old Dominion University
> Norfolk, VA 23529
> 757 683 5803
> 757 683 3038 (fax)
> weinstein at odu.edu
> http://www.lions.odu.edu/~lweinste/
>
> _______________________________________________
> Tpe mailing list
> Tpe at jlab.org
> https://mailman.jlab.org/mailman/listinfo/tpe
>