From rmarson@aoc.nrao.edu Fri Oct 26 17:11:28 2001
Date: Fri, 26 Oct 2001 16:22:08 -0600
From: Ralph Marson <rmarson@aoc.nrao.edu>
To: aips2-naug@cv3.cv.nrao.edu
Subject: msplot status

Hi Everyone,
   Ed's recent email reminded me that I was going to send you all an
update on msplot. 

Firstly I'll tell you what has happened since my message of October 16.
* A widget now appears that will translate the cursor position into,
  field id (and name), spectral window id (and freq.), time (as the
  time axis is inadequate), and interferometer number into 
  antenna & station names. 
* You can now use a box region to select a region that is smaller than
  one pixel on either (or both) axes. Previously you needed to use the
  point region to do this. 
* The problem of the viewer not releasing the temporary image has
  been minimized. It may still occur but in my experience much less
  frequently. If it is happening to you often you should tell me and I
  will increase the priority on this defect.

I am quite happy with the performance of msplot, in raster mode, on
synthesis data sets. I will need to test it some more on single dish
measurement sets. I also need to improve the speed at which the flag
columns are copied.

I'll now move on to addressing some of the comments in Greg's message
of October 17.
> Note that the
> data is rotated by 90 degrees and mirror-imaged compared to 
> what is shown in TVFLG.  Antenna numbers go up the Y axis
> starting from 1, while time increases to the right.  To find out
> about a pixel place the cursor over it.  The Y axis value
> will appear as something like 6022.  This decodes to baseline
> 6 with 22.  The formula is Y=antenna1*1000 + antenna2.  Ralph
> is working on a little widget to make it easier to decode the
> output from the viewer.
Time is by default placed on the x-axis because this will usually be
the longest axis and your screen has the most pixels along this
axis. However it is easy to change using the gui brought up when you
press the 'Adjust' button. These options are saved to disk and will
not be reset until you use msplot on a different measurement set.

The widget that translates interferometer number into antenna & pad
names is also done (as described above).

> One problem that we noticed right away was that antennas that
> were out at the beginning of the run get dropped for the 
> entirety, even if they come back in later.  This was bugged by 
> Ralph:
> 
> Bug: AIPS++ defect AOCso02253: Request gridding different data 
> 
> And we think this same defect was also causing aips++ to crash
> (abnormal child termination).  But it was still possible for me
> to call up msplot from glish.  Which is what the remainder of
> this e-mail is based on.
Both the problem and the crashes related to it have been fixed.

> Another problem immediately obvious was that the startup of msplot is a
> bit slow, about 2 minutes for a standard VLA continuum observation
> lasting about 1.8 hours (212000 visibilities).   About 50 seconds
> is spent copying the flag table to enable the user to revert to
> a previous flagging if so desired.  Ralph says it takes so long to
> copy this table because it goes through glish.  He is thinking
> about ways to optimize it.
As Ed mentioned I have sped things up by about 50% but hope to do much
better. I expect this will happen next week

> We also noticed a lot of error messages being directed to the
> glish interface.  Ralph will be turning some of these off.  Others
> will probably stick around.
You should not see any of these.

> A feature quite notable in its absence is the ability to display 
> the rms of amplitude, or an amplitude difference.  The only display
> options are amplitude and phase.  Time averaging doesn't appear to
> be an option either so it will be interesting to see what happens
> when something more substantial than this little test data set 
> is loaded.  
Greg has submitted an enhancement request for this, but it will not
happen in this release.

> Interactive clipping of high visibilities is another useful feature
> that is currently missing in action.
Again this feature is not planned for this release.

> The scope of the edits can be modified using the buttons on the viewer.
> This covers what is needed, but there is a minor bug that prevents
> all correlations from being flagged.  Ralph is working on this and
> I think the fix will be in tomorrow's "daily".  
As far as I can remember this is fixed.

Finally I urge you all to test msplot. Many of you know I will be
leaving the aips++ project on December 1 and the maintenance of msplot
will be picked up by someone else. My experience is that it takes some
time to become familiar with this code and as I have this experience
comments that you make now can be acted upon much quicker than
comments made after I move to alma.

Ralph