Special NAUG - 2007-Feb-02


* data selection -
    • Need to modify min match to wild carding
    • replace msselect parameter

CASA 2007 Calendar

  • Update: Crystal will be in Socorro from Feb 4 to Feb
  • ALMA TST5 Pre-testing starts Feb 8

User Support - current week

  • 1. logger

It takes 10 to 15 sec for the logger to appear after running casapy. This is much longer than before. My laptop has 500M of memory and is doing other things, so this my be the problem. Sometimes, after appearing for less than one second, the logger disappears. The unix ps command shows 19571 pts/10 00:01:03 casalogger after this happens. I have no idea where it is and how to get it.

casalog.log can get very large and should be cleaned up. We're looking at performance issues (for tens of thousands of lines) but if you purge it, or move it after a session, it should be okay. You can call up a new logger, should it go down, by typing 'casalogger' at the CASA prompt (in next stable).

  • 2. tasklist

The list of available tasks that are shown when casa begins is not identical to the list when you type . For example, flagdata has become flagger. Do these lists come from one basic list of names?

Not very sophisticated but they aren't derived but are updated by hand. Fixed the flagger-flagdata issue.

  • 3. Segmentation fault:This should be fixed globally now. The RPMs were apparently made without a flag that we need for error handling. This has been corrected.
zvis was not defined, so that vis was not properly set.  Task should
have not segmented.  I did get a message when I set vis=zvis that the
data set is not defined, but it should also have caught this problem
when I mistakenly ran correct anyway.

taskname = 'correct'
vis = zvis
msselect = ''
gaintable = 'f1.cal'
bptable = 'bp1.cal'
gainselect = 'FIELD_ID==1'
gaincurve = False
opacity = False

Fixed vis=zvis, but the error message:
 Sat Jan 27 15:06:16 2007      WARN Calibrater::setdata:
Selection is empty: reverting to sorted MeasurementSet
was uninformative.  Execution was finished, but I'm not sure what
happened or
what was wrong.

Bottom line:  'correct' needs a little more bullet-proofing
  • 4. SPLIT:
I am confused about the nature of nchan, start, step and width

In gaincal, for example, to get a solution for channels 10 to 50 I used,
    nchan = 41; start=10; step=1
In split, to obtain these data averaged in the output, I needed
    nchan = 1; start=10; step=41
    shouldn't this be width??

   I believe this is correct, but the syntax is is confusiong.
   The use of BCHAN, ECHAN, NCHAV (maybe NSTEP?) is much easier to

Also, the script

taskname = 'split'
vis = zvis
outputvis = 'P2229_S0_avg.ms'
fieldid = 1
spwid = 0
nchan = 1
start = 10
step = 41
timebin = "-1s"

gives me the identical output for
timebin = "-1s"   and
timebin = "300s".

So, it doesn't average the data.  I looked at split.py and saw
start=start,step=step,whichcol=datacolumn).  Is timebin missing after

By the way, why not just define timebin in units of seconds, input as
a floating point number.
  • 5. Viewer:

contouring: I think that the proper definition of the contour levels in viewer should have high priority. The level type fractional gives 1% contours only, with only integer contour available (although negatives are allowed but at the present time they cannot be different than the positive contour). I don't know what the definition of abs contour levels mean. For any reasonably high dynamic range image, I don't know how to get contour levels at the 3-sigma limit, for example.

You cannot choose specific negative contour levels. They must be the same as the contour level. Both position and negative contours should be specified in contour level.

I would like to blink between two contour images. It seems that only raster blinking is supported. In fact, with the two images loaded into the view, I don't know how to change back and forth between contoured images. I tried the register button to include one or the other image, but that killed viewer.

Deleting image: When I try to remove images in the view, using the 'Data' button on the left top of the view, choosing close, and then choosing the image to close, about 40% of the time the view aborts and has to be started over.

There is virtually no documentation on the viewer in the cookbook and online. I don't think there are any popups with the buttons to help document what they do.

Is there any scripting interface with the viewer as there was in aips++. Can I view a contour levels with defined specifications (like contour levels, area, lebeling, other things (see TVLOD)before invoking 'viewer' so that I only have to go the adjust button for additional specifications of the image properties.

  • 6. exportfits:

I wanted to export an image so I could view it in aips. The task exportfits is not available when the list of available tasks comes up when casapy is started. However, when in casapy it much more convenient to type to refresh my memory of what is available - and there is exportfit listed. I can also type taskhelp and I think this list is independent. For example the flagdata task is only listing when casapy is started. If these task availability listings do not come from the same list, they should. This may hold for other information as well. You don't want the same information listed in duplicate lists. This is identical to my flagger comment above.
  • 7. Image tool and imagefits:

The uvfits was in the image tool in aips++. Unfortunately, this tool has not been pythonized yet, so there is no way of getting an image out of casa in casapy. You have to go the aips++, I think. Perhaps, in the short term, the uvfits reader and writer can be incorporated into the ia (image analysis utilities tool). I successfully made the fits file using aips++.
  • 8. Catalog;filecatalog is now available, though we encourage the use of the unix shell as this is the simplest most direct route.

I like the aips++ catalog routine and some of the interaction with it. Can this be implement at some level? Most of my deletions of incorrect gain tables and measurement sets were done in unix which is okay.
  • 9. Clean:

The clean boxes work! Thanks.

Some parameter considerations: I think bmaj, bmin and bpa (clean beam size) should be included.

It is not at all clear that if you want to clean in the same way as IMAGR in aips, then you should choose the alg='csclean'. The other options will only clean accurately the inner quarter area and are more like APCLN in AIPS which hardly anyone uses. These options need more documentation in several places.

Additional parameters needed in clean are timerange, antennas

Do you need uvfilter at all? If uvfilterbmaj = 0.0, that is the number to decide on filtering. Also, I think users would be more comfortable with uvtaper (the inverse relationship) rather than the filtering.

rmode is unfamiliar to most of use. It needs a bit more information in help clean

  • 10. Calibration files:

The need for tying calibration files to the relevant data set is becoming more clear. For example, fluxscale has the visibility data set as input because it needs the visibility data set to translate fieldids into source names. As yet, plotcal does not point to a measurment set, and this is the reason fieldid and antennaid cannot be translated.

My general feeling is that these calibration tables should become tables within the structure of the measurement set. These are already many table directories under the measurement set, so why not add calibration tables. In talking with George about various calibration concerns, he mentioned that Robert Lucas is working on the calibration tables for the asdm and quick-look, and the casa group is waiting to see this implementation. I believe there is still the consideration of how 'independent' the calibration tables should be with respect to the measurement set. This is an important decision.
  • 11. Casa parameter defaults:
I notice that the default parameters for casa parameters are listed in
the casa.py file in the appropriate directory, in this case


I guess we are stuck with global defaults, rather than different defaults
for different tasks.  I would also recommend changes some of the defaults
such as:

fieldid = -1   instead of fieldid = 0
datacolumn = 'data'   instead of  datacolum = 'corrected'
field is in the list twice
multiplot = true    instead of   multiplot = false
alg = 'csclean'     instead of   alg = 'hogbom'  (ie APCLN)
opacity = false     instead of   opacity = true
gaincurve = false   instead of   gaincurve = true
plotsymbol = '.'    instead of   plotsymbol = ','
timebin = 0.0       instead of   timebin = '-1s'  (see above comments on
timerange = ['01-JAN-1980/00:00:00.0','01-JAN-2010/00:00:00.0']
                   instead of   timerange = ''
  (It is difficult finding the various syntax for timerange.  If you
use the
   default above, it is obvious what format to use if you really want
to use

verbose = true      instead of   verbose = false
  • 12. plotcal

I suspect plotcal is being worked on, but here are some suggestions which might not yet be considered.

a. The plotted points are too large and blend into each other. Need a parameter to scale the plotting character size.

b. As discussed above, the input vis is needed to translate id numbers

c. A range parameter for the x- and y- axes are needed
  • 13. fluxscale

Since fluxscale already points to the measurement set, the fluxscale pythong script should automatically run setjy on this source(s). At the present time, the flux that fluxscale determines is not available, I believe, so that in running setjy on your own, you have to write in the value(s) specifically. Perhaps a flag in fluxscale indicating whether to apply or not apply this value to the measurement set is useful.
  • 14. HELP information?

I can see the text information for the inputs to any task, and making changes is simple. But, where is the information used for the help files? I believe that these should be an easily accessible and changeable, so that changes to the inp and help functions for any task can be more easily coordinated.

  • DS: Enhancement request:

In the filler, please flag autocorrelations by default. Add a parameter to the task inputs that you can fill them if you want. (request by Mark Claussen and it is a good one but should be low priority)
  • User error reporting:
setjy ('vis=src.ms',field='1331+305')
 # typo created major error messages that are not helpful:
 # Exception Reported: Table
 #does not exist
 #Fri Jan 26 20:22:29 2007      WARN :
 #No table specified, please open first
 #Exception Reported: Table
 #does not exist
 # etc...
   # just say that the vis file "vis=src.ms" doesn't exist and leave
     it at that.  

  • Asthetics change request:
browsetable ' src.ms'
 # why does this come up as a small box in a large white screen?
 # why not bring it up so the table fills the available space?  

  • bug:
 # gives an error - details not reported here because I've already
   worked them with Joe:
##### Joe figured it out, selection between fieldid and field is not
      handed off correctly.  eventually get rid of fieldid but not
      until casa has wildcard parameters in the field names.  Minimum
      match has too many problems getting multiple sources.  

  • bug:
plotcal gets TPPlotter error when trying to plot a gain table:
    msselect='FIELD_ID IN [0] && DATA_DESC_ID IN [0,1]',
 # OK, 2 good solutions found

 # a bunch of code traceback, I get a blank screen.
   # Fri Jan 26 22:10:23 2007    SEVERE :
   # Exception Reported: TPPlotter: From Python.
   - problem is repeatable. 
  • Inconsistency:
Note: plot cal is marching through the antennas in numerical
      sequential order, This suggests that it is showing antenna_id
      rather than antenna_name like plotxy.  This should be made to be
  • Enhancement request:Behaves as for other plots.
When I stop iterating in plotcal or plotxy, the image goes away.  I
would like it to stay up on the screen so I can see the last plot. 

  • Problem with understanding:
# why do I need a visibility file for fluxscale?  Isn't it still just
  operating on the cal tables????  This is disturbing.  
It is used for referencing information from the original MS. This is also planned to be used for then setting the value without requiring an additional step

  • bug:
fluxscale causes segmentation fault:
fluxscale(vis='src.ms',caltable=' src.gcal',fluxtable='src.fcal',
    reference='1331+305', transfer=['1820-254','2007+404','1925+211','2015+371'])
 # Fri Jan 26 23:15:23 2007      WARN :
 # The following transfer fields have no solutions available:
 # Fri Jan 26 23:15:23 2007    NORMAL : Found reference field(s): 1331+305
 # Segmentation fault (core dumped)
   # src.fcal not created.
# Summary shows that all field names are correct...  solutions were
  obtained for field-ids 0,1,3,5,7 which are the sources I list
  above.  So this doesn't appear to be a wrong transfer field. 

# restart casa and try again:
fluxscale(vis='src.ms ',caltable='src.gcal',fluxtable='src.fcal',
 # Fri Jan 26 23:23:42 2007    SEVERE Calibrater::fluxscale:
 # Caught Exception:  Cannot find solutions for transfer field(s)
 # but this time it did not cause a seg fault.
 # maybe since the transfer fields were appended there is a problem?  
  • Review last week's work
    • Dailys in CV:
    • Ed:
      • - clarification on antenna based flagging
      • - log mouse copying
      • - exportuvfits improvements
      • X refant confusion (need to confirm)
      • - fluxscale
    • David:
      • X bandpass documentation
      • X autoflag segfault
      • X autoflag parameter type issue (fixed in tomorrow's daily)
      • X gaincal defaults Changed defaults to include calwt True for opacity,gaincurve
      • - importvla bandname issue
    • Debra:
      • X msplot problemsSee above
      • - mplot data selection issues
      • X regression failure (used old script)
      • X plotxy problems (.matplotlib problem)
      • X setjy confused?fieldid/field selection hierarchy confusing. Consolidated to single parameter='field' which will take either an integer or list of integers or a string.
    • ipython/python versioning - use 'special' ipythonrc_casa
      • Writes to local directory .ipython_casa directory in local directory. Should this be the same place as the others, i.e., $HOME/...?
    • EF: seg faults - data corruption or other? We think this was a combination of effects from possible actual data corruption between conflicting versions (running the pipeline version), rpm built without proper flags, and an actual bug in fluxscale.

Alpha Release notes

Enhancements/Defects resolved

New functionality

-- JosephMcMullin - 26 Jan 2007 * flag_casapy.html: flag_casapy.html

* flag_casapy.html: flag_casapy.html

-- JosephMcMullin - 28 Jan 2007
Topic revision: r5 - 2013-02-13, PatrickMurphy
This site is powered by FoswikiCopyright © by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
Ideas, requests, problems regarding NRAO Public Wiki? Send feedback