CASA 4.3.1 / Pipeline Cycle3 R1 release EVLA pipeline accepted
Pipeline F2F meeting in Socorro, pipeline working group and developers
Parallelizing CF calculation.
Execution on AWS of full pol imaging & w-projection.
Implemented the first version of importasdm + partition. There is still an issue when running it with simple_cluster (TBV), but it works fine with MPI. Updates will be posted to CAS-7369.
Several discussions about the inclusion of spw selection in the online flags of importasdm. The online system started to populate the spectralWindowId in the Flag.xml file, but did not coordinate the support of this in CASA. They also need to include a numSpectralWindow in Flag.xml in order to get the spw selections in the flag commands. ICT-1736, CAS-7623.
Verified test_setjy on MMS. Reported to Tak that some errors only happen in RHEL 5, when the tearDown method is called. Using shutil.rmtree() seems to be the issue. The tests pass in RHEL 6.
Trying to build CASA from scratch on OSX 10.9.
Preparing a list of topics to be discussed next week in the HPC working meeting in Socorro.
(CAS-7329) Casacore LatticeHistograms gives a segfault when setting a range outside of data range: Fixed but unable to commit because of casacore freeze.
(CAS-7621) Residual image not correct when fitting a single change of an image cube with a 2D fit: Fixed (RTT)
(CAS-7584) no random number seed in ia.addnoise(): Responded to reporter's objection to name of new ia.addnoise() parameter of "seeds" vs "seed", pointing out that the underlying RNG for LatticeAddNoise in casacore requires two seedS.
More conversations re: visstat. Folks (this has now come up several times) seem to be under the impression that I wrote the current visstat code, or at least know the intentions behind it. Answer: False. I did not write the current visstat code, nor do I know its intent, nor do I even suspect that there is a design document explaining that information. I do know that at least some of the current visstat code uses LatticeStatistics, which was a poor design choice IMO. I do know that visstat should be reimplemented to use the new stats framework, which I did write. My ability to answer questions is limited to that domain as far as visstat is concerned.
(CAS-7628) msmd.scansforintent returns all matching scans in a multi-obsid dataset, regardless of the obsid parameter. Fixed (RTT)
(CAS-7622) msmd.*forscan(s) methods should default to "any" obsid. Currently seeking approval from major msmd stakeholders re: reporter's change in default behavior proposal.
(CAS-7257) discrepancy between spectra extracted by imval and by spectral profiler in viewer: Responded to reporter that I did not know the root cause of the discrepancy he reported, that I only could tell him that the numerical results determined by the viewer and imval were in agreement. He later confirmed this to be the case and closed the ticket.
Spoke with Urvashi about more ia warnings that are produced by imaging code.
Wasted a half day trying to debug test_msmd failure on OSX that I had forgotten I have a fix for but cannot commit because of casacore freeze. Looking into removing this test from automated test suite so I don't waste similar amounts of time on this red herring in the future since I'm obviously having issues remembering and juggling in my mind all the fixes I now have for casacore that I cannot commit.
(CAS-5524) import uvfits should do make consistent when getting visibilities in STOKES form: Asked for and was provided with sample data set that exhibited the issue
(CAS-3934) importuvfits needs to trap and flag 'NaN's coming into the MS: began investigation into this issue
(CAS-7287) Extensive testing of sort column behaviors in context of solving modes, in prep for the migration of solving to the VI2/VB2. Most interesting finding: MSIter is effectively ignoring the user-specified time interval unless TIME is the last (fastest) column specified for the virtual sort (because the formation of SORTED_TABLE is not using the interval to pace the TIME column). This appears to be fixable, in a way that should enable fairly flexible virtual sorting (e.g., all spws for a time range vs. all time for a spw, among others) without disproportionate I/O performance risks, especially if proper quantization of tiles and flexibility in the tile cache is also introduced.
Continued adding test programs for tclean and fixing issues along the way (save model plumbing, starting model plumbing, choosing model when multiple image fields overlap, log messages, etc...)
Some code cleanup in ImageStore and fixing miscellaneous normalization issues ( since it's the same code for single/multi facets, single/multi taylor term, gridders with and without weightimage, serial/parallel ).
Tested TT's automasking function and added basic PB based mask merging
Updating local data repository to begin checking in test program material.
more familiarization with mstransform code
removed more compiler warnings from flagging and mstransform
helped triaging a local plotms + mpi issue
Investigation and testing on setjy with an MMS after the report of the setjy failure on MMS. Fixed the sub-parameter defaulting issue for SSObject setjy case.
Investigated CAS-7484 ("tclean crashes CASA when using spw frequency selections with gaps"). The error seems to occur with multiple channel selections in spw, not because of there is a gap in the selection. It failed with array conformance error in Major cycle process when vb.visCube and vb.modelVisCube ended up with different shapes.
Friday NAOJ Meeting
Introduction to Renaud
visstat code investigation and discussion with Masaya
sdimaging: support datacolumn selection ( CAS-7586): I found the current implementation, reasonable. Resolved without working after discussing with Takeshi.
sdimaging: beam size definition of MSes without REST_FREQUENCY column ( CAS-7598): done fix. wil resolve after adding a couple of unit tests.
extend averaging weight support ( CAS-7450): investigated statwt code. Hoping to talk to George M. next week.
worked on CAS-4771 (gaussian smoothing task)
reported a bug in visstat (wrong value for quartile)