Takeshi, Mark, Susan, Darrell, George, Kumar, Patrick,
Jeff, Justo, Sandra, David, Wes, Urvashi, Scott, Tak, Kana,
Shinnosuke, Sanjay, Wataru, Jim, Lindsey
Jeff will turn the AIs into tickets.
9:00 (Urvashi) Refactored imaging classes
- AI: (Low Priority) Move Correlation to Stokes Calculation out of synthesis and into a more common place. (Where?)
- Comment: Should the deconvolvers be templated to allow complex later?
9:45 (Jim) New VI/VB
- AI: (Jim) Add a set of "blessed" sorts which people can use.
- AI: (George) -> ! CalSort
- AI: (Kumar) ! ImagerSort -> Standard One
- AI: (Jim) Can we check for pathology against current data set?
- AI: (Patrick) Define the sorts that plotMS uses
- AI: (Justo) Flagging issue (mapping flags set on data back to original MS)
- Do the simple things first (catch cases we couldn't handle)
- Then think about the hard case.
- Comment: This should be the preferred access to all MSs
- Tools: Should look and see if this is appropriate.
- MS ! MetaData -> Probably not appropriate
10:45 (Kana) Status of ASAP Package
- Decision: Developers will need to build ASAP.
- Scott has been working on a script to simplify the full end-to-end build. See CAS-5397.
- AI: (Scott) Set up local cache of ASAP (and casacore) to prevent slow checkout from ATNF - See CAS-5362.
- AI: (Jeff) to have SD pipeline profiled on Socorro cluster, particularly looking at IO shape.
- AI: (Kana) Investigate using VI/VB2 framework for MS processing.
11:15 (Jim) CASA Core Meeting
- AI: (Jeff) Plan to replace Regex with std::regex throughout package
- Comment: Need to evaluate impact on compilers
: (Jim) Look into whether we could we make the statics atomics or remove them?
13:45 (George, Jim) Cal Library and Virtual Corrected Data
- AI: (George, Sandra, Urvashi) Consider using a formal parser (like lexx, pyparsing, or PLY) to parse this and maybe flagdata.
- AI: (Jeff, Jim) Once that's been established, look into a common syntax that can be used (XML?).
Wednesday 19 June
08:30 (Sandra) Flagging module
- Decision: Viewer and !PlotMS should use flag_data interface to set flags
- AI: (Justo) to put back translating of flags into transform iterator (CAS-3679)
- AI: (Darrell, George, Urvashi, Justo, Sandra) to define interface
- AI (Jim): Review design of ! FlagVersion and see if it needs to be updated. Discuss with CASA Core Group (Ger et al., Sandra et al.)
- Decision (Sandra): Suggested changes are endorsed by the developers, Sandra to produce a summary of the shortcomings and the method to address them.
- AI: (Sandra) to implement a new action=unapply in flagdata and modify the existing unapply in flagcmd.
09:15 (Kern) Weight and Weight Spectrum
- Decision: All procedural code should move to using weight spectrum
- AI: (Jim) VI/VB framework to only expose weight spectrum
- AI: (Jim) To consider a compressed storage manager (horizontal incremental) for the weight spectrum.
09:45 (Sanjay) Moving Selection into the VI/VB
- Decision: ! MsIter or ! VisIter to have a constructor / factory that accepts a selection as well as MS and creates ref MS internally. Selection object should be accessible (RO) through the VI/VB interface.
10:30 (Scott) Changes to Build and Distribution System
- Decision: Test branch is deprecated, we will build the trunk to produce test package. See CAS-5396.
- Decision: Stable branches will be generated as needed, rather than always created. See CAS-5399.
1:15 (Scott) Third Party Packages
- Decision: Third party changes to be approved by Architect and Manager. Due two months before the release. See CAS-5401.
11:15 (All) Automated Test System
- Decision: (With Dissent) Per commit e-mail notifications will continue to go only to the suspect.
- Comment: Developers should not be relying on regression test to catch problems.
- Decision: Testing of tools may happen at the gcwrap Python layer.
- Decision: All new code shall have a test committed with it that exercises it. The preference is that you test at as low a level as possible.
- AI: (New Tester) Put the full suite of functional test (runUnitTest) into the nightly testing infrastructure.
- AI: (New Tester) Review existing regressions and remove obsolete tests, add current ones.
- Decision: Any new functional test must be added to unittests_list.txt file
16:30 (Kumar) Definition of REST frame
- Decision: We will define a frame called SOURCE which is defined as the co-moving frame with the source object.
- AI: (Jeff) Work with Butler to define amplitude correction.
09:15 (All) Requirements for plotters, API, etc graphics.pdf
- AI: (Jeff, Jim) Decide on path forward for GUI efforts
- Java/ Freechart
- Agg (Gui Less Operation)
- Decision: Darrell's implementation of the plotter is our path forward in generic plotting. Improvements to features and architecture need to be identified.
- AI: (Jim, Patrick, Susan, Darrell, Jeff) Follow up and clarify overall architecture for various tools.
- AI (Jeff): Establish what is actually needed for "publication-quality" plots.
- Decision: Agreed that indentation style should be 4 spaces (no tabs) for both Python and C++.
13:30 (Justo, Sandra) MMS, Pre-Imaging Parallelization, and mstransform
- AI: (Sandra) In MMS consider having top level subtables not be links, and then doing "flag row" on the subtables (such as SPW, etc) to show what is really in each subMS. (Lower priority)
- AI: (Justo, Sandra) Look at issue with locking and the pointing table in MMS
- Decision: In OpenMP Pragma statements default(none) is required.
Parallelization and HPC
- AI: (Jim) Investigate ! ZeroMQ and discuss with Andrew Grimshaw and folk at GB as possible alternative to MPI.
- Decision: We will investigate replacing the parallel_go framework with MPI or ! ZeroMQ. dBus will continue to be supported for "user level" communication. SAMP will be introduced as necessary to support VO functionality.
9:00 Environment and Startup:
- AI: (Jeff) Define tasks and order to address issues in current system.
10:15 Jim Jacobs CASA and the Environment
- Decision: We should not create any new environment variables. If we have a choice in external packages between environment variables and API interface we should prefer the API.
- AI: (All) Provide feedback on environment variables on CASA SE wiki by August 1.
- Decision: The design for handling these provided by Jim is accepted, to be implemented and integrated.
- Decision: taskinit functionality to be renamed to globaltoolinit, only used in interactive (casapy.py) implementation.
- AI: (Jeff?) Correctly handle passing of C++ exceptions into the python layers. Lower priority.
- Decision: No method should be longer than 80 lines, classes should top out at 50 - 75 methods. This applies to both Python and C++.
- Comment: Reminder that all tasks must use internal tools.
- Comment Consider modifying task system to use a "factory" method that generates the interface from the XML on the fly.
- Comment: Look at decreasing startup time and decreasing number of open files at startup.
- AI: (Wes) Document task XML definitions.
- Decision: Add in long description to the parameters in the Task XML.
- Decision: Include ability to include XML and build up a parameter dictionary.
- Decision: Parameter help files should transition to being auto-generated.
- Decision: We will not implement "expert" parameters as a filter in the task interface.
- AI: (Scott / Darrell) Get the doxygen documentation into nightly build process. See CAS-5400.
- AI: (Jim) Review project book and architecture document and develop TOC (framework) for a living document.
- AI: (Jeff) Organize a place for the memo series.