Monday Morning Meeting 8/26/19
- DIAL-IN NUMBERS & PASSCODES:
- IP: 192.33.117.12##8110
- Phone: (434) 817-6524
News / Management
- No meeting next week - NRAO holiday
- CUC agenda/material development
- ADASS poster content generation
- CNGI
- Trade study report updates and discussion this week
- Conceptual Design document underway
- CASA 6
- Full attention to 6.0 initial release tarball issues
- CASA team should transition to 6.x development environment
- Dave and Darrell please provide directions on Confluence
Development
- Calibration and Flagging
- Imaging
- Data Visualization
- Import / Export
- Infrastructure
- Simulation
- CASA 6
- What is the formal modus-operandi for developers to get fix or development tested-validated and PR'ed ?
- Documentation: examples with "import..." (potentially many and very spread throughout the docs) would need updates and decisions on what should be public or not. For example, flagdata says:
import flaghelper as fh
antdic = fh.readAntennaList(antfile)
Paths have changed, the PYTHONPATH is not overpolluted any longer, and in this example flaghelper.py is now private.
Build / Testing
- Infrastructure
- Unit
- Verification
- Validation
User Liaison Updates
- Helpdesk/feedback users:
- Frequent crashes CASA 5.6.0 on OSX 10.13.4 (ALMA 16093; Adam Ginsburg)
- CASA commands do not work on Samba-mounted drive (ALMA 16103; Adam Ginsburg)
- In mstransform with multiple MSs, and when specifying both regrid=True and timeaverage = True, the resulting flux and velocity width of the line are wrong (possibly because not all of the spw’s have correct Doppler tracking applied) (ALMA 16021; Ian Czekala)
- CASA 6 with Python 3.6 could not import libraries when using scipy 1.2.1, only with scipy 1.3.1 (Casey Law).
- Since most people start with a virgin ~/.casa tree, it would be very useful to have a (commented) version of the init.py and rc file that can be placed in the ~/.casa tree. I would suggest to add examples of those in the $CASAROOT/etc or data directory, or share? (Peter Teuben)
Pipeline Updates
Single Dish Development (8/22/19)
- Wataru Kawasaki
- Wrote sample codes using PySpark and Dask to fit/subtract polynomial baseline from SD spectra as part of CNGI study
- Used parquet as input data format
- Elapsed time for baselining looks linearly scaling both for PySpark and Dask
- Dask is a bit (about 50%) faster?
- The codes yet to be parallelised. Partitioning parquet data needed in advance?
- Takeshi Nakazato
- 5.6 closeout
- 5.7/6.1 planning
- investigating resurrected debug message in recent prerelease package -> created CAS-12641
- Renaud Miel
- Suminori Nishie