BlueJeans: https://bluejeans.com/426716450

attending: John Swinbank, Krzysztof Findeisen, Scott Daniel, Unknown User (cmorrison), Meredith Rawls, Unknown User (gkovacs)

apologies: Eric Bellm

Topics for discussion

  • LSST2019 DIA sessions: lessons learned, feedback received, etc (Meredith Rawls, others):
    • Many people are interested in using ap_pipe or something similar on their data.
      • e.g. the obs_goto people (James Mullaney)
      • Meredith is considering whether she should follow up on these.
    • Meredith gives comprehensible talks!
      • And in general, many of the talks seemed like they were pitched at a high level.
    • We didn't learn much about the DRP differencing approach.
    • Lots of science collabs don't have a clear ideas about what data products we will actually be providing.
    • Lots of people are interested in dispersed fluxes in the difference images, but this is not one of our deliverables.
  • Porting ImageDifferenceTask to Gen 3 middleware (Krzysztof Findeisen, Unknown User (gkovacs)):
    • “Gabor and I discussed porting ImageDifferenceTask to Gen 3 with Nate and Jim. Supporting calexp templates in Gen 3 will be hard, but the discussion generated some out-of-the-box ideas I’d like to discuss with the AP group”, says Krzysztof Findeisen.
    • The good news: the Gen 3 system has been built from the ground to let you match up regular images and coadds; should work fine.
    • The less stupendous news: the system cannot do differencing of one calexp against another.
      • We would have to write our own quantum graph generator to support this case.
      • How important is this use case? Maybe in LOY1, but mainly for testing.
    • Some suggestions:
      • The Gen 3 system lets us input prerequisite datatypes. Can we use this system, and treat calexp templates as something provided outside of the AP pipeline?
      • Generate coadds made of just one calexp.
        • Note that this involves two warps, rather than just one warp for calexp-calexp as we have at the moment.
    • Have split image difference task so there will be separate tasks for coadds and calexps.
    • It's not clear how much work the Middleware Group is actually planning to do for this.
    • We discussed possible implementations at some length, but for this discussion to converge it seems like we need some concrete use cases.
      • Which we have punted to next week for discussion with Eric Bellm.
  • Table round:
    • Meredith Rawls worked with Ian Sullivan to get an initial ap_pipe run with DCR templates. “Looks pretty good”; a few fewer detections than with the compare warp templates.
    • Scott Daniel has been doing commissioning stuff (testing photometric repeatability on HSC data).
    • Krzysztof Findeisen nothing else.
    • Unknown User (gkovacs) has a discussion with Robert L at the PCW. He provided some suggestions about handling the case when the PSFs in the template and the science image are similar. Fixing this would mean resampling the images to a finer resolution. Discussions ongoing.
    • Unknown User (cmorrison) talked with Andy Salnikov. Moving the PPDB database to Cassandra. Chris has moved some functions into the database wrapper, so they can get ported too.
  • Review CI (https://chronograf-demo.lsst.codes/):
    • Done.