Fix Version/s: None
Sprint:Science Pipelines DM-S15-4, Science Pipelines DM-S15-5, Science Pipelines DM-S15-6
Team:Data Release Production
This is a transfer for changesets for HSC-1062.
Unlike most of the HSC backport issues for multiband deblending, these changes will require significant modification the LSST side, because we need to apply them to the new forced measurement framework in meas_base rather than the old, HSC-only one in meas_algorithms and pipe_tasks.
Jim, would you mind giving this a review. You should see all the u/lauren/
DM-1954 branches on the tab to the right.
A Jenkins build was successful.
Review complete! I really only had minor suggestions, most of which were doc fixes. All comments in the PRs for meas_base and obs_subaru (other packages looked fine).
meas_extensions_photometryKron's tests/Kron.py still needs to be updated (it was probably missed because that package is not yet in the CI system). I have pushed a partial fix to tickets/
DM-1954 but have not yet figured out the whole story. Any help or hints would be appreciated.
A quick update on this issue. The work is essentially done and the forced coadd photometry with HeavyFootprint attachment is now working. This is based on a comparison of measurements from the reference catalogs (created on
DM-1945) to the HSC-I forced_src
catalogue having selected on merge_measurement_i = True. Most of the flux comparisons are identical (I looked at base_PsfFlux_flux, base_SdssShape_flux, and base_CircularApertureFlux_5). I'd attach plots, but they are truly flatlines, so not very interesting. One exception is for ext_photometryKron_KronFlux_flux where there are a few outliers. The flux differences are very small (the biggest was ~0.08 mag, with most <0.01 mag). Jim Bosch kindly confirmed that similar outliers also show up with the HSC stack, so they are not a result of something going awry with this port. Buildbot just let me know I have a few tests to update in other packages, so I will wait until that is done before putting this in for review (likely to Perry Gee) early next week.