Fix Version/s: None
Sprint:DRP S19-4, DRP S19-5
Team:Data Release Production
It appears that unless config.forceKnownWcs=True for the AstrometryTask (called in processCcd), then the referenceSelection task is not actually run.
These selectors are called only in loadAndMatch (https://github.com/lsst/meas_astrom/blob/b71a5648e2899270502e19d7bb47e8521766682d/python/lsst/meas/astrom/ref_match.py#L91) and not in solve (https://github.com/lsst/meas_astrom/blob/b71a5648e2899270502e19d7bb47e8521766682d/python/lsst/meas/astrom/astrometry.py#L145), and loadAndMatch is only called if forceKnownWcs is True: https://github.com/lsst/meas_astrom/blob/b71a5648e2899270502e19d7bb47e8521766682d/python/lsst/meas/astrom/astrometry.py#L137
Further investigation has led to the realization that there are two source selectors that may be called in certain circumstances due to duplication of configs.
This investigation led to
RFC-589, and this is now the implementation ticket for that RFC:
- The sourceSelector be removed from the matcher tasks (namely, MatchOptimisticBTask and MatchPessimisticBTask)
- The sourceSelection in AstrometryTask be changed to a SourceSelectorRegistry.
- The source selection and reference selection will be called by AstrometryTask prior to the match-and-fit iterations. This will also make things (marginally) faster because in the current usage the (identical) source selection is repeated for each iteration.