In ticket DM-4410 I am porting over some code which will cause the detection task to turn on an option such that detection will now return footprints which are grown by some factor of the psf. This code has existed for some time but was off by default because there was no deblender to separate sources which were merged, and was intended to be turned on once a deblender was present. We want this option to be on by default as without it some footprints of detected objects may be returned as unrealistically small, causing some plugins to fail. We now have the deblender and want to turn this grow option on by default, like in HSC.
However turning on grow footprints caused the unit test for processCcd to fail. In the case of no grown footprints the processCcd unit test detected 169 sourced. When grow footprints was turned on, only 159 sources were detected. This makes sense as some close objects had their footprints merged. It became apparent however that the deblender was not running. When modifying the unit test to use the deblender, 188 sources were detected, which again makes sense as the basic detection task would not know how to separate merged sources and with the addition of deblending we should expect additional sources.
This all leads to the question, if we are turning on an option to grow all footprints (to solve unrealistically small footprints) should we also enable deblending by default? If we do, the processccd unit test will be updated to reflect this change. However if we do not turn on deblending by default, perhaps we should not grow footprints as there may be unexpected behavior for people running the stack. I am voting for turning on deblending by default with this ticket as they are logically connected. Does anyone have any feeback on this?