Details
-
Type:
Bug
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: ap_association
-
Labels:None
-
Story Points:4
-
Epic Link:
-
Sprint:AP S21-4 (March), AP S21-5 (April)
-
Team:Alert Production
-
Urgent?:No
Description
A recent run of the ap_verify_hits2015 dataset is throwing errors with points being outside of the ccd bounding box. Likely this is coming from DiaForcedSource photometry measurements where the previous behavior was to run a NaN value instead of crashing.
Ticket is to investigate (confirm where the error is from) and fix the problem.
[Updated title. See below]
Attachments
Issue Links
Activity
Looks good, please see my comments on the PR. My main question is whether we should catch all InvalidParameterErrors, or only the ones that were caused by this particular bug. I assume we would want to catch them all, but we might want to hit them for now so that we can try to understand where they are coming from.
Initially the thought was that this problem was occurring in DiaForcedSource based on the output error. However, this error appears to be thrown by PackageAlertsTask when center of the BBox is off image. This is caused (strangely) by the DiaSource centroid itself being off image in the diffIm. I'll add some sort of try, except clause to catch this for now (likely return None for the image cuttout), however it might be worth looking into why this centroid is off image. This may be related to previous issues (DM-26032) discovered in processing where negative detections some times contained peaks that were 100s or 1000s of pixels outside of the source footprint.
For now I'll test the None return and add a warning if it is used. I'll also check a few other failures to see if it is caused by similar problems with the DiaSource centroid.