Details
-
Type:
Story
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: obs_base
-
Labels:None
-
Story Points:1
-
Epic Link:
-
Sprint:AP S19-4
-
Team:Alert Production
Description
It looks like <exposure>_wcs datasets are coming from the header (which may be an approximation) instead of the true WCS in the FITS binary tables. Fix this, and add support for more components if it's easy to do so.
Note that using the header was once thought to be necessary to do this efficiently, but in fact there's always been a trick to get anything out of the binary tables efficiently (due originally to Paul Price, I think): load a 1-pixel subimage of the exposure, and pull the component out of that.
However, since DM-15500 we've had an ExposureFitsReader class that provides direct and efficient access to all of those components without that trick, and that's what we should use here.
Attachments
Issue Links
- relates to
-
DM-16386 Recent data using obs_decam retrieves incorrect wcs with butler.get("calexp_wcs")
- Done
Activity
Nevermind: my vistInfo implementation also results in an error in obs_test (see latest Jenkins) during validation:
E TypeError: getVisitInfo() missing 2 required positional arguments: 'location' and 'dataId'
|
Unless someone has a fix for this, I would rather revert and leave implementing the non-Wcs things to another ticket.
Tim found a fix, which fixes my local obs_test. New Jenkins: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/29526/pipeline
Looks fine to me. It might be possible to factor out the commonality between these functions, but only at the expense of going back to a closure/lambda, I think, which would reduce debuggability.
Thank you for the review. I agree that it would be nicer to have a common way to handle this, but I don't see an obvious approach.
Merged and done.
Ok, I updated it for Filter and VisitInfo. Calib is going away shortly (and was only ever persisted in the header), and Detector and BBox are more complicated, and can't be done in the trivial manner I used here.
New Jenkins run: https://ci.lsst.codes/blue/organizations/jenkins/stack-os-matrix/detail/stack-os-matrix/29525/pipeline