Fix Version/s: None
Sprint:AP S22-1 (December)
Following the success of
DM-32245, it is pretty apparent that all the various flavors of ApPipe.yaml should use multi-tract difference imaging by default. At present, ApPipe.yaml is single-tract and ApPipeMultiTractFakes.yaml is multi-tract (with fakes, as the name suggests).
This ticket is to standardize the situation further:
- ApPipe should use multi-tract difference imaging
- Rename ApPipeMultiTractFakes (and its camera-specific derivatives) to just ApPipeFakes
- Remove lsst-imSim's ApPipeFakes because it's already all fake
- Use yaml parameters throughout for setting e.g. fakesType and coaddName
- Ensure all pipeline imports (especially in camera-specific pipelines) are as straightforward as possible
- Run each camera-specific pipeline to make sure nothing broke
This work is being completed on PREOPS-931.
LOL no it's not, we're going to move it here, stay tuned.
- Ken updated ap_pipe/pipelines/ApPipe.yaml to use ImageDifferenceFromTemplateTask (multi-tract differencing) plus the necessary GetMultiTractCoaddTemplateTask prerequisite, in lieu of single-tract ImageDifferenceTask
- Meredith updated the ap_pipe fakes-related pipelines to be consistent with this, as the ticket description suggests
- Krzysztof helped us realize the three ap_verify datasets needed to have updated pipelines, since they all have their own custom pipelines that import from ap_pipe/pipelines, so now there are three more PRs
- TO DO (Ken): test runs of (1) ap_verify.py on both CI datasets and (2) ApPipe.yaml on a small HSC dataset to make sure nothing horrible happened
- Assuming nothing is broken, the work to tidy up various pipelines in ap_verify to use the new ap_pipe fakes pipelines and tidy up the overall ApVerifyWithFakes situation (and/or the messiness of per-dataset pipelines in general!) is punted to
Let me know if I've missed anything or misrepresented anything!
Links to all PRs, because Jira is at least as confused as I am
ap_pipe and the three ap_verify_<dataset> packages have all been merged now.
I'm listing this as a blocker for DM-28628, which will probably involve putting both new calibs and new templates into the dataset. The templates generated by
DM-29857use the finer-grained hsc_rings_v1 skymap, and therefore cannot be processed without multi-tract support.
(We can update just the calibs, if need be, but I think having everything up-to-date would be better.)