Details
-
Type:
Bug
-
Status: Done
-
Resolution: Done
-
Fix Version/s: None
-
Component/s: pipe_tasks
-
Labels:
-
Story Points:1
-
Epic Link:
-
Team:Alert Production
Description
The option doNImage is very handy for coadd assembly, as it creates an "nImage" of each patch with pixel values that show how many images (i.e., warps) will be assembled at that location. The nImages are saved as fits files. However, things get a little complicated with dcrAssembleCoadd's nImages.
This ticket is to change the present situation so only one nImage is created one time. The current state of affairs with doNImage (approximately in order):
- dcrAssembleCoadd calls regular assembleCoadd, which creates an nImage for all patches in the deepCoadd directory before beginning assembly
- These nImages are overwritten with nImages that have pixel values 3x larger than before (in this example, 3 is the number of DCR subfilters being used)
- Another set of nImages is created in the dcrCoadd directory alongside the dcr coadds, which are identical to the original nImages
Somewhat related to this, when one uses the butler to load a DCR coadd and does dcrCoadd.getInfo().getCoaddInputs().visits, it lists each constituent visit 3 times (i.e., nSubfilter times). This is probably also a bug.
Attachments
Activity
| Field | Original Value | New Value |
|---|---|---|
| Priority | Undefined [ 10000 ] | Major [ 3 ] |
| Epic Link |
|
| Story Points | 1 |
| Reviewers | Meredith Rawls [ mrawls ] | |
| Status | To Do [ 10001 ] | In Review [ 10004 ] |
I confirmed that nImages in the deepCoadd directory are now only created once assembly has succeeded. Great, and thank you!
I will note the issue with dcrCoadd.getInfo().getCoaddInputs().visits printing each visit nSubfilter times still exists, but this is not a blocker or critical problem.
| Status | In Review [ 10004 ] | Reviewed [ 10101 ] |
| Resolution | Done [ 10000 ] | |
| Status | Reviewed [ 10101 ] | Done [ 10002 ] |
I'll create a separate ticket to fix the dcrCoadd.getInfo().getCoaddInputs().visits duplication bug.
The DCR nImages should have values equal to the number of subfilters * the number of images, which is necessary since the values must be integers.
The bug where the nImage is written once at the start of processing, and then overwritten at the end should now be fixed.