DM-20746: Defer to camera geometry instead of raw formatter for intiial WCS.

[TallJimbo]

This is a small cleanup (reducing reliance on a middleware interface we'd prefer to stop treating as public) that should only be merged after the 'camera' datasets in all LATISS repos have been rebuilt with post-DM-20746 code.

[cmsaunders]

Looks good.

[cmsaunders]

Rereading the comments on the Jira ticket, I see that this is not supposed to be merged with the other PRs.

[kfindeisen]

Is this still relevant?

(Also, typo in the commit message)

[TallJimbo]

Yes, this is still relevant, but I haven't been paying attention to whether the precondition has been met (though if prompt processing no longer cares about LATISS, that's another way to satisfy it).

I've rebased and fixed the commit message.

[kfindeisen]

We are still using LATISS data for integration testing, but I have no idea how to recognize a "rebuilt camera".

Will it fail spectacularly if we try to run it on an incompatible repo?

[TallJimbo]

Yes, I expect you'll see astrometry-solving failures a large fraction of the time if the camera hasn't been rebuilt.

If the camera has been rebuilt, you should see:

butler.get("camera").getFocalPlaneParity()

return True instead of False. The reason this is only relevant for LATISS (and DECam) is that our other cameras will always return False.

[kfindeisen]

Basic tests (upload.py) work fine. A larger run of 38 LATISS visits produced 3 instances of NoPsfStarsToStarsMatchError, 2 of ObjectSizeNoGoodSourcesError, and 1 of BadSubtractionError.

A run of 38 (independently selected) LATISS visits on latest produced 5 instances of NoPsfStarsToStarsMatchError, 2 of ObjectSizeNoGoodSourcesError, 2 of BadSubtractionError, and 1 of PsfexTooFewGoodStarsError. So I think we're good?

[TallJimbo]

That seems like a fairly high failure rate, but I don't have a good baseline for LATISS. Is this /repo/main? I can just go inspect whatever repo is relevant manually, and then we can be sure.

[TallJimbo]

It looks like the camera in /repo/main has not been rebuilt since this change. I'll go see if I can figure out who is maintaining calibrations for LATISS and ask if we can just do that.

[kfindeisen]

Re: the failure rate, the imaging pipeline has never been optimized for LATISS. That's why I did the latest run (i.e., without this PR) for comparison.

We have our own dev repo, but I don't know when or from where the camera was last sourced. @hsinfang might know.