No more AstroPy compatible WCS from the Butler (LSSTCam onwards)

[ColinOrionChandler]

The functionality to make an AstroPy WCS is effectively gone from difference images, visit images, and others.

This means our Butler standardizer won't work for making WorkUnits for LSSTCam.

Possibly images from Prompt Processing still have the old (but inaccurate) FITS WCS that we currently depend on, but DRP will not.

There is a thread on Rubin Slack about this topic, re: the RuntimeError "WCS does not have an attached FITS approximation" seen with DP1, then again with LSSTCam. (DP1 had a WCS inserted to handle this issue, but the WCS isn't great from what I'm seeing.)

[maxwest-uw]

😬

is it possible for us to get a list of say 100 points from across the Butler (+ the four corners of the image) and then use astropy's fit_wcs_from_points function to get a good approximation? it's a similar process to what we use in the actual reprojection code to go from original->ebd space

[wilsonbb]

For a more complete discussion of the issues where this does and does not work, see https://community.lsst.org/t/how-to-use-wcss-in-dp1-and-commissioning-processing/10769

@maxwest-uw suggestion seems like to me it could work in the butler standardizer by just

1. Choose 100 random (x,y) pairs in the image, rand_xy
2. sky_coords = pixelToSky(rand_xy)
3. Call wcs = fits_wcs_from_points(rand_xy, sky_coords)

But don't know if this would be accurate enough for our purposes