Jim Bosch added a comment - 28/Jun/18 12:52 PM - edited Here's Russell Owen 's initial proposal for how to do both of these operations together; we now think it's missing a way to provide CRVAL to the iwcToSky transformation, but it may otherwise be complete: import math   import astshim as ast   from lsst.afw.geom import SkyWcs from lsst.afw.cameraGeom import PIXELS, FIELD_ANGLE     crval = ... camera = ... detector = ... pixelsToIwcRad = camera.getTransform(detector.makeCameraSys(PIXELS), FIELD_ANGLE) radToDeg = ast.ZoomMap( 2 , 180 /math.pi) pixelsToIwcDeg = pixelsToIwcRad.then(radToDeg) skyToIwc = ast.WcsMap( 2 , ast.AST__TAN, 1 , 2 , “PV1_1=..., PV1_2=..., ...“) pixelFrame = ast.Frame( 2 , “Domain=PIXELS”) iwcFrame = ast.Frame( 2 , “Domain=IWC”) skyFrame = ast.SkyFrame(“Domain=SKY”) skyFrame.setSkyRef(crval) skyFrame.setSkyRefIs(“Ignored”) frameDict = ast.FrameSet(pixelFrame, pixelsToIwcDeg, iwcFrame) frameDict.addFrame(“IWC”, skyToIwc.getInverse(), skyFrame) wcs = SkyWcs(frameDict)  

Jim Bosch added a comment - 28/Jun/18 1:00 PM Also from slack ( John Parejko ), this seems like an even better starting point: https://github.com/lsst/jointcal/blob/master/src/SimpleAstrometryModel.cc#L117 It's not the most elegant solution (going through FITS metadata to construct the iwcToSky transform), but I bet it would work, and it may even be better long term than trying to work out which AST mappings we'd want to construct directly to make a composed mapping that is equivalent to a FITS TAN WCS.

John Parejko added a comment - 28/Jun/18 1:26 PM Where does the jointcal code "go through FITS metadata"?

Jim Bosch added a comment - 28/Jun/18 1:29 PM Inside the makeSkyWcs call, so it's not actually jointcal's problem, I suppose.

Robert Lupton added a comment - 29/Jun/18 1:34 PM  I believe that this function works (it might be easier once e.g. makeIntermediateWorldCoordsToSky is available). The API is higher than the examples given above, but I think that it needs to be supported as part of this ticket (the details of the API could change, of course, but I deliberately chose to pass a Point2D and Angle and would be a little unhappy if this were changed as that seems to me to be the natural inputs for the user ). When DM-14980 is resolved the API will not need to include the camera.   import astshim import lsst.afw.cameraGeom as cameraGeom import numpy as np # only for pi!   def getWcsFromDetector(camera, detector, boresight, rotation = 0 * afwGeom.degrees): """Given a camera, detector and (boresight, rotation), return that detector's WCS   Parameters - - - - - - - - - - camera: `lsst.afw.cameraGeom.Camera` The camera containing the detector detector: `lsst.afw.cameraGeom.Detector` A detector in a camera boresight: `lsst.afw.geom.Point2D` The boresight of the observation rotation: `lsst.afw.geom.Angle` The rotation angle of the camera   The rotation is "rotskypos" , the angle of sky relative to camera coordinates ( from North over East) """ trans = camera.getTransform(detector.makeCameraSys(cameraGeom.PIXELS), detector.makeCameraSys(cameraGeom.FIELD_ANGLE)) polyMap = trans.getMapping() radToDeg = astshim.ZoomMap( 2 , 180 / np.pi) # convert from radians to degrees polyMap = polyMap.then(radToDeg)   pixelFrame = astshim.Frame( 2 , "Domain=PIXELS" ) iwcFrame = astshim.Frame( 2 , "Domain=IWC" )   frameDict = astshim.FrameDict(pixelFrame, polyMap, iwcFrame)   crpix = afwGeom.PointD( 0 , 0 ) crval = afwGeom.SpherePoint( * boresight, afwGeom.degrees) cd = afwGeom.makeCdMatrix( 1.0 * afwGeom.degrees, - (rotation + 90 * afwGeom.degrees), True ) iwcToSkyWcs = afwGeom.makeSkyWcs(crpix, crval, cd)   iwcToSkyMap = iwcToSkyWcs.getFrameDict().getMapping( "PIXELS" , "SKY" ) skyFrame = iwcToSkyWcs.getFrameDict().getFrame( "SKY" )   frameDict.addFrame( "IWC" , iwcToSkyMap, skyFrame)   wcs = afwGeom.SkyWcs(frameDict) return wcs  

Russell Owen added a comment - 10/Jul/18 1:27 PM - edited I added one new function: makeSkyWcs(pixelsToFieldAngle, orientation, flipX, crval, projection="TAN") rather than the two requested functions. I did this because it is not enough to have the IWC to sky transform, one must also construct an appropriate ast::SkyFrame with the SkyRef and SkyRefIs properties set, and I wanted to avoid users having to deal with that or with FrameDicts . Note that "orientation" is with respect to the focal plane, not CCD x, y. This can be helpful for cameras such as HSC that have a mix of CCD rotations, because the same angle can be used for all CCDs. However, it does require some care on the part of the user.

Russell Owen added a comment - 10/Jul/18 1:30 PM For the record: the FITS intermediate world coordinates frame (IWC) is the same as field angle, except as follows: IWC is in degrees, whereas field angle is in radians flipX affects IWC, but not field angle orientation affects IWC, but not field angle

Jim Bosch added a comment - 10/Jul/18 3:14 PM Looks good.  Some comments about tests and names on the PR. I'll also note that I do think that the use of SpherePoint for the boresight position here is preferable to the Point2d in  Robert Lupton 's proposed signature above; we should use Angle for angles, and SpherePoint is the way to do that for coordinate tuples. But the functionality to go straight from a Detector to a SkyWcs in Robert Lupton 's proposed signature definitely is desirable, even if it's just a trivial forward to the function implemented here that operates on the Transform .  Given that that signature would be changing soon anyway on DM-14980 , I'm happy to just add it when I take that ticket on, but I also wouldn't be opposed to Russell Owen adding it here.  I'll also admit that I haven't thought yet about whether it'd be problematic to have that function here from a circular dependency standpoint (I don't know which way the afw.geom<->afw.cameraGeom dependency runs now that we've moved some stuff into lsst.geom).  Regardless, we'll want that signature somewhere.

Russell Owen added a comment - 11/Jul/18 4:11 PM Thank you for the helpful review Jim Bosch