Dithering and nodding

[siriobelli]

This is a very exciting project! I have a suggestion regarding dithering (sub-pixel offsets) and nodding (large offsets to improve background subtraction). It would be nice if these modes were natively supported and did not require the data to be resampled one extra time. A clean way to do this is to include the spatial offset of each frame in the rectification solution. Ideally, both wavelength calibration and spatial rectification should be 2D mappings from observed pixel coordinates (x,y) to final rectified coordinates (wavelength, actual "vertical" position on the sky in arcsec). This way the 2D spectrum needs to be resampled only once throughout the data reduction. I wrote a detailed description of this workflow here: https://ui.adsabs.harvard.edu/abs/2018MNRAS.478.2097B -- this may be useful for inspiration, particularly for near-infrared data (but the associated code won't be as useful as it was written in IDL and not in an object-oriented way).

[tepickering]

thank you for this and for the link to the paper! we have talked about this informally in the past, but it's good to finally have an issue in place as a marker to implement this. there's a bit more infrastructure that needs to be put in place first, but this is something specreduce should definitely support.

[jehturner]

This is where we're going with the spectroscopic reduction at Gemini, using gwcs. Being able to have a single resampling step (or even none) is more-or-less the point of issue #20. Thanks for the paper (can't recall off hand whether I've seen it before).