Rebuilding the IRAF identify fitcoords stack

One of the most essential tasks in spectroscopic data reduction is extraction of a 1D spectrum from a 2D image. This is the basic first step of nearly all optical, UV, and IR spectroscopic data reduction packages.

For long-slit spectra, a very nice way to do the 1D extraction is to create a complete 2D mapping of pixel coordinates in a CCD to spectral + spatial coordinates. This is at least partly going to be dealt with in the "generalized WCS" system, but it is not yet clear how that system is going to be created. Open questions:

• How do you measure the mapping from pixel to spatio-spectral coordinates?
  • IRAF used identify for the spectral axis.
  • apextract required a trace of a stellar spectrum to determine the spectral curvature
• Within identify, tools are needed to match peaks to values within a spectral line catalog.
  • Local peak fitting is a first step, which can be done with astropy.modeling
  • The standard IRAF system of interactively identifying lines works, but is awful. Can it be systematically replaced?
  • reidentify using a set of input guesses is the most effective way to do this for stable spectrographs
• Some version of fitcoords is needed to determine the best-fit polynomial distortion map
  • fitcoords had a fairly useful graphical interface, but many improvements in data selection techniques should be made
• Reprojection to a linear coordinate system is usually desirable. This can be achieved using scipy.ndimage.map_coordinates once the coordinate map has been determined.