Allow other ways to store wavelength Solutions

[simontorres]

This topic was brought up at the UNC meeting. The argument is that, since in order to write a linear solution in the header we are linearizing the data therefore we are re-sampling it. The request was to allow the data to be written as a table of wavelength and intensity (maybe multi-extension fits).

I need to look this up in more details. I would discard writting non-linear solutions in the header since the documentation is poor and is really complicated.

[cbaorion]

I agree.

[simontorres]

I'm still not in favor of writing fits files as tables because the model information is lost.

• To recover it would be necessary to do a fit, again information is lost and errors not tracked.
• Some people might not need to recover the model information.

As an alternative solution for this is to use adsf+gwcs, I'm still not fully confidant with these tools but in general they allow to serialize arrays and models.

For example, to write:

from asdf import AsdfFile

f = AsdfFile()

print(f.tree)
# will print {}

# adding attributes, poly is an astropy.modeling.Model instance, wavelength
# and intensity are arrays.
f.tree['model'] = poly
f.tree['wavelength'] = wavelength
f.tree['flux'] = intensity

# finally write to a file
f.write_to('testfile.asdf')

To read:

ff = asdf.open('testfile.asdf')
# how access attributes or fields in the tree.
ff.tree['model'](range(len(intensity)))

[simontorres]

We need to revisit this because of #384

[simontorres]

This is kind of redundant, please focus attention in #384