Feature request: redshift support for SpectrumCollection

[weaverba137]

To help plan future development, I would like to request support for a redshift property in SpectrumCollection, functionally similar to Spectrum.redshift.

Note that #1318 is related.

[dhomeier]

In principle the property already exists, as both redshift and radial_velocity are SpectralCoord attributes with Spectrum offering direct references to its spectral_axis properties, and of course convenience functions like shift_spectrum_to.
So there is not much standing in the way of offering the same in SpectrumCollection, maybe with the small caveat that many instances may indeed represent spectra with all identical shifts, but in principle each source and spectral axis could of course also have its individual shift, e.g. extracts from an IFU.
This could even be handled by SpectrumCollection by giving it not a single redshift, but an array of the same shape – but not (easily) within the existing representation, which uses a single SpectralCoord as spectral_axis and only allows single (scalar) parameters for that (see checks in from_spectra).
I found that SpectralCoord actually accepts arbitrary array-like inputs for redshift and radial_velocity, but it's unclear to me how those are supposed to function. Anyway this is within the astropy domain...
Unfortunately this also does not offer an easy path to moving methods like shift_spectrum_to to spectral_axis to avoid code duplication, but it could still be added here at least for scalar shifts.

[rosteen]

I just opened #1332 to add support for scalar redshift/radial_velocity in SpectrumCollection. As @dhomeier noted, although SpectralCoord currently allows defining these properties as arrays, the array case errors out when using the available methods to shift to new values and would require some upstream work to handle that case properly.

[weaverba137]

I'm confused, the reason I opened this issue in the first place was to add support for a redshift property that has a different redshift for each spectrum. But you seem to be only supporting scalar redshift in the PR. Could you elaborate on this?

[rosteen]

I'm confused, the reason I opened this issue in the first place was to add support for a redshift property that has a different redshift for each spectrum. But you seem to be only supporting scalar redshift in the PR. Could you elaborate on this?

That's why I'm not closing this issue with #1332. As I said above, supporting array redshift needs some upstream work to be done first, so for now I'm implementing what I can here and will build on it later.

[weaverba137]

O sorry! #1332 closes #1318, not this one. Thank you.