Use SpectalFitting.jl as the framework

[barrettp]

Have you considered using SpectralFitting.jl as the framework for this project? The long term goal of SpectalFitting is to be a general purpose, multiwavelength spectral fitting package.

[icweaver]

Hi all, happy holidays!

Sure, were you just thinking of punting the fitting capabilities here over to SpectralFitting.jl. or was there something else that folks had in mind?

At least currently, my understanding is that we were investigating whether Spectra.jl could be a lightweight base package to provide some base types for PhotometricFilters.jl.

Do we see this as also potentially being the case for SpectralFitting.jl, or were there some examples in mind for the opposite? Just trying to get a sense of where folks are at, thanks!

[barrettp]

The reason for starting the Steering Council meetings is to discuss such issues, coordinate efforts amongst various similar packages to reduce duplicated efforts, and to provide the community with a common and coherent set of astronomical packages.

However, in answer to your question, yes I believe that the Spectra.jl fitting capabilities should be based on the SpectralFitting.jl ones, i.e., SpectralFitting.jl should be used as the framework. A requirement of SpectralFitting.jl is to be able to fit all types of data such as binned array data from CCD and CMOS detectors to event data from X-ray detectors and to use the proper statistics for each case including proper modeling of the background.

[icweaver]

Sure, looking forward to discussing more at the meeting

[cgarling]

I haven't done much work on this package (just trying to help out Ian a bit) but this package seems best positioned to be a lightweight interface-style package for doing simple, quick-look style analyses. In particular integrating with DustExtinction.jl and PhotometricFilters.jl would be nice. My interest is actually more theoretical (i.e., taking model stellar spectra from MARCS, PHOENIX, etc., applying dust extinction, and integrating them through observational filters to calculate bolometric corrections), and dealing with theoretical spectra does not require the complications that SpectralFitting.jl includes to deal with observed spectra.

Looking at SpectralFitting.jl it is clear it aims to be more detailed and robust in its implementations (e.g., compare their spectrum type to ours). As such it doesn't really make sense to me to base a simple package (Spectra.jl) on a more complicated package (SpectralFitting.jl), it seems to me they have pretty different focuses.

IMO the continuum fitting, line flux, and equivalent width implementations currently here are not really competitive with what SpectralFitting.jl or something like specutils is doing.

[icweaver]

Yea, just to expand on my previous comment a bit:

At least currently, my understanding is that we were investigating whether Spectra.jl could be a lightweight base package

I just see this package as a convenience wrapper for carrying around bits of wavelength, flux, and potentially some metadata information. Not much more. Along that line, I'd be happy to move out the analysis-specific stuff that Chris mentioned

[icweaver]

I think we've got a good handle on this now. Closing in favor of #41 and incoming PRs from Fergus. Related discussion in #24