Roadmap

[mileslucas]

Below is a broad-strokes road map of the design of Spectra.jl

Types

General spectrum types

• AbstractSpectrum
• Spectrum <: AbstractSpectrum
  This would be a type very similar to python's SpecUtils.spectrum1d. It contains a spectral axis (wavelength) and flux. This flux could be signified with a standard array or a Measurements.jl array signifying measurement uncertainty. Note that this is a design choice over having an explicit sigma array within the Spectrum type.
• CountSpectrum <: Spectrum
  Similar to the above but carries an extra array around with the counts such that proper addition between spectra could happen.
• EchelleSpectrum <: AbstractSpectrum
  Not sure the best way to implement this. It seems like a natural extension of Spectrum but into two dimensions. Is it appropriate to keep this a separate type, or should Spectrum have an intrinsic dimension and an echelle spectrum would just be a Spectrum{..., 2}

These types should be easy to create and have decent interoperability with FITSIO.jl and WCS.jl.

Operations

I would definitely enjoy some input on these, I have limited experience

• Extinction
  Use DustExtinction.jl
• Broadening
  • Instrumental
  • Rotational
    In general, I'm not certain whether it's appropriate to put these explicitly here or put the underlying broadening mechanics in AstroBase and provide the explicit instrumental or rotational broadening ops here (cc @giordano).
• Doppler Shift
• Resampling
  Currently using Interpolations.jl but running into some problems. May switch to Diereckx.jl. It will be nice to have a function that can resample to a minimum doppler content. The way I've implemented this before is to determine the minimal $\Delta \nu$ and create a subsequent log-lambda spaced grid.
• Rescale
  A few ideas- rescale to a given integrated flux (as a value) or to another spectrum. Can optionally use a bandpass to only scale in a certain area (see photometry below)
• Photometry
  This would be similar to what PySynphot offers, where a user specifies a bandpass (either a standard like V or HST-OIII, or a Distributions-like object with a pdf) and we can get the respective integrated flux density.

Spectral Library Interface

I'm not certain if this is appropriate here or in a separate package, but it would be nice to have standardized interfaces to common model libraries (like PHOENIX). My gut says this should be a different package, but I'm not convinced yet. For an example of such an interface take a look at the work I've done here.

[icweaver]

Howdy from the future. Just wanted to check in with folks to see if this package is still needed today? It looks like the RvSpectML organization and Korg.jl, among others, have this space pretty well covered now, but it would be useful to have someone with domain experience weigh in.

Will remove this special casing from our JuliaAstro infra testing untill we hear back otherwise

[cgarling]

For my part I would not take any action at the moment. Weve talked about using this package for PhotometricFilters.jl and are still considering it. I also think its good to have some simple stuff here (e.g., the simple line fitting and reddening utilities which I don't think Korg or RVSpectML have in their public APIs? Korg uses a different fitting method)

For this package to be registered it will need a name change (Spectra.jl is taken) and probably some other work. But it does not seem to be redundant and so I would argue for keeping it.

[icweaver]

Sweet, this is exactly the feedback we need, thanks Chris! Just saw JuliaAstro/PhotometricFilters.jl#32

Having a concrete use case like this motivates me to try and get this package over the finish line now

[icweaver]

EchelleSpectrum <: AbstractSpectrum
Not sure the best way to implement this. It seems like a natural extension of Spectrum but into two dimensions. Is it appropriate to keep this a separate type, or should Spectrum have an intrinsic dimension and an echelle spectrum would just be a Spectrum{..., 2}

Note to self: Looks like specutils is going with the combined route now. Could be nice to follow that