When working on #1289, I realized that the ChromaticObject.atRedshift feature doesn't always make much sense.
If the ChormaticObject is fairly simple, like (gal * sed).atRedshift(z) then atRedshift just applies the redshift to the SED and everything is fine. But in that case, it's just as easy for the user to call atRedshift on the SED directly: gal * sed.atRedshift(z).
However, if the object is a convolution of a chromatic galaxy with a chromatic PSF, then it's unclear what atRedshift should even mean? Probably you just want to shift the galaxy part, but the current implementation also changes the wavelength dependence of the PSF part, which seems likely an error. It also doesn't really work properly with photon shooting.
Rather than try to find an implementation that does this sensibly, it seems more straightforward to just require the user to apply the redshift to the galaxy SED instead. So this PR deprecates the use of atRedshift on Chromatic objects as well as putting a redshift parameter in the gal field of a config dict, unless the galaxy is achromatic (in which case redshift is just an attribute, so it's not problematic).
When working on #1289, I realized that the ChromaticObject.atRedshift feature doesn't always make much sense.
If the ChormaticObject is fairly simple, like
(gal * sed).atRedshift(z)thenatRedshiftjust applies the redshift to the SED and everything is fine. But in that case, it's just as easy for the user to callatRedshifton the SED directly:gal * sed.atRedshift(z).However, if the object is a convolution of a chromatic galaxy with a chromatic PSF, then it's unclear what
atRedshiftshould even mean? Probably you just want to shift the galaxy part, but the current implementation also changes the wavelength dependence of the PSF part, which seems likely an error. It also doesn't really work properly with photon shooting.Rather than try to find an implementation that does this sensibly, it seems more straightforward to just require the user to apply the redshift to the galaxy SED instead. So this PR deprecates the use of
atRedshifton Chromatic objects as well as putting aredshiftparameter in thegalfield of a config dict, unless the galaxy is achromatic (in which case redshift is just an attribute, so it's not problematic).