float issues with wavelength axis in synthesize

[andrew-saydjari]

Here is a reproducer

ex_atm = Korg.interpolate_marcs(5750, 2.0, 1.0)
ex_spec = synthesize(ex_atm, lines, format_A_X(1.0), 3805.0, 5645.0)
extrema(ex_spec.wavelengths)

The result is :
(3805.0000000000005, 5645.0)

[andrew-saydjari]

This persists even when passing Ints, which is very concerning.

[andrew-saydjari]

ex_spec = synthesize(ex_atm, lines, format_A_X(1.0), 3805, 5645, 1)
extrema(ex_spec.wavelengths)

Also fails

[ajwheeler]

I will keep this in mind as something to improve on, but tbh it doesn't seem critical. It's already the case that to analyze real data you must buffer the wl range on either side in order to get the LSF right.

Open to being convinced that I should care more, though.

[andrew-saydjari]

I just hate Float issues like this and try to avoid them because I find they cause no end of annoying edge cases that bother me later down the road (and can be hard to run down). It is also non-Julian to not respect my type choice if I pass Ints.

Off my philosophical soapbox, I agree with your assessment of where your time investment is valuable and think you should back burner it, but keep the issue open.

[ajwheeler]

Definitely not claiming it's Julian. I would like to fix this at some point. Just being honest that it's not a priority. I'll definitely leave the issue open.