You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The goal of this issue would be for synphot itself to not have to download any data - i.e. all data part of the package - while extension packages like stsynphot would still be able to use the existing data-download machinery. The idea being that in theory synphot should have pretty minimal data requirements, but stsynphot for example by design requires quite a lot of data that shouldn't be included in the package.
Currently the data files that require downloading (which can be most easily seen in the current synphot.cfg) fill into three categories: 1) filter bandpasses 2) extinction laws 3) a Vega spectrum for vega mags. These could be dealt with via three different strategies:
The filter bandpasses might be "compressed" by using tynt (https://github.com/bmorris3/tynt), or a similar approach. That ends up representing the bandpasses via Fourier coefficients and thereby requiring a lot less space. tynt seems to manage a reasonable reconstruction for ~500 filters with 150kb, so the 12 built into synphot could be reproduced with just a few KB, easy to include in the package itself. (EDIT: Also see Parameterize non-HST filter curves #38)
Extinction laws could be farmed out to an external package - see Use separate package for extinction curve #132 for more. This is probably better to do anyway since it then becomes more flexible and adaptable to the future and other extinction models appear.
The Vega spectrum is 270K (or 150K gzipped). Since it's very rarely likely to be updated, that seems reasonable to keep as an shipped-with-the-package data file (maybe ungzip it at install or import time or something). Then for users who really care a lot about their choice of Vega, Provide a way to give a favored Vega spectrum #209 can give them a way to use a different one.
(Prompted by a discussion @pllim and I had offline, but I think I've faithfully recorded what we discussed.)
The text was updated successfully, but these errors were encountered:
The goal of this issue would be for
synphot
itself to not have to download any data - i.e. all data part of the package - while extension packages likestsynphot
would still be able to use the existing data-download machinery. The idea being that in theorysynphot
should have pretty minimal data requirements, butstsynphot
for example by design requires quite a lot of data that shouldn't be included in the package.Currently the data files that require downloading (which can be most easily seen in the current synphot.cfg) fill into three categories: 1) filter bandpasses 2) extinction laws 3) a Vega spectrum for vega mags. These could be dealt with via three different strategies:
The filter bandpasses might be "compressed" by using
tynt
(https://github.com/bmorris3/tynt), or a similar approach. That ends up representing the bandpasses via Fourier coefficients and thereby requiring a lot less space.tynt
seems to manage a reasonable reconstruction for ~500 filters with 150kb, so the 12 built intosynphot
could be reproduced with just a few KB, easy to include in the package itself. (EDIT: Also see Parameterize non-HST filter curves #38)Extinction laws could be farmed out to an external package - see Use separate package for extinction curve #132 for more. This is probably better to do anyway since it then becomes more flexible and adaptable to the future and other extinction models appear.
The Vega spectrum is 270K (or 150K gzipped). Since it's very rarely likely to be updated, that seems reasonable to keep as an shipped-with-the-package data file (maybe ungzip it at install or import time or something). Then for users who really care a lot about their choice of Vega, Provide a way to give a favored Vega spectrum #209 can give them a way to use a different one.
(Prompted by a discussion @pllim and I had offline, but I think I've faithfully recorded what we discussed.)
The text was updated successfully, but these errors were encountered: