DRAFT: feature: partial caching of chromatic Fourier design matrix

[jeremy-baier]

This PR adds the ability to cache portions of the construction of the chromatic Fourier design matrix.
Initially testing on NG15yr data shows a 10x speed up to the recalculation of the chromatic Fmat construction.
This constitutes anywhere between a 3% to 15% speed up of the overall likelihood calculation for single pulsar noise when sampling in the chromatic index, alpha.

As the function docstrings say, the actual caching occurs via the @function decorator, which now wraps the later function, caching the initially passed arguments which include the achromatic portion of the Fourier design matrix and the division of TOA observational frequencies by a reference frequency.

This will be implemented in the chromatic_noise_block() of enterprise_extensions for automatic caching when sampling in the chromatic index.

[AaronDJohnson]

@jeremy-baier linting failed. Can you run the linter? I think enterprise may be on a slightly older version of flake8/black. So just check to see what version it is using..

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 85.41%. Comparing base (efae9b0) to head (88e9bfa).
Report is 4 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #400      +/-   ##
==========================================
+ Coverage   85.33%   85.41%   +0.07%     
==========================================
  Files          13       13              
  Lines        3158     3167       +9     
==========================================
+ Hits         2695     2705      +10     
+ Misses        463      462       -1     

Files with missing lines	Coverage Δ
enterprise/signals/gp_bases.py	73.56% <100.00%> (+3.05%)	⬆️

... and 1 file with indirect coverage changes

---

Continue to review full report in Codecov by Sentry.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 447ddae...88e9bfa. Read the comment docs.

[jeremy-baier]

@AaronDJohnson looks like everything is passing now except one part of the code cov. Does that just mean I have to make a test for the 2 added functions?

[AaronDJohnson]

@jeremy-baier codecov measures the number of lines covered by tests. So more tests will increase codecov % as long as those lines that you're covering are not already tested.

[jeremy-baier]

@AaronDJohnson I added a test which covers both of the added functions. This seemed to satisfy the codecov bot.
I think this PR is ready for merge now!

[jeremy-baier]

marking this as a draft for a second because I need to check one part of the compatibility with enterprise_extensions

[vhaasteren]

@jeremy-baier , I marked it as a draft by just using the feature. Or did you by now verified this is compatible?

[jeremy-baier]

hi @vhaasteren, thanks for following up on this.
so i verified that it was not compatible with enterprise_extensions’s chromatic noise block, and was waiting to see if I could find a way to make it compatible. so far i have not (and i tried for a while). however, this code is still useful as it does provide a speed up. it’s just that one can’t use that speed up and set their model up with enterprise_extensions.
if someone else, want’s to try to make this work with enterprise_extensions, i would gladly accept some help here. otherwise, i think it is worth merging as is. (and the user just can’t use the chromatic noise block).