Autodiff Support and Allocation Cleanup

[jmurphy6895]

Added fixes and tests for a set of autodiff backends and cleaned up some allocations

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 100.00%. Comparing base (1ad140f) to head (f7fa5e6).
⚠️ Report is 1 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff            @@
##              main       #12   +/-   ##
=========================================
  Coverage   100.00%   100.00%           
=========================================
  Files           30        30           
  Lines         1313      1313           
=========================================
  Hits          1313      1313           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[ronisbr]

Hi @jmurphy6895 !

Awesome PR! Thanks!

One question: can we make the functions verbose by default? I am only trying to avoid changing the default behavior for those who do not want differentiation.

[ronisbr]

By the way, any idea why the tests are failing?

[jmurphy6895]

I'll go through and resolve the failing tests, my guess is it has to do with new versions that got freed by removing them from the compat I'll try resolve by end of week

[jmurphy6895]

One question: can we make the functions verbose by default? I am only trying to avoid changing the default behavior for those who do not want differentiation.

I actually added this change to make the function non-allocating. The kwargs there weren't exposed in any of the other functions so I made it false by default, with the thought that anyone modifying that function would have access to it.

I can try bubbling it up to more common interfaces via a kwarg if that's desired, when I went and tried it the first time it just ended being in a lot of places. What are your thoughts?

Edit: What if it was true by default in the function call itself but in the interfaces it was defaulted to false? I don't think a user can trigger it without working directly with the function

[ronisbr]

Hum, if it removes allocations, maybe it is better to let it false by default. Since it is just a waring, maybe I can consider it non-breaking.

[jmurphy6895]

The tests are failing due to changes to Zygote 0.6 -> 0.7, I address them here JuliaSpace/ReferenceFrameRotations.jl#35. I can add in the gating for certain tests on nightly similar to JuliaSpace/SatelliteToolboxGravityModels.jl#4, if the work-around there is acceptable

[ronisbr]

Yes, the work around seems perfect!

[ronisbr]

Thanks!