Upgrade to SciMLOperators

[vpuri3]

plan: SciML/SciMLOperators.jl#142

To be clear, DiffEqOperators is not being deprecated in this PR. I am only adding support for SciMLOperators.jl by moving AbstractSciMLOperator and many of the traits to SciMLOperators.jl and making DiffEqOperators a subtype. That way, all the stuff in SciMLOperators.jl will be supported in downstream packages.

[codecov]

Codecov Report

Merging #377 (3ae3c78) into master (9ee42e0) will increase coverage by 0.70%.
The diff coverage is 68.18%.

@@            Coverage Diff             @@
##           master     #377      +/-   ##
==========================================
+ Coverage   57.61%   58.31%   +0.70%     
==========================================
  Files          44       44              
  Lines        3324     3306      -18     
==========================================
+ Hits         1915     1928      +13     
+ Misses       1409     1378      -31     

Impacted Files	Coverage Δ
src/SciMLBase.jl	77.77% <ø> (ø)
src/operators/basic_operators.jl	53.93% <ø> (ø)
src/problems/basic_problems.jl	86.95% <ø> (ø)
src/problems/ode_problems.jl	57.14% <ø> (ø)
src/operators/operators.jl	25.00% <50.00%> (+14.28%)	⬆️
src/scimlfunctions.jl	59.70% <66.66%> (+0.12%)	⬆️
src/operators/common_defaults.jl	58.53% <75.00%> (-15.00%)	⬇️
src/remake.jl	78.84% <0.00%> (+0.96%)	⬆️
... and 4 more

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[vpuri3]

@ChrisRackauckas

[vpuri3]

oops looks like some tests failed

[vpuri3]

looks like none of the errors are related to the changes made here. @xtalax can you help?

[xtalax]

@ChrisRackauckas can confirm that the flux issue is present on master, excepting invalidations LGTM

[vpuri3]

Okay, I took a look at all the errors:

In OrdinaryDiffEq.jl, the test is failing because of low accuracy. However the offending @test, however, is comparing floats with = instead of \approx, and has no tolerance value.

https://github.com/SciML/OrdinaryDiffEq.jl/blob/2426acd6342a17add6e9fee979bf7f53d2d97270/test/regression/psos_and_energy_conservation.jl#L48

I reran that entire test group on my personal mac (with this branch of SciMLBase.jl), and the test passed. Then I did the same thing on a linux cluster and it got no errors. So this must be a numerical fluke.

For DiffEqFlux, the test is erroring on the NeuralCDDE function call. I am getting that error with when I test DiffEqFlux without this PR so this doesn't seem related to my work.

https://github.com/SciML/DiffEqFlux.jl/blob/7e52962f011f1a9ebf39a05c0f48ffd3a9b659df/test/neural_de_lux.jl#L274

[vpuri3]

yay all tests have passed! @ChrisRackauckas this is good to go whenever you have time.

[vpuri3]

anything else @ChrisRackauckas ?

[vpuri3]

@ChrisRackauckas anything else that needs to be done here, or are we just waiting on tests?

[ChrisRackauckas]

Add LinearSolve.jl to the downstream tests

[vpuri3]

hold on - just noticed LinearSolve was in downstream tests all along:

SciMLBase.jl/.github/workflows/Downstream.yml

Lines 28 to 37 in f9e2da0

	- {user: SciML, repo: LinearSolve.jl, group: Core}
	- {user: SciML, repo: OrdinaryDiffEq.jl, group: Interface}
	- {user: SciML, repo: OrdinaryDiffEq.jl, group: Integrators}
	- {user: SciML, repo: OrdinaryDiffEq.jl, group: Regression}
	- {user: SciML, repo: StochasticDiffEq.jl, group: Interface1}
	- {user: SciML, repo: StochasticDiffEq.jl, group: Interface2}
	- {user: SciML, repo: StochasticDiffEq.jl, group: Interface3}
	- {user: SciML, repo: StochasticDiffEq.jl, group: AlgConvergence}
	- {user: SciML, repo: Sundials.jl, group: All}
	- {user: SciML, repo: LinearSolve.jl, group: All}

[vpuri3]

And the test did pass. So I'll remove the line
line 28 - - {user: SciML, repo: LinearSolve.jl, group: Core} , because the test on line 37 - {user: SciML, repo: LinearSolve.jl, group: All} should be sufficient

[vpuri3]

So we should be good to go @ChrisRackauckas ?

[vpuri3]

ok LinearSolve test passed again:

      Testing LinearSolve tests passed 

https://github.com/SciML/SciMLBase.jl/actions/runs/4093861110/jobs/7059588503

[ChrisRackauckas]

and making DiffEqOperators a subtype.

Where is this done?