Keep non-array connector variable collected

[baggepinnen]

The PR

• fix: fix guesses for Real*Input and Real*Output #291

was breaking, in particular, the removal of

u = collect(u)

from RealInput and RealOutput broke Multibody.jl

Since we have both RealInput and RealInputArray, where the latter is using symbolic arrays, I propose that we keep the old behavior of RealInput intact, i.e., revert the breaking change.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 51.57%. Comparing base (200b13a) to head (f4ada5b).
Report is 2 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #296      +/-   ##
==========================================
- Coverage   61.56%   51.57%   -9.99%     
==========================================
  Files          46       46              
  Lines        1470     1778     +308     
==========================================
+ Hits          905      917      +12     
- Misses        565      861     +296     

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

[AayushSabharwal]

I think tests might fail here. The reason to not scalarize this is so that the non-scalarized form is available through guesses. Once tests pass here, I'll also run MTK tests (using SciML/ModelingToolkit.jl#2676) with this checked out, since it has the potential to break those too

[baggepinnen]

The reason to not scalarize this is so that the non-scalarized form is available through guesses.

Why is this needed, wouldn't using RealOutputArray be an option if you need the non-scalarized version?

[baggepinnen]

Regarding the test failure here

  LoadError: Initial condition underdefined. Some are missing from the variable map.
  Please provide a default (`u0`), initialization equation, or guess
  for the following variables:
  
  Any[P₊input₊u(t)]

I have tests passing locally on the https://github.com/AayushSabharwal/ModelingToolkit.jl/tree/as/linearization-initialization branch of MTK so this PR will have to wait on

• fix: create and solve initialization system in linearization_function ModelingToolkit.jl#2676

[baggepinnen]

tests are passing on MTK master branch