Generate events in functions; to avoid non-discrete booleans.

[HansOlsson]

Closes #4209

I don't know if significant enough to push for release.

[HansOlsson]

model Model1b
  replaceable package Medium = Modelica.Media.R134a.R134a_ph "Medium model"
    annotation (choicesAllMatching=true);
  Medium.ThermodynamicState state_a;
  Modelica.Units.SI.AbsolutePressure p=Medium.p_default;
  Modelica.Units.SI.SpecificEnthalpy h=Medium.h_default;

equation 
  //state_a = Medium.setState_phX(Medium.p_default, Medium.h_default);
  state_a = Medium.setState_phX(p, h);
end Model1b;

still fails in Dymola 2024x - even with the proposed changes. Would be good to have this model as test model.

Are other functions affected as well?

No, but it seems that I messed up the annotations a bit, it should now be resolved.

[HansOlsson]

still fails in Dymola 2024x - even with the proposed changes. Would be good to have this model as test model.

I added a test-model, but made it similar to an existing one instead.

[beutlich]

I don't know if significant enough to push for release.

For me it seems like a MSL bug. What's the opinion of the Media officers?

[beutlich]

Unfortunately, the new test model ModelicaTest.Media.TestOnly.R134a_setState_phX raises a AssignmentFromNonDiscreteToDiscrete warning in SimulationX 4.4 and fails to simulate with all tested solvers due to a singular system of equations. It does not matter if this PR is applied or not.

[HansOlsson]

Unfortunately, the new test model ModelicaTest.Media.TestOnly.R134a_setState_phX raises a AssignmentFromNonDiscreteToDiscrete warning in SimulationX 4.4 and fails to simulate with all tested solvers due to a singular system of equations. It does not matter if this PR is applied or not.

Ok, that is problematic. I assume the failure to simulate and assignment-error are related; indicating that the current version of the function doesn't work, and not some unrelated error in the example model.

I don't know the exact reason - is it related to calling GenerateEvents indirectly (i.e., one function with GenerateEvents calling another one)? If so it might be possible to rewrite it as:

  state := ThermodynamicState(phase=
         if ((h < bubbleEnthalpy(SaturationProperties(psat=p,Tsat=0)) 
          or (h > dewEnthalpy(SaturationProperties(psat=p,Tsat=0))) or (p > R134aData.data.FPCRIT)) then 1
         else 2 , p=p, h=h, d=density_ph(p, h), T=temperature_ph(p, h));

(I cannot test that in SimulationX.)

So what to do:

• Just add
• Delay until the next release.
• Add the changes in MSL - but don't add the test.
• Use the weird rewriting above - even if horrible copy-pasta. One could argue that this indirect calling isn't fully specified, whereas this rewriting should be ok.

[hubertus65]

I would be in favor to add the rewrite proposed by Hans. It makes the event generation explicit. Could you add it to the PR?

[HansOlsson]

I would be in favor to add the rewrite proposed by Hans. It makes the event generation explicit. Could you add it to the PR?

Should be done now. Please re-test

[HansOlsson]

@gkurzbach can you give some feedback? As I see it, the changes give clear benefits

[casella]

This PR should be back-ported to maint/4.1.0

[Esther-Devakirubai]

Backporting by #4541