assert in initial section

[Volker Waurich]

Hi,

In OpenHydraulics.Components.Cylinders.DoubleActingCylinder
You will find:

initial equation 
  assert(cylinderChamberHead.s_rel>=0,"Initial position is smaller than zero");

I suppose that this means to evaluate the assert in the initial system only. Is this valid Modelica? Shall we support this? Currently, this assert hinders some tests to simulate to stopTime.

[Lennart Ochel]

@vwaurich That means this kind of asserts are evaluated for each time step?

[Volker Waurich]

yes. It dont think that we separate asserts between simulation and initialization dae. At least in BackendDAECreate.lower, every assert ends up in the removedEqs section.

[Lennart Ochel]

I guess it is then just the same for other initial equations that end up in removedEqs. I will have a look.

[Lennart Ochel]

Okay, it seems that we should improve BackendDAECreate.lower a bit. However, we need to fix a front end issue first: The assert is moved to an equation section as following example demonstrates:

loadString("
model A
  Real x(start=1.0, fixed=true) = -der(x);
initial equation
  assert(time < 0.5, \"Test assert\");
initial algorithm
  assert(time < 0.6, \"Test assert\");
end A;
"); getErrorString();
instantiateModel(A); getErrorString();

// class A
//   Real x(start = 1.0, fixed = true) = -der(x);
// initial algorithm
//   assert(time < 0.6, \"Test assert\");
// equation
//   assert(time < 0.5, \"Test assert\");
// end A;

Is there a volunteer who wants to fix the front end? :)

[Adrian Pop]

I'll take the front-end part. Seems like missing an initial flag propagation.

[Adrian Pop]

We don't have DAE.INITIAL_ASSERT we only have DAE.ASSERT. I will add it, but it will need to be handled in the back-end.
Can one have terminate() as initial equation?

[Lennart Ochel]

I don't think that initial terminate is necessary, but I'm not sure. Anyway, I will take care of the back end part.

[Adrian Pop]

I added it anyway. How about when equations in the initial section?

[Lennart Ochel]

There are no when equations in initial equations (only when initial() in the dynamic equations).

[Adrian Pop]

Fixed the front-end part in: b8fe18/OMCompiler.
Added test in: 079d78/OpenModelica-testsuite.
@lochel: take it away :)

[Lennart Ochel]

Thanks @adrpo. I just fixed classification of initial equations and added support for initial terminate statements to the c runtime: ​OMCompiler#1537.

[Lennart Ochel]

The coverage tests look good - OpenHydraulics went up a bit.