Invalid model because of lookup in partial class

[casella]

Please check Buildings.Fluid.HeatExchangers.DXCoils.BaseClasses.Examples.NominalCondition. OpenModelica reports that

[Buildings 7.0.0/Fluid/HeatExchangers/DXCoils/BaseClasses/NominalCondition.mo:36:3-40:32:writable] 
Error: Medium is partial, name lookup is not allowed in partial classes.

The package Medium is declared to be a replaceable partial class, but is not subsequently redeclared to a non-partial one before being used. This is forbidden by the Modelica Specification, Section 5.3.2: "The class we look inside may not be partial in a simulation model".

[mwetter]

Correct, a non-partial media need to be assigned.

[casella]

When you fix this, can you pls. back-port it to a 7.0.1 maintenance branch, so we can refer to that for our work?

[mwetter]

Would a 7.0.1 branch be helpful for you, otherwise we would not likely do one but rather focus on getting 8.0.0 out, which has already a lot of fixes.

[mwetter]

@casella : I think we have a bigger problem: Consider

class Cl

  model A
    B myB(redeclare package Medium = Modelica.Media.Air.MoistAir);
  end A;

  model B
    replaceable package Medium = Modelica.Media.Interfaces.PartialMedium;

    // The statement below is not valid, and leads to an error in checking the model B in OMEdit.
    parameter Modelica.SIunits.SpecificEnthalpy h_default = Medium.specificEnthalpy_pTX(
      p=Medium.p_default,
      T=Medium.T_default,
      X=Medium.X_default) "Default enthalpy for the selected medium";
  end B;

end Cl;

which fails the model check for B in OMEdit. Also, Modelica.Fluid.Vessels.ClosedVolume fails to check, as does Pipes.StaticPipe, Valves.ValveLinear and probably many others in the MSL.

The above pattern is used throughout Modelica.Fluid.

[casella]

Would a 7.0.1 branch be helpful for you, otherwise we would not likely do one but rather focus on getting 8.0.0 out, which has already a lot of fixes.

See email

[casella]

@casella : I think we have a bigger problem: Consider

class Cl

  model A
    B myB(redeclare package Medium = Modelica.Media.Air.MoistAir);
  end A;

  model B
    replaceable package Medium = Modelica.Media.Interfaces.PartialMedium;

    // The statement below is not valid, and leads to an error in checking the model B in OMEdit.
    parameter Modelica.SIunits.SpecificEnthalpy h_default = Medium.specificEnthalpy_pTX(
      p=Medium.p_default,
      T=Medium.T_default,
      X=Medium.X_default) "Default enthalpy for the selected medium";
  end B;

end Cl;

which fails the model check for B in OMEdit. Also, Modelica.Fluid.Vessels.ClosedVolume fails to check, as does Pipes.StaticPipe, Valves.ValveLinear and probably many others in the MSL.

The above pattern is used throughout Modelica.Fluid.

That is a problem with the 'Check' feature, which tries to instantiate the model, and of course it can't because it still refers to partial models. We should introduce some relaxations so that instantiation of partial stuff proceeds, see #6217.

Then, we should also report that the model is locally balanced, despite some equations potentially being missing from the partial class. This is covered by #3977, which is not currently our highest priority.

[mwetter]

@casella : In view of https://trac.openmodelica.org/OpenModelica/ticket/6217, I assume this won't need to be fixed in the Buildings library for the maint_7.0.x branch, correct?

[casella]

@mwetter, it will still need to be fixed.

• When you check a model (e.g. a standalone component in a library) it is OK to have missing parameters and/or unredeclared partial classes. The model is not invalid, it is incomplete, but that's expected in a library.
• When you compile a model to simulate it, everything should be in place, including parameter values and non-partial redeclared classes. In particular, as I wrote in the first comment, the Modelica specification requires that "The class we look inside may not be partial in a simulation model".

[mwetter]

@casella : I am not sure we talk about the same: Once https://trac.openmodelica.org/OpenModelica/ticket/6217 is fixed, then models like the example Cl will work, and hence also Buildings 7.0.0/Fluid/HeatExchangers/DXCoils/BaseClasses/NominalCondition.mo should work, as will the models in Modelica.Fluid, which currently (almost all?) fail to check with OMEdit.

Hence, I think keeping models with the pattern as in the example Cl is fine. If not, 100's of models will need to be changed, and the whole design principle of .Fluid in which a user has to redeclare the medium prior to simulation won't work.

Consequently, I would think we don't need to apply this patch to the maint_7.0.x branch. If you prefer to apply the patch so other tests can proceed, I'd be happy to do so.

[casella]

@casella : I am not sure we talk about the same

Neither am I 😄

Once https://trac.openmodelica.org/OpenModelica/ticket/6217 is fixed, then models like the example Cl will work

That depends on your definition of work.

Cl is a technically a model, but it only contains class definitions; there are no components, variables, or equations in it. As such, it is valid and can be legally compiled it into a simulation executable, but it does not compute anything, so it is rather useless as such. In fact, you can both check and compile Cl successfully with today's OMC, and what you'll get out of this is nothing, as expected.

I guess it makes more sense to consider Cl as a package instead, and then check and compile its models Cl.A and Cl.B.

• Cl.A is a valid simulation model, because there are no partial classes in it. As such, it can be both checked and simulated successfully with today's OMC
• Cl.B is a valid model, so it should check successfully, and it will once we fix #6217. However, it is not a valid simulation model, because it calls a function from a partial package, so it lacks some information to be fully defined and, according to Sect. 5.3.2. of the spec, cannot be compiled into simulation code. Indeed, OMC correctly complains that "[Cl: 11:5-14:69]: Medium is partial, name lookup is not allowed in partial classes.". As such, it cannot be compiled.

Component models in the Modelica.Fluid library are like Cl.B, so they can't be simulated as such, but they need their partial medium model to be redeclared.

That said, my original diagnostic on Buildings.Fluid.HeatExchangers.DXCoils.BaseClasses.Examples.NominalCondition was wrong, my fault. That model declares a parameter nomCon of type Buildings.Fluid.HeatExchangers.DXCoils.BaseClasses.NominalCondition, but indeed it redeclares the partial medium package, so I guess there is something wrong with OMC that does not recognize this.

I opened #6221 to address this issue.

[mwetter]

@casella : Great, now we both agree. In this case I will leave maint_7.0.x as is.

[casella]

Yep. Maybe it could still get useful later, when we compare the results.

[casella]

@mwetter I question the label "OpenModelica" here. This issue was because of code which is not compilant to the Modelica Specification. You can't blame OpenModelica because it follows the rules 😃