Issue with records in Chemical library

[Francesco Casella]

There are several test cases failing in the ​Chemical library, see, e.g., ​Chemical.Examples.EnzymeKinetics, that fail for the same reason:

Chemical_Chemical.Examples.EnzymeKinetics.c:149:3: error:
use of undeclared identifier \'P_stateOfMatter_SubstanceData\'\n  P_stateOfMatter_SubstanceData tmp1;

The flattened model shows the default constructor function

function Chemical.Examples.EnzymeKinetics.P.stateOfMatter.SubstanceData "Automatically generated record constructor for Chemical.Examples.EnzymeKinetics.P.stateOfMatter.SubstanceData"
  input Real MolarWeight = 0.01801528;
  input Real z = 0.0;
  input Real DfG = 0.0;
  input Real DfH = 0.0;
  input Real gamma = 1.0;
  input Real Cp = 0.0;
  input String[1] References = {""};
  input Real DfG_25degC_1bar = 0.0;
  input Real DfH_25degC = 0.0;
  input Boolean SelfClustering = false;
  input Real SelfClustering_dH = 0.0;
  input Real SelfClustering_dS = 0.0;
  input Real density = 1000.0;
  output SubstanceData res;
end Chemical.Examples.EnzymeKinetics.P.stateOfMatter.SubstanceData;

maybe something goes wrong because it contains an array of Strings which is still unsupported?

[Francesco Casella]

Note that these models were successful when handled by the old frontend, as they got broken on master when we ​switched on the new frontend by default.

I hope this holds a clue to fix the problem. What has changed from OF to NF that negatively affects the code generation later on?

[Francesco Casella]

Here are some more clues I got on Windows with the latest nightly build. When flattening the model with the NF, I get

  P.molarEntropyPure = Chemical.Examples.EnzymeKinetics.P.stateOfMatter.molarEntropyPure(
  P.stateOfMatter.SubstanceData(0.01801528, 0.0, -123947.8094685, 0.0, 1.0, 0.0, {""}, 
    0.0, 0.0, false, 0.0, 0.0, 1000.0),
  P.temperature, P.pressure, P.electricPotential, P.moleFractionBasedIonicStrength, {});

and subsequently this C-code

/*
equation index: 210
type: SIMPLE_ASSIGN
P.molarEntropyPure = Chemical.Examples.EnzymeKinetics.P.stateOfMatter.molarEntropyPure(P.stateOfMatter.SubstanceData(0.01801528, 0.0, -123947.8094685, 0.0, 1.0, 0.0, {""}, 0.0, 0.0, false, 0.0, 0.0, 1000.0), solution.temperature, solution.BasePressure, 0.0, 0.0, {})
*/
void Chemical_Examples_EnzymeKinetics_eqFunction_210(DATA *data, threadData_t *threadData)
{
  TRACE_PUSH
  const int equationIndexes[2] = {1,210};
  P_stateOfMatter_SubstanceData tmp0;

which gives trouble because P.stateOfMatter.SubstanceData is not defined, while Chemical.Examples.EnzymeKinetics.P.stateOfMatter.SubstanceData is. I guess the flat code should instead be:

  P.molarEntropyPure = Chemical.Examples.EnzymeKinetics.P.stateOfMatter.molarEntropyPure(
  Chemical.Examples.EnzymeKinetics.P.stateOfMatter.SubstanceData(
    0.01801528, 0.0, -123947.8094685, 0.0, 1.0, 0.0, {""}, 
    0.0, 0.0, false, 0.0, 0.0, 1000.0),
  P.temperature, P.pressure, P.electricPotential, 
  P.moleFractionBasedIonicStrength, {});

When using the OF, the flat model is

P.molarEntropyPure = Chemical.Components.Substance$P.stateOfMatter.molarEntropyPure(
  /*.Chemical.Components.Substance$P.stateOfMatter.SubstanceData*/(P.substanceData), 
  P.temperature, P.pressure, P.electricPotential, P.moleFractionBasedIonicStrength, {});

leading to this C-code

/*
equation index: 416
type: SIMPLE_ASSIGN
P.molarEntropyPure = Chemical.Components.Substance$P.stateOfMatter.molarEntropyPure((*.Chemical.Components.Substance$P.stateOfMatter.SubstanceData*)(P.substanceData), solution.temperature, solution.BasePressure, 0.0, E.solution.I, {})
*/
void Chemical_Examples_EnzymeKinetics_eqFunction_416(DATA *data, threadData_t *threadData)
{
  TRACE_PUSH
  const int equationIndexes[2] = {1,416};
  string_array tmp43;
  string_array_create(&tmp43, ((modelica_string*)&((&data->simulationInfo->stringParameter[2] /* P.substanceData.References[1] PARAM */)[calc_base_index_dims_subs(1, 1, ((modelica_integer) 1))])), 1, 1);
  data->localData[0]->realVars[166] /* P.molarEntropyPure variable */ = omc_Chemical_Components__omcQSubstance_24P_stateOfMatter_molarEntropyPure(threadData, omc_Chemical_Components__omcQSubstance_24P_stateOfMatter__omcQSubstanceData_24P_24substanceData(threadData, data->simulationInfo->realParameter[59] /* P.substanceData.MolarWeight PARAM */, data->simulationInfo->realParameter[64] /* P.substanceData.z PARAM */, data->simulationInfo->realParameter[55] /* P.substanceData.DfG PARAM */, data->simulationInfo->realParameter[57] /* P.substanceData.DfH PARAM */, data->simulationInfo->realParameter[63] /* P.substanceData.gamma PARAM */, data->simulationInfo->realParameter[54] /* P.substanceData.Cp PARAM */, tmp43, data->simulationInfo->realParameter[56] /* P.substanceData.DfG_25degC_1bar PARAM */, data->simulationInfo->realParameter[58] /* P.substanceData.DfH_25degC PARAM */, data->simulationInfo->booleanParameter[11] /* P.substanceData.SelfClustering PARAM */, data->simulationInfo->realParameter[60] /* P.substanceData.SelfClustering_dH PARAM */, data->simulationInfo->realParameter[61] /* P.substanceData.SelfClustering_dS PARAM */, data->simulationInfo->realParameter[62] /* P.substanceData.density PARAM */), data->localData[0]->realVars[5] /* solution.temperature STATE(1) */, data->simulationInfo->realParameter[139] /* solution.BasePressure PARAM */, 0.0, data->localData[0]->realVars[112] /* E.solution.I DUMMY_STATE */, _OMC_LIT15);
  TRACE_POP
}

that can be successfully compiled.

@perost, could you fix the NF so that it doesn't refer to undefined record constructor functions in the function call?

[Per Östlund]

Fixed in 8b85d032.

[Francesco Casella]

Great!

We'll check the ​EnzymeKinetics test tomorrow.

[Adrian Pop]

I guess is no need to wait for tomorrow:
​https://libraries.openmodelica.org/branches/history/master/2021-01-10%2012:09:32..2021-01-12%2001:14:51.html

[Francesco Casella]

EnzymeKinetics now ​works fine, alongside with other 9 models in the library that now simulate.

Other models now pass the code generation phase, but then fail because of event issues, e.g. ​HemoglobinTitration or ​Joels57, but this is another story, see #6339.