Missing default connection equations for unconnected flow/effort connectors in expandable connectors

[casella]

Description

Consider this MWE package:

package TestExpandableConnectors3
   
  expandable connector C
  end C;

  model ModelA
    C c;
    Modelica.Electrical.Analog.Basic.Ground ground1, ground2, ground3;
  equation
    connect(ground1.p, c.p1);
    connect(ground2.p, c.p2);
    connect(ground3.p, c.p3);
  end ModelA;
  
  model ModelB
    C c;
    Modelica.Electrical.Analog.Basic.Ground ground1;
  equation
    connect(ground1.p, c.p1);
  end ModelB;

  model System
    ModelA m1;
    ModelA m2;
    ModelB m3;
  equation
    connect(m1.c, m2.c);
    connect(m1.c, m3.c);
  end System;
end TestExpandableConnectors3;

Checking System gives:

Class TestExpandableConnectors3.System has 30 equation(s) and 32 variable(s).

Inspection of the flattened model reveals that connector m3.c contains all the flow variables obtained by gathering the components of the expandable connectors:

  Real m3.c.p1.i "virtual variable in expandable connector";
  Real m3.c.p2.i "virtual variable in expandable connector";
  Real m3.c.p3.i "virtual variable in expandable connector";

Unfortunately, only one equation is generated from the connect(ground1.p, c.p1); statement in ModelB

  m3.ground1.p.i - m3.c.p1.i = 0.0;

leaving m3.c.p2.i and m3.c.p3.i orphans of their matched equations.

As I understand, what are missing here are the default equations for unconnected ports in m3:

  m3.c.p2.i = 0.0;
  m3.c.p3.i = 0.0;

[casella]

Keeping @albertoleva in the loop

[casella]

Possibly related to discussion #12160

[perost]

We discussed this issue during the meeting, but afterwards I realised I was still confused about why the model shouldn't be unbalanced since all the flow variables are connected as inside connectors due to the connect equations in System.

But it turns out that I'd missed one very crucial thing in the specification (9.2): flow variables added when augmenting expandable connectors should be added as connection sets both as inside and as outside. So we need to keep track of what's added during augmentation and add sets with such flow variables as outside too in addition to the usual inside ones.

[casella]

But it turns out that I'd missed one very crucial thing in the specification (9.2): flow variables added when augmenting expandable connectors should be added as connection sets both as inside and as outside.

Yes, that is the crucial point.

So we need to keep track of what's added during augmentation and add sets with such flow variables as outside too in addition to the usual inside ones.

Yep.

If we follow my proposal in #12160, we could eventually avoid adding a lot of these variables and connection sets. Given the currently standing specification, however, we have to track them all, even though they will lead to a lot of alias equations and a lot of trivial c.f = 0 equations.

[casella]

@albertoleva tomorrow's nightly build should work fine with expandable connectors.