Enumeration array with bindings flagged as structurally singular

[Per Östlund]

OMC says that the following model is structurally singular, which it shouldn't be. I'm quite sure it used to work a couple of months ago.

model M
  type E = enumeration(one, two);
  Real x[E] = {1, 2};
end M;

The model works if you don't use an enumeration though:

model M
  Real x[2] = {1, 2};
end M;

[Lennart Ochel]

I will have a look.

[Martin Sjölund]

Postponed until 1.9.1

[Lennart Ochel]

I tried to step throw the code, but it is way too complex. I strongly recommend to revise and simplify it a lot. But that will need some time.
Jens, can you have a look or just give me a hint?

[Lennart Ochel]

The error message is the following:

Error: Model is structurally singular, error found sorting equations 
 1: x[1] = 1.0;
2: x[2] = 2.0;
 for variables 
 x[M.E.two](1), x[M.E.one](2)
Error: Internal error Transformation Module PFPlusExt index Reduction Method Pantelides failed!

It seems that the variables get differently expanded for the variable vector and inside the equations. Hence, both cannot get mapped anymore.

[Lennart Ochel]

This is a problem with the preOptModule <inlineArrayEqn>. After it get applied, the system changes to the following and it is not anymore possible to match the system as I pointed out above:

pre-optimization module inlineArrayEqn:


Variables (2)
========================================
1: x[M.E.two]:VARIABLE() .M, .Real type: Real [2]
2: x[M.E.one]:VARIABLE() .M, .Real type: Real [2]


Equations (2, 2)
========================================
1/1 (1): x[1] = 1.0
2/2 (1): x[2] = 2.0

[Lennart Ochel]

Before that preOptModule get applied, the equation was the following

Equations (1, 2)
========================================
1/1 (2): x = {1.0, 2.0}

The equation should get expanded to the following

1/1 (1): x[M.E.one] = 1.0
2/2 (1): x[M.E.two] = 2.0

[Adrian Pop]

Maybe expandCref replaces enums with integers.
Is a good question if we should keep the enum as they are in the back-end or replace them with integers. From my side I think we should keep them as enums.

[Lennart Ochel]

Replying to adrpo:

Maybe expandCref replaces enums with integers.
Is a good question if we should keep the enum as they are in the back-end or replace them with integers. From my side I think we should keep them as enums.

I agree. Probably we should keep the enumerations. Hence, the equation shoulg get expended as I proposed above.

[Lennart Ochel]

Fixed in r19589.