[Diro7]

Hello,

I'm trying to read from file and store a given number N of dimensioned scalar parameters, where N is defined at runtime by the user.
Such parameters have all the same dimensions (T^-1) and are supposed to be later used in a set of N equations (each of the same form, hence my wish to "vectorize" somehow the solver).

What I tried so far looks like this:

1. read a scalar list from a dictionary entry and assign it to an auxiliary variable of type List<scalar>
2. declare a List<dimensionedScalar> variable
3. iterate over the List<scalar> variable and append to the List<dimensionedScalar> variable a corresponding dimensionedScalar with the desired dimensionSet.

dimensionSet dimLambda(0, 0, -1, 0, 0, 0, 0);

List<scalar> lambda_(dict.lookup("lambda"));

List<dimensionedScalar> lambda;

forAll(lambda_, i)
{
    lambda.append( dimensionedScalar("lambda" + name(i + 1), dimLambda, lambda_[i]) );
}

where the dict entry looks like this:

lambda              ( val1 val2 ... valN );

What I find is that the the code compiles and the dictionary is correctly read.
At runtime, however, I get this error:

--> FOAM FATAL ERROR: 
Different dimensions for =
     dimensions : [0 0 0 0 0 0 0] = [0 0 -1 0 0 0 0]


    From function bool Foam::dimensionSet::operator=(const Foam::dimensionSet&) const
    in file dimensionSet/dimensionSet.C at line 171.

FOAM aborting

which I guess occurs due to the append instruction in the forAll loop.

What am I doing wrong here?
Does anyone know alternative (i.e. simpler or more straightforward) ways to achieve my goal?

Thank you in advance for the help,
Andrea

[ybapat]

You are assigning non-dimensional quantity to dimensional quantity. Dimensions should be consistent.

[Diro7]

Hi Yogesh!

Thank you for the reply.
You are right, the error seems to tell that what I'm trying to do somehow implies the assignment of a dimensioned quantity to a previous variable with different dimensions.

What I'm missing is why this happens and whether am I just doing something wrong or should I use a different approach.
Could you elaborate a little bit more, if you have any suggestion?

Andrea

[HPE]

The dimensions of 'lambda' and 'lambda_' are not the same. Any arithmetic among 'dimensionedScalar's require having the same dimensions.

If the size 'lambda' is equal to the number of cells of your mesh, I would definitely use 'GeometricFields'/' instead of bare 'List'.

Note that you can also state the dimensions instead of using a dimensionSet, e.g. 'dimless/dimTime' instead of 'dimLambda'.

[Diro7]

Hi HPE,

thank you for the reply.

Originally Posted by HPE

The dimensions of 'lambda' and 'lambda_' are not the same. Any arithmetic among 'dimensionedScalar's require having the same dimensions

Yes, this is what the error says. Is there a way to instantiate a List<dimensionedScalar> object with the desired dimensions avoiding this behavior?

Originally Posted by HPE

If the size 'lambda' is equal to the number of cells of your mesh, I would definitely use 'GeometricFields'/' instead of bare 'List'.

In the original solver, I had several different 'lambda1', ... , 'lambdaN' dimensionedScalar parameters which were plugged in N analogous equations.
Now I would like to make it a bit more flexible by having the user specifying N at run-time without modifying the source code to change it.
Since OpenFOAM supports lists, it seemed quite straightforward to me to use them to aggregate analogous parameters with the same dimensions.
Since they are constant scalar parameters, I didn't see the need to store them as fields. How would you suggest to use GeometricFields?


Andrea

[HPE]

Out of my head (for ESI-OF):

dimensionedScalar dS("dS", dimless/dimTime, dict.get<scalar>("dS"));

List<dimensionedScalar> dSL
(
10, // some size
dS
);

[Diro7]

Originally Posted by HPE

Out of my head (for ESI-OF):

dimensionedScalar dS("dS", dimless/dimTime, dict.get<scalar>("dS"));

List<dimensionedScalar> dSL
(
10, // some size
dS
);

I already tried something similar, but it doesn't work. It compiles but gives the usual run-time error complaining about dimensions.

I'm not very into OF implementation details, but it could seem like some sort of bug, as (if I'm getting it right, at least) from documentation List< T > should allow a

List (const label, const T &)

constructor, but dimensionedScalar type just seems not compatible for some reason.

Anyway, I'm speaking of OF7 and I don't know if other OF versions behave differently.

Andrea

[olesen]

Originally Posted by Diro7

I already tried something similar, but it doesn't work. It compiles but gives the usual run-time error complaining about dimensions.

I'm not very into OF implementation details, but it could seem like some sort of bug, as (if I'm getting it right, at least) from documentation List< T > should allow a

List (const label, const T &)

constructor, but dimensionedScalar type just seems not compatible for some reason.

Anyway, I'm speaking of OF7 and I don't know if other OF versions behave differently.

Andrea

If you are remotely interested in performance you should not be using a List of dimensionedScalar. This carries the overhead of a name and dimensions for every single element!



If you are using a mesh or something equivalent, you should consider a DimensionedField. If you need a List only (no mesh reference available), you can easily define your own Dimensioned List.


/mark