Originally posted by: SystemAdmin

[philippe said:]

I've a problem with array lengths in OPL.

From the C++ API one can do something like this:


struct Container
{
  std::vector<float> floats;
};

void test()
{
  std::vector<Container> containers;
 
  IloEnv env;
  IloArray<IloNumVarArray> vars( env );

  for ( int i = 0; i < containers.size(); ++i ) {<br />    vars.add( IloNumVarArray( env ) );

    for ( int j = 0; j < containers&#91;i&#93;.floats.size(); ++j ) {<br />      vars[i].add( IloNumVar( env ) );
    }
  }
}


The lengths of the inner float vectors don't have to be the same. Works fine and saves some variables.

Not so with OPL:


tuple Container
{
  {float} floats;
}

{Container} containers = ...;

float vars[i in containers][j in i.floats] = 0;


The code gives the following error:
[b]
Variable indexer size not allowed for a generic array.
[/b]

Any idea how to work around this? Otherwise I've to find out the longest float set and constrain the unused variables to something reasonable...

Thanks for any hints.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[jfk said:]

hello,
I hate bringing bad news: but you can't do dynamic arrays in OPL, you have to create the whole 2D array with the max length (sort of suggested already).


cheers
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[philippe said:]

Ok, thanks for the answer. At least I know it for sure now ;-)

Is there any hope to have this feature implemented in a future release or is it a language limitation that won't be overcome?
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[jfk said:]

Hello,
the clear and loud answer is: yes, in the future it will be there. However, what is not clear when it can happen... It brings lots of benefits not just convenience...

cheers

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[philippe said:]

Thanks for the info. I'm really looking forward to it.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[afleischer said:]

Hi,

the suggested way today when you deal with sparse structures is to use tuple sets.

Alex
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[afleischer said:]

Let me give you some details:



tuple Container
{
  {float} floats;
}

{Container} containers = ...;

tuple t
{
Container c;
float f;
}

{t} s={<i,j> | i in containers, j in i.floats};

and then you have a light array

float vars[s] = 0;


#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

[philippe said:]

Thanks for the details!

As long as the feature isn't available I'll use the suggested workaround.
#DecisionOptimization
#OPLusingCPLEXOptimizer