Originally posted by: SystemAdmin

Hi,

it seems that there is a different interpretation of breakpoints for piecewise() and stepwise().
Consider the following small model:

dvar int+ x; dexpr 

float obj_pwf =  piecewise
{ 0->10; 1 
} (0, 0) x; dexpr 

float obj_swf =  stepwise
{ 0->10; 10 
} x;   dexpr 

float obj = obj_pwf + obj_swf; minimize obj; subject to
{ x == 10; 
}   execute
{ writeln(
"     x = ", x); writeln(
"    obj      = ", obj); writeln(
"    obj_pwf  = ", obj_pwf); writeln(
"    obj_swf  = ", obj_swf); 
}

which yields some strange output :

// solution (optimal) with objective 0 x = 10 obj      = 10 obj_pwf  = 0 obj_swf  = 10

First, OPL claims "solution (optimal) with objective 0" although we have "obj=10"
(what changes to "solution (optimal) with objective 10" if we maximize instead)

Moreover, what is the reason for handling the breakpoints differently? In case
of pwf the "old slope" is applied whereas for swf the "new step" is applied?
Quite confusing ...
Best regards
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer

Hi,

apart from the strange objective we get in the IDE (which should be equal to obj), this looks normal.

dexpr float obj_pwf =  piecewise{ 0->10; 1 } (0, 0) x;
dexpr float obj_swf =  stepwise{ 0->10; 10 } x;

In obj_pwf, the slope is 0 til x=10 and then the slope becomes 1
In obj_swf, the value is 0 before x=10, 10 excluded, and then the value becomes 10.

regards
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

Alex,
you are right.
Thanks
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer