Originally posted by: UDOPS

Is there a rounding function, to a defined number of decimals? I see the "round" function in the documentation, but it is actually an int() or trunc() function.

Right now I am using Math.round(### * 1000)/1000 to obtain rounding to three decimals.
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

Even that does not work all the time.
Try

execute{
  var r = 9102.888199999999;
  writeln(" --> ", Math.round( r * 1000)/1000); 
}

which yields

--> 9102.888000000001

Question to the OPL-experts: Does the problem come from
the arithmetics or from the writeln?

Regards
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

Even more confusing

78019.90029999999 --> 78019.89999999999

#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: UDOPS

I think I can answer the second example, it is floating point error. I see it frequently, especially in the output of Excel Solver problems which otherwise possess the integrality property (transportation models from textbooks).

The "000000001" add-on is the result of floating point chip manipulations.

I have also seen this in older FileMaker Pro databases, where a random addition of two integers a+b would yield (a+b).00000000001. This was quite a stress for invoicing and accounting databases!
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: SystemAdmin

Thanks for the remarks.
I agree, it is somewhat annoying.

Regards
Norbert
#DecisionOptimization
#OPLusingCPLEXOptimizer

Originally posted by: UDOPS

So, to wrap up, there is no appropriate function in OPL/ILOG for rounding to a specified number of significant figures?
#DecisionOptimization
#OPLusingCPLEXOptimizer