This class is used to simplify error handling logic within operator and function implementations. Note - OperationEval.evaluate() and Function.evaluate() method signatures do not throw this exception so it cannot propagate outside.

Here is an example coded without EvaluationException, to show how it can help:

public Eval evaluate(Eval[] args, int srcRow, short srcCol) {
  // ...
  Eval arg0 = args[0];
  if(arg0 instanceof ErrorEval) {
      return arg0;
  }
  if(!(arg0 instanceof AreaEval)) {
      return ErrorEval.VALUE_INVALID;
  }
  double temp = 0;
  AreaEval area = (AreaEval)arg0;
  ValueEval[] values = area.getValues();
  for (int i = 0; i < values.length; i++) {
      ValueEval ve = values[i];
      if(ve instanceof ErrorEval) {
          return ve;
      }
      if(!(ve instanceof NumericValueEval)) {
          return ErrorEval.VALUE_INVALID;
      }
      temp += ((NumericValueEval)ve).getNumberValue();
  }
  // ...
 }
 

In this example, if any error is encountered while processing the arguments, an error is returned immediately. This code is difficult to refactor due to all the points where errors are returned.
Using EvaluationException allows the error returning code to be consolidated to one place.

public Eval evaluate(Eval[] args, int srcRow, short srcCol) {
  try {
      // ...
      AreaEval area = getAreaArg(args[0]);
      double temp = sumValues(area.getValues());
      // ...
  } catch (EvaluationException e) {
      return e.getErrorEval();
  }
}

private static AreaEval getAreaArg(Eval arg0) throws EvaluationException {
  if (arg0 instanceof ErrorEval) {
      throw new EvaluationException((ErrorEval) arg0);
  }
  if (arg0 instanceof AreaEval) {
      return (AreaEval) arg0;
  }
  throw EvaluationException.invalidValue();
}

private double sumValues(ValueEval[] values) throws EvaluationException {
  double temp = 0;
  for (int i = 0; i < values.length; i++) {
      ValueEval ve = values[i];
      if (ve instanceof ErrorEval) {
          throw new EvaluationException((ErrorEval) ve);
      }
      if (!(ve instanceof NumericValueEval)) {
          throw EvaluationException.invalidValue();
      }
      temp += ((NumericValueEval) ve).getNumberValue();
  }
  return temp;
}
 

It is not mandatory to use EvaluationException, doing so might give the following advantages:
- Methods can more easily be extracted, allowing for re-use.
- Type management (typecasting etc) is simpler because error conditions have been separated from intermediate calculation values.
- Fewer local variables are required. Local variables can have stronger types.
- It is easier to mimic common Excel error handling behaviour (exit upon encountering first error), because exceptions conveniently propagate up the call stack regardless of execution points or the number of levels of nested calls.