Question

Is it surprising that the following code outputs a wrong value for startTime?

public class Temp {
    public static void main(String args[]){
       float duration = (float) 2.0;
       long endTime = 1353728995;
       long startTime = 0;
       startTime = (long) (endTime - duration);
       System.out.println(startTime);
}
}

Answer 1

It doesn’t output the wrong value. It just outputs a value that you don’t expect.

The important bit is what happens here:

endTime - duration

This is actually evaluated as:

(float) endTime - duration;

That’s where data is lost. The closest float value to 1353728995 is 1353729024, and the closest float to the result of subtracting 2 is still 1353729024.

This all follows the language specification. JLS section 15.8.2 (additive operators) states that binary numeric promotions is applied to the operands.

Section 5.6.2 (binary numeric promotion) starts off like this:

When an operator applies binary numeric promotion to a pair of operands, each of which must denote a value that is convertible to a numeric type, the following rules apply, in order:

• If any operand is of a reference type, it is subjected to unboxing conversion (§5.1.8).

• Widening primitive conversion (§5.1.2) is applied to convert either or both operands as specified by the following rules:

  • If either operand is of type double, the other is converted to double.

  • Otherwise, if either operand is of type float, the other is converted to float.

  • …

So following those rules, the long value of endTime is converted to float, and then the subtraction is performed in float arithmetic.

Remember that a float only provides 7 significant figures of accuracy, and the rest is fairly obvious, I hope.

Note that without the cast of the result to long, the compiler makes it clearer what’s going on:

Test.java:8: error: possible loss of precision
    long startTime = endTime - duration;
                             ^
  required: long
  found:    float
1 error

That makes it fairly clear that the result is going to be a float, which should have raised warning bells about how the operation was going to be performed, and what accuracy might be expected.