How to convert float to int with Java

I used the following line to convert float to int, but it's not as accurate as I'd like:

 float a=8.61f;
 int b;

 b=(int)a;

The result is : 8 (It should be 9)

When a = -7.65f, the result is : -7 (It should be -8)

What's the best way to do it ?

Using Math.round() will round the float to the nearest integer.

Actually, there are different ways to downcast float to int, depending on the result you want to achieve: (for int i, float f)

• round (the closest integer to given float)

  i = Math.round(f);
    f =  2.0 -> i =  2 ; f =  2.22 -> i =  2 ; f =  2.68 -> i =  3
    f = -2.0 -> i = -2 ; f = -2.22 -> i = -2 ; f = -2.68 -> i = -3
  

  note: this is, by contract, equal to (int) Math.floor(f + 0.5f)

• truncate (i.e. drop everything after the decimal dot)

  i = (int) f;
    f =  2.0 -> i =  2 ; f =  2.22 -> i =  2 ; f =  2.68 -> i =  2
    f = -2.0 -> i = -2 ; f = -2.22 -> i = -2 ; f = -2.68 -> i = -2
  

• ceil/floor (an integer always bigger/smaller than a given value if it has any fractional part)

  i = (int) Math.ceil(f);
    f =  2.0 -> i =  2 ; f =  2.22 -> i =  3 ; f =  2.68 -> i =  3
    f = -2.0 -> i = -2 ; f = -2.22 -> i = -2 ; f = -2.68 -> i = -2
  
  i = (int) Math.floor(f);
    f =  2.0 -> i =  2 ; f =  2.22 -> i =  2 ; f =  2.68 -> i =  2
    f = -2.0 -> i = -2 ; f = -2.22 -> i = -3 ; f = -2.68 -> i = -3
  

For rounding positive values, you can also just use (int)(f + 0.5), which works exactly as Math.Round in those cases (as per doc).

You can also use Math.rint(f) to do the rounding to the nearest even integer; it's arguably useful if you expect to deal with a lot of floats with fractional part strictly equal to .5 (note the possible IEEE rounding issues), and want to keep the average of the set in place; you'll introduce another bias, where even numbers will be more common than odd, though.

See

http://mindprod.com/jgloss/round.html

http://docs.oracle.com/javase/6/docs/api/java/lang/Math.html

for more information and some examples.