Practical and performance differences between for vs. foreach in Java (and a generic timer-testing class to boot)

(This is also posted as an answer on stackoverflow. See the bottom of this post for the time-difference testing class.)

The foreach loop, added in Java 5 (also called the "enhanced for loop"), is equivalent to using a java.util.Iterator–it’s syntactic sugar for the same thing. Therefore, when reading each element, one by one and in order, a foreach should always be chosen over an iterator, as it is more convenient and concise.

foreach

for(int i : intList)  {
   System.out.println("An element in the list: " + i);
}

Iterator

Iterator<Integer> intItr = intList.iterator();
while(intItr.hasNext())  {
   System.out.println("An element in the list: " + intItr.next());
}

There are situations where you must use an Iterator directly. For example, attempting to delete an element while using a foreach can (will?) result in a ConcurrentModificationException.

foreach vs. for: Basic differences

The only practical difference between for and foreach is that, in the case of indexable objects, you do not have access to the index. An example when the basic for loop is required:

for(int i = 0; i < array.length; i++)  {
   if(i < 5)  {
      //Do something special
   }  else  {
      //Do other stuff
   }
}

Although you could manually create a separate index int-variable with foreach

int idx = -1;
for(int i : intArray)  {
   idx++;
   ...
}

it is not recommended, since variable-scope is not ideal, and the basic for loop is simply the standard and expected format for this use-case.

foreach vs. for: Performance

When accessing collections, a foreach is significantly faster than the basic for loop’s array access. When accessing arrays, however–at least with primitive and wrapper-arrays–access via indexes is dramatically faster.

Timing the difference between iterator and index access for primitive int-arrays

Indexes are 23-40 percent faster than iterators when accessing int or Integer arrays. Here is the output from the testing class at the bottom of this post, which sums the numbers in a 100-element primitive-int array (A is iterator, B is index):

[C:\java_code\]java TimeIteratorVsIndexIntArray 1000000
Test A: 358,597,622 nanoseconds
Test B: 269,167,681 nanoseconds
B faster by 89,429,941 nanoseconds (24.438799231635727% faster)

[C:\java_code\]java TimeIteratorVsIndexIntArray 1000000
Test A: 377,461,823 nanoseconds
Test B: 278,694,271 nanoseconds
B faster by 98,767,552 nanoseconds (25.666236154695838% faster)

[C:\java_code\]java TimeIteratorVsIndexIntArray 1000000
Test A: 288,953,495 nanoseconds
Test B: 207,050,523 nanoseconds
B faster by 81,902,972 nanoseconds (27.844689860906513% faster)

[C:\java_code\]java TimeIteratorVsIndexIntArray 1000000
Test A: 375,373,765 nanoseconds
Test B: 283,813,875 nanoseconds
B faster by 91,559,890 nanoseconds (23.891659337194227% faster)

[C:\java_code\]java TimeIteratorVsIndexIntArray 1000000
Test A: 375,790,818 nanoseconds
Test B: 220,770,915 nanoseconds
B faster by 155,019,903 nanoseconds (40.75164734599769% faster)

[C:\java_code\]java TimeIteratorVsIndexIntArray 1000000
Test A: 326,373,762 nanoseconds
Test B: 202,555,566 nanoseconds
B faster by 123,818,196 nanoseconds (37.437545972215744% faster)

I also ran this for an Integer array, and indexes are still the clear winner, but only between 18 and 25 percent faster.

For collections, iterators are faster than indexes

For a List of Integers, however, iterators are the clear winner. Just change the int-array in the test-class to

List<Integer> intList = Arrays.asList(new Integer[] {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100});

and make the necssary changes to the test-function (int[] to List<Integer>, length to size(), etc)

[C:\java_code\]java TimeIteratorVsIndexIntegerList 1000000
Test A: 3,429,929,976 nanoseconds
Test B: 5,262,782,488 nanoseconds
A faster by 1,832,852,512 nanoseconds (34.326681820485675% faster)

[C:\java_code\]java TimeIteratorVsIndexIntegerList 1000000
Test A: 2,907,391,427 nanoseconds
Test B: 3,957,718,459 nanoseconds
A faster by 1,050,327,032 nanoseconds (26.038700083921256% faster)

[C:\java_code\]java TimeIteratorVsIndexIntegerList 1000000
Test A: 2,566,004,688 nanoseconds
Test B: 4,221,746,521 nanoseconds
A faster by 1,655,741,833 nanoseconds (38.71935684115413% faster)

[C:\java_code\]java TimeIteratorVsIndexIntegerList 1000000
Test A: 2,770,945,276 nanoseconds
Test B: 3,829,077,158 nanoseconds
A faster by 1,058,131,882 nanoseconds (27.134122749113843% faster)

[C:\java_code\]java TimeIteratorVsIndexIntegerList 1000000
Test A: 3,467,474,055 nanoseconds
Test B: 5,183,149,104 nanoseconds
A faster by 1,715,675,049 nanoseconds (32.60101667104192% faster)

[C:\java_code\]java TimeIteratorVsIndexIntList 1000000
Test A: 3,439,983,933 nanoseconds
Test B: 3,509,530,312 nanoseconds
A faster by 69,546,379 nanoseconds (1.4816434912159906% faster)

[C:\java_code\]java TimeIteratorVsIndexIntList 1000000
Test A: 3,451,101,466 nanoseconds
Test B: 5,057,979,210 nanoseconds
A faster by 1,606,877,744 nanoseconds (31.269164666060377% faster)

In one test they’re almost equivalent, but clearly, with collections, iterator wins.

This post is based on two answers I wrote on stackexchange:

Some more information: http://stackoverflow.com/questions/2113216/which-is-more-efficient-a-for-each-loop-or-an-iterator

The full testing class

I created this compare-the-time-it-takes-to-do-any-two-things class after reading this question on stackoverflow

   import  java.text.NumberFormat;
   import  java.util.Locale;
/**
   &lt;P&gt;{@code java TimeIteratorVsIndexIntArray 1000000}&lt;/P&gt;

   @see  &lt;CODE&gt;&lt;A HREF=&quot;http://stackoverflow.com/questions/180158/how-do-i-time-a-methods-execution-in-java&quot;&gt;http://stackoverflow.com/questions/180158/how-do-i-time-a-methods-execution-in-java&lt;/A&gt;&lt;/CODE&gt;
 **/
public class TimeIteratorVsIndexIntArray  {
   public static final NumberFormat nf = NumberFormat.getNumberInstance(Locale.US);
   public static final void main(String[] tryCount_inParamIdx0)  {
      int testCount;
      //Get try-count from command-line parameter
         try  {
            testCount = Integer.parseInt(tryCount_inParamIdx0[0]);
         }  catch(ArrayIndexOutOfBoundsException | NumberFormatException x)  {
            throw  new IllegalArgumentException("Missing or invalid command line parameter: The number of testCount for each test. " + x);
         }

      //Test proper...START
         int[] intArray = new int[] {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44, 45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88, 89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100};

         long lStart = System.nanoTime();
            for(int i = 0; i < testCount; i++)  {
               testIterator(intArray);
            }
         long lADuration = outputGetNanoDuration("A", lStart);

         lStart = System.nanoTime();
            for(int i = 0; i < testCount; i++)  {
               testFor(intArray);
            }
         long lBDuration = outputGetNanoDuration("B", lStart);

         outputGetABTestNanoDifference(lADuration, lBDuration, "A", "B");
   }
      private static final void testIterator(int[] int_array)  {
         int total = 0;
         for(int i = 0; i < int_array.length; i++)  {
            total += int_array[i];
         }
      }
      private static final void testFor(int[] int_array)  {
         int total = 0;
         for(int i : int_array)  {
            total += i;
         }
      }
      //Test proper...END

//Timer testing utilities...START
   public static final long outputGetNanoDuration(String s_testName, long l_nanoStart)  {
      long lDuration = System.nanoTime() - l_nanoStart;
      System.out.println("Test " + s_testName + ": " + nf.format(lDuration) + " nanoseconds");
      return  lDuration;
   }

   public static final long outputGetABTestNanoDifference(long l_aDuration, long l_bDuration, String s_aTestName, String s_bTestName)  {
      long lDiff = -1;
      double dPct = -1.0;
      String sFaster = null;
      if(l_aDuration > l_bDuration)  {
         lDiff = l_aDuration - l_bDuration;
         dPct = 100.00 - (l_bDuration * 100.0 / l_aDuration + 0.5);
         sFaster = "B";
      }  else  {
         lDiff = l_bDuration - l_aDuration;
         dPct = 100.00 - (l_aDuration * 100.0 / l_bDuration + 0.5);
         sFaster = "A";
      }
      System.out.println(sFaster + " faster by " + nf.format(lDiff) + " nanoseconds (" + dPct + "% faster)");
      return  lDiff;
   }
//Timer testing utilities...END
}
Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s