Which is more efficient, a for-each loop, or an iterator?

Question

Which is the most efficient way to traverse a collection?

List  a = new ArrayList();
for (Integer integer : a) {
  integer.toSt         


        

Answer 1

If you are just wandering over the collection to read all of the values, then there is no difference between using an iterator or the new for loop syntax, as the new syntax just uses the iterator underwater.

If however, you mean by loop the old "c-style" loop:

for(int i=0; i<list.size(); i++) {
   Object o = list.get(i);
}

Then the new for loop, or iterator, can be a lot more efficient, depending on the underlying data structure. The reason for this is that for some data structures, get(i) is an O(n) operation, which makes the loop an O(n²) operation. A traditional linked list is an example of such a data structure. All iterators have as a fundamental requirement that next() should be an O(1) operation, making the loop O(n).

To verify that the iterator is used underwater by the new for loop syntax, compare the generated bytecodes from the following two Java snippets. First the for loop:

List<Integer>  a = new ArrayList<Integer>();
for (Integer integer : a)
{
  integer.toString();
}
// Byte code
 ALOAD 1
 INVOKEINTERFACE java/util/List.iterator()Ljava/util/Iterator;
 ASTORE 3
 GOTO L2
L3
 ALOAD 3
 INVOKEINTERFACE java/util/Iterator.next()Ljava/lang/Object;
 CHECKCAST java/lang/Integer
 ASTORE 2 
 ALOAD 2
 INVOKEVIRTUAL java/lang/Integer.toString()Ljava/lang/String;
 POP
L2
 ALOAD 3
 INVOKEINTERFACE java/util/Iterator.hasNext()Z
 IFNE L3

And second, the iterator:

List<Integer>  a = new ArrayList<Integer>();
for (Iterator iterator = a.iterator(); iterator.hasNext();)
{
  Integer integer = (Integer) iterator.next();
  integer.toString();
}
// Bytecode:
 ALOAD 1
 INVOKEINTERFACE java/util/List.iterator()Ljava/util/Iterator;
 ASTORE 2
 GOTO L7
L8
 ALOAD 2
 INVOKEINTERFACE java/util/Iterator.next()Ljava/lang/Object;
 CHECKCAST java/lang/Integer
 ASTORE 3
 ALOAD 3
 INVOKEVIRTUAL java/lang/Integer.toString()Ljava/lang/String;
 POP
L7
 ALOAD 2
 INVOKEINTERFACE java/util/Iterator.hasNext()Z
 IFNE L8

As you can see, the generated byte code is effectively identical, so there is no performance penalty to using either form. Therefore, you should choose the form of loop that is most aesthetically appealing to you, for most people that will be the for-each loop, as that has less boilerplate code.

Answer 2

We should avoid using traditional for loop while working with Collections. The simple reason what I will give is that the complexity of for loop is of the order O(sqr(n)) and complexity of Iterator or even the enhanced for loop is just O(n). So it gives a performence difference.. Just take a list of some 1000 items and print it using both ways. and also print the time difference for the execution. You can sees the difference.

Answer 3

Iterator is an interface in the Java Collections framework that provides methods to traverse or iterate over a collection.

Both iterator and for loop acts similar when your motive is to just traverse over a collection to read its elements.

for-each is just one way to iterate over the Collection.

For example:

List<String> messages= new ArrayList<>();

//using for-each loop
for(String msg: messages){
    System.out.println(msg);
}

//using iterator 
Iterator<String> it = messages.iterator();
while(it.hasNext()){
    String msg = it.next();
    System.out.println(msg);
}

And for-each loop can be used only on objects implementing the iterator interface.

Now back to the case of for loop and iterator.

The difference comes when you try to modify a collection. In this case, iterator is more efficient because of its fail-fast property. ie. it checks for any modification in the structure of underlying collection before iterating over the next element. If there are any modifications found, it will throw the ConcurrentModificationException.

(Note: This functionality of iterator is only applicable in case of collection classes in java.util package. It is not applicable for concurrent collections as they are fail-safe by nature)

Answer 4

To expand on Paul's own answer, he has demonstrated that the bytecode is the same on that particular compiler (presumably Sun's javac?) but different compilers are not guaranteed to generate the same bytecode, right? To see what the actual difference is between the two, let's go straight to the source and check the Java Language Specification, specifically 14.14.2, "The enhanced for statement":

The enhanced for statement is equivalent to a basic for statement of the form:

for (I #i = Expression.iterator(); #i.hasNext(); ) {
    VariableModifiers(opt) Type Identifier = #i.next();    
    Statement 
}

In other words, it is required by the JLS that the two are equivalent. In theory that could mean marginal differences in bytecode, but in reality the enhanced for loop is required to:

• Invoke the .iterator() method
• Use .hasNext()
• Make the local variable available via .next()

So, in other words, for all practical purposes the bytecode will be identical, or nearly-identical. It's hard to envisage any compiler implementation which would result in any significant difference between the two.

Answer 5

foreach uses iterators under the hood anyway. It really is just syntactic sugar.

Consider the following program:

import java.util.List;
import java.util.ArrayList;

public class Whatever {
    private final List<Integer> list = new ArrayList<>();
    public void main() {
        for(Integer i : list) {
        }
    }
}

Let's compile it with javac Whatever.java,
And read the disassembled bytecode of main(), using javap -c Whatever:

public void main();
  Code:
     0: aload_0
     1: getfield      #4                  // Field list:Ljava/util/List;
     4: invokeinterface #5,  1            // InterfaceMethod java/util/List.iterator:()Ljava/util/Iterator;
     9: astore_1
    10: aload_1
    11: invokeinterface #6,  1            // InterfaceMethod java/util/Iterator.hasNext:()Z
    16: ifeq          32
    19: aload_1
    20: invokeinterface #7,  1            // InterfaceMethod java/util/Iterator.next:()Ljava/lang/Object;
    25: checkcast     #8                  // class java/lang/Integer
    28: astore_2
    29: goto          10
    32: return

We can see that foreach compiles down to a program which:

• Creates iterator using List.iterator()
• If Iterator.hasNext(): invokes Iterator.next() and continues loop

---

As for "why doesn't this useless loop get optimized out of the compiled code? we can see that it doesn't do anything with the list item": well, it's possible for you to code your iterable such that .iterator() has side-effects, or so that .hasNext() has side-effects or meaningful consequences.

You could easily imagine that an iterable representing a scrollable query from a database might do something dramatic on .hasNext() (like contacting the database, or closing a cursor because you've reached the end of the result set).

So, even though we can prove that nothing happens in the loop body… it is more expensive (intractable?) to prove that nothing meaningful/consequential happens when we iterate. The compiler has to leave this empty loop body in the program.

The best we could hope for would be a compiler warning. It's interesting that javac -Xlint:all Whatever.java does not warn us about this empty loop body. IntelliJ IDEA does though. Admittedly I have configured IntelliJ to use Eclipse Compiler, but that may not be the reason why.

Answer 6

The difference isn't in performance, but in capability. When using a reference directly you have more power over explicitly using a type of iterator (e.g. List.iterator() vs. List.listIterator(), although in most cases they return the same implementation). You also have the ability to reference the Iterator in your loop. This allows you to do things like remove items from your collection without getting a ConcurrentModificationException.

e.g.

This is ok:

Set<Object> set = new HashSet<Object>();
// add some items to the set

Iterator<Object> setIterator = set.iterator();
while(setIterator.hasNext()){
     Object o = setIterator.next();
     if(o meets some condition){
          setIterator.remove();
     }
}

---

This is not, as it will throw a concurrent modification exception:

Set<Object> set = new HashSet<Object>();
// add some items to the set

for(Object o : set){
     if(o meets some condition){
          set.remove(o);
     }
}