In the university we were given the following code sample and we were being told, that there is a memory leak when running this code. The sample should demonstrate that this is a situation where the garbage collector can't work.
As far as my object oriented programming goes, the only codeline able to create a memory leak would be
items=Arrays.copyOf(items,2 * size+1);
The documentation says, that the elements are copied. Does that mean the reference is copied (and therefore another entry on the heap is created) or the object itself is being copied? As far as I know, Object and therefore Object[] are implemented as a reference type. So assigning a new value to 'items' would allow the garbage collector to find that the old 'item' is no longer referenced and can therefore be collected.
In my eyes, this the codesample does not produce a memory leak. Could somebody prove me wrong? =)
import java.util.Arrays;
public class Foo
{
private Object[] items;
private int size=0;
private static final int ISIZE=10;
public Foo()
{
items= new Object[ISIZE];
}
public void push(final Object o){
checkSize();
items[size++]=o;
}
public Object pop(){
if (size==0)
throw new ///...
return items[--size];
}
private void checkSize(){
if (items.length==size){
items=Arrays.copyOf(items,2 * size+1);
}
}
}
The code sample doesn't produce a leak. It's true that when you call pop(), the memory isn't freed for the appropriate object - but it will be when you next call push().
It's true that the sample never releases memory. However, the unreleased memory is always re-used. In this case, it doesn't really fit the definition of memory leak.
This will produce memory that isn't freed. However, if you ran the loop again, or a hundred thousand million times, you wouldn't produce more memory that isn't freed. Therefore, memory is never leaked.
You can actually see this behaviour in many malloc and free (C) implementations- when you free memory, it isn't actually returned to the OS, but added to a list to be given back next time you call malloc. But we still don't suggest that free leaks memory.
In the pop() method, the item on the size (i.e, items[size-1]) is not set to NULL. As a result, there still exists reference from objects items to items[size-1], although size has been reduced by one. During GC, items[size-1] won't be collected even if there is no other object pointing to it, which leads to memory leak.
It's not a priori true that there's a memory leak here.
I think the prof has in mind that you're not nulling out popped items (in other words, after you return
items[--size]
, you probably ought to setitems[size] = null
). But when the Foo instance goes out of scope, then everything will get collected. So it's a pretty weak exercise.This example is discussed in
Effective Java
byJoshua Bloch
. The leak is when popping elements. The references keep pointing to objects you don't use.