Java memory leak with Strings? Why is this consuming and crashing

Question

Lists of Integer, Long, execute ok. However when using String the code below consumes over 5GB of memory when an entry at most should be in the neighborhood of 8 bytes per String, equating to only ~700MB. It also runs infinitely never throwing heap out of bounds. What is going on?

        List<List<String>> arrayList = new ArrayList<List<String>>();
        long offset = 1000;
        long size = 83886080;
        int max = Integer.MAX_VALUE - 100;
        long subloops = 1;
        if(size > max)
        {
            subloops = size / max;
        }

        int temp = 0;
        long count = 1;
        long start = System.nanoTime();
        for(int j=0; j<subloops; j++)
        {
            temp = (int)(size % max);
            arrayList.add(new ArrayList<String>(temp));
            List<String> holder = arrayList.get(j);
            for (long i = 0; i < temp; i++)
            { 
                holder.add(Long.toString(offset + count));
                count++;
            }
            size -= temp;
        }

        long finalTime = System.nanoTime() - start;
        System.out.println("Total time = " + finalTime);
        System.out.println(count);
        //for reference the max item length in bytes ends up being 8
        System.out.println(Long.toString(offset+count).getBytes().length);

You know the base memory overhead of a `String` is, like, at least 24 bytes? — Louis Wasserman, Oct 12 '16 at 04:18
8 bytes per String? On 64bit JVM with CompressedOops, you'ree looking at 12 bytes of header per String, and each one of those has a `char[]` with 16 bytes of header. Strings also have an int to cache their hash, that's another 4 bytes. That's 32 bytes per object just of overhead. Then each char is another 2 bytes. — yshavit, Oct 12 '16 at 04:24

score 3 · Accepted Answer · edited May 23 '17 at 12:16

Memory footprint of a String involves the memory overhead of an object plus the fields of the object. See this answer for detail: What is the memory consumption of an object in Java?

The String object has two instance fields in Java 8:

char value[]
int hash

Assuming 64-bit Java with compressed OOPS, that means memory is:

String:
  Object header:         12 bytes
  Reference to char[]: +  4 bytes
  Integer value:       +  4 bytes
  Data size:           = 20 bytes
  Total aligned size:    24 bytes
char[]:
  Object header:         12 bytes
  Array length:        +  4 bytes
  Characters:          +  2 bytes * length
  Data size (len=8):   = 32 bytes
  Total aligned size:    32 bytes

Add another 4 bytes for the reference to the string (stored in the ArrayList), and you get a total size for String of 8 characters: 60 bytes

Creating 83,886,080 strings then uses 5,033,164,800 bytes = 4.7 Gb.

Java memory leak with Strings? Why is this consuming and crashing

1 Answers1