Hi,
The int in Entry is a 32bit value so it takes 4 bytes of memory.
In a 32bit JVM, a reference, ie. a pointer takes 4 bytes. In a 64bit
JVM it takes 8 bytes.
HashMap's Entry consumes on a 32bit jvm:
4 bytes for the class identifier,
4 bytes for the object identifier,
4 bytes for the key
4 bytes for the value
4 bytes for the hash
4 bytes for next
a total of 24 bytes.
and on a 64bit jvm
4 bytes for the class identifier,
4 bytes for the object identifier,
8 bytes for the key
8 bytes for the value
4 bytes for the hash
8 bytes for next
36 bytes.
IntHashMap's entry takes:
4 bytes for the class identifier,
4 bytes for the object identifier,
4 bytes for key
4 bytes for value
4 bytes for next
20 total on 32bits
or
4 bytes for the class identifier,
4 bytes for the object identifier,
4 bytes for key
8 bytes for value
8 bytes for next
28 total on 64bits
java.util.AbstractMap$SimpleEntry<K,V> and consequently
java.util.concurrent.ConcurrentHashMap$WriteThroughEntry
only has serialVersionUID, key and value.
which is 24 bytes on 32bit and 32bytes on 64bit.
As an aside, on my core2, the 32-bit hotspot seems to align objects on
8byte boundaries, so 20 and 24 bytes both consume 24 bytes.
My personal conclusion would be that if there is a memory saving
advantage to using IntHashMap, it is negligible. Further, the
accidental complexity IntHashMap introduces outweighs the memory
savings.
Br,
Tuomas
Quoting Johan Compagner <[email protected]>:
the serialized size doesnt have to be that different if you use Integer key
objects because hashmap has its own special serialization where it doesnt
have to store everything thats in mem
but just the key/value combination
But in memory we have an Entry object:
HashMap.Entry implements Map.Entry {
final Object key;
Object value;
final int hash;
Entry next;
IntHashMap.Entry
{
final int key;
Object value;
Entry next;
so you see thats one reference less per entry.
Also the lookup could be faster because it doesnt have to calculate the hash
on for example strings.
The IntHashMap just skips over the object key hash and gives you the option
to give that hash yourself.
This only works ofcourse for the same type of objects where the hash is
always different for any object, so it wont work as a replacement of String
key, but it does for Integer key
johan
On Fri, Feb 13, 2009 at 09:43, Tuomas Kärkkäinen
<[email protected]>wrote:
Hi,
I couldn't gather from the javadoc when IntHashMap should be preferred over
a regular HashMap, i.e. how many objects per map, what type of objects etc.
I wrote a little program (copy pasted at the bottom) to see how much the
size difference was for IntHashMap, java.util.HashMap and
java.util.concurrent.ConcurrentHashMap.
for 1 instance of new org.apache.wicket.Page(){};
for 10 instances of new org.apache.wicket.Page(){};
for 100 instances of new
org.apache.wicket.ajax.form.AjaxFormSubmitTestPage();
Br,
Tuomas
the results are as follows, respectively:
--------------------------------------------------
map implementation: org.apache.wicket.util.collections.IntHashMap -- size
of outputstream: 54 bytes.
map implementation: java.util.HashMap -- size of outputstream: 57 bytes.
map implementation: java.util.concurrent.ConcurrentHashMap -- size of
outputstream: 282 bytes.
HashMap is 5.556% larger than IntHashMap.
ConcurrentHashMap is 422.222% larger than IntHashMap.
--
map implementation: org.apache.wicket.util.collections.IntHashMap -- size
of outputstream: 369 bytes.
map implementation: java.util.HashMap -- size of outputstream: 399 bytes.
map implementation: java.util.concurrent.ConcurrentHashMap -- size of
outputstream: 624 bytes.
HashMap is 8.130% larger than IntHashMap.
ConcurrentHashMap is 69.106% larger than IntHashMap.
--
map implementation: org.apache.wicket.util.collections.IntHashMap -- size
of outputstream: 24272 bytes.
map implementation: java.util.HashMap -- size of outputstream: 24572 bytes.
map implementation: java.util.concurrent.ConcurrentHashMap -- size of
outputstream: 24797 bytes.
HashMap is 1.236% larger than IntHashMap.
ConcurrentHashMap is 2.163% larger than IntHashMap.
----------
/*
* Licensed to the Apache Software Foundation (ASF) under one or more
* contributor license agreements. See the NOTICE file distributed with
* this work for additional information regarding copyright ownership.
* The ASF licenses this file to You under the Apache License, Version 2.0
* (the "License"); you may not use this file except in compliance with
* the License. You may obtain a copy of the License at
*
* http://www.apache.org/licenses/LICENSE-2.0
*
* Unless required by applicable law or agreed to in writing, software
* distributed under the License is distributed on an "AS IS" BASIS,
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
* See the License for the specific language governing permissions and
* limitations under the License.
*/
package org.apache.wicket.util.collections;
import java.io.IOException;
import java.util.HashMap;
import java.util.concurrent.ConcurrentHashMap;
import org.apache.wicket.Page;
import org.apache.wicket.util.io.ByteArrayOutputStream;
import org.apache.wicket.util.io.WicketObjectOutputStream;
import org.apache.wicket.util.tester.WicketTester;
public class Tester
{
private static final int NUMBER_OF_PAGES = 1;
public static void main(String[] args) throws Exception
{
int intHashMapSize = writeMap(populatedIntHashMap());
int hashMapSize = writeMap(populatedHashMap());
int concurrentHashMapSize =
writeMap(populatedConcurrentHashMap());
double hashMapIsLargerPercetange =
((Double.valueOf(hashMapSize) / intHashMapSize) - 1) * 100;
double concurrentHashMapIsLargerPercetange =
((Double.valueOf(concurrentHashMapSize) / intHashMapSize) - 1) * 100;
System.out.printf("HashMap is %.3f%% larger than
IntHashMap.%n", hashMapIsLargerPercetange);
System.out.printf("ConcurrentHashMap is %.3f%% larger than
IntHashMap.%n",
concurrentHashMapIsLargerPercetange);
}
private static int writeMap(Object map) throws IOException
{
ByteArrayOutputStream out = new ByteArrayOutputStream();
WicketObjectOutputStream objectOutputStream = new
WicketObjectOutputStream(out);
objectOutputStream.writeObject(map);
System.out.printf("map implementation: %s -- size of
outputstream: %d bytes.%n",
map.getClass().getName(), out.size());
return out.size();
}
private static HashMap<Integer, Page> populatedHashMap()
{
new WicketTester();
HashMap<Integer, Page> map = new HashMap<Integer, Page>();
for (int i = 0; i < NUMBER_OF_PAGES; i++)
{
map.put(i, createPage());
}
return map;
}
private static ConcurrentHashMap<Integer, Page>
populatedConcurrentHashMap()
{
new WicketTester();
ConcurrentHashMap<Integer, Page> map = new
ConcurrentHashMap<Integer, Page>();
for (int i = 0; i < NUMBER_OF_PAGES; i++)
{
map.put(i, createPage());
}
return map;
}
private static IntHashMap<Page> populatedIntHashMap()
{
new WicketTester();
IntHashMap<Page> map = new IntHashMap<Page>();
for (int i = 0; i < NUMBER_OF_PAGES; i++)
{
map.put(i, createPage());
}
return map;
}
private static Page createPage()
{
Page page = new Page()
{
};
page.detach();
return page;
}
}
Igor Vaynberg wrote:
exactly what the javadoc says, its a map that does not need to store a
key object. so it is smaller than a regular hashmap when serialized or
kept in memory.
-igor
On Thu, Feb 12, 2009 at 1:40 PM, Tuomas Kärkkäinen
<[email protected]> wrote:
Hi,
I was looking at IntHashMap in wicket trunk.
I couldn't figure out what it does.
The implementation of hashcode for Integer is value, and the autoboxing
of
int to Integer comes out to Integer.valueOf(int) which is implemented as
an
array lookup for Integers in the range of -128 to 127, and beyond that
it's
just new Integer(int). Page ids start at zero for each page in each
session
so most of the time they will be between zero and 127.
Br,
Tuomas
