From subsecret
Jump to: navigation, search
Line 82: Line 82:
!Map
!Map
!Memory usage (mb)
!Memory usage (mb)
!Insertion time (sec.)
!get(...) (sec.)
!containsKey(...) (sec.)
!containsValue(...) (sec.)
!Removing keys and validating (sec.)
|-
|-
|Java HashMap
|Java HashMap
Line 107: Line 110:


'''JVM 1.7'''
'''JVM 1.7'''
{| class="wikitable"
!Map
!Memory usage (mb)
!get(...) (sec.)
!containsKey(...) (sec.)
!containsValue(...) (sec.)
!Removing keys and validating (sec.)
|-
|Java HashMap
|2483
|7.8
|-
|Colt HashMap
|1709
|7
|-
|Colt (Native) HashMap
|1709
|6.1
|-
|Colt (IntInt) HashMap
|1304
|4.7
|-
|Colt (Quick IntInt) HashMap
|1304
|4.8
|-
|}

Revision as of 12:40, 22 June 2013

Extension of Colt HashMap using Generics for storage of Objects

This is based on part of the code from Colt Project: http://acs.lbl.gov/software/colt/ The Colt Project is made before generics was introduced in Java. This means that the HashMap from Colt Project cannot be used as a drop-in replacement for Java's HashMap.

This version contains only the HashMap part of Colt Project (version 1.2.0) and adds the support for generics, so that it can now be used as a drop-in replacement for Java's HashHamp

Download

JAR also contains source code

Version File Description
1.00 http://files.subsecret.dk/ColtHashMap100.jar Initial release

Changes to original code

  • Removed all non-hashmap relevant code
  • Fixed bug where OpenLongObjectHashMap.indexOfValue(...) and OpenIntObjectHashMap.indexOfValue(...) used "==" rather than equals for Object types.
  • Created new interfaces and classes for wrapping Colt Project HashMap in Java's Map<K,V> interface.

Usage

Class Key type Value type
ColtIntHashMap Int Object
ColtIntIntHashMap Int Int
ColtQuickIntIntHashMap Int Int
ColtIntDoubleHashMap Int Double
ColtLongHashMap Long Object
ColtDoubleIntHashMap Double Int

Existing code can easily be changed to use new HashMaps's. By changing to Colt's HashMap there will be a memory saving because of more efficient storage of keys and possibly values. However the penalty for autoboxing is still present unless using put(...) and get(...) on the native type.

//From
Map<Integer, String> map = new HashMap<Integer,String>();
map.put(3,"test");

//To
ColtIntHashMap<String> native = new ColtIntHashMap<String>();
Map<Integer, String> map = native;
map.put(3,"test");

//For maximum performance, its necessary to use the original type when putting and retrieving from the map.
native.put(3,"test"); 
//Caveat: When using ColtIntHashMap<Integer> the compiler cannot infer which of the two put methods you want, and you need to use putNative(...) method
native.putNative(3,"test");

Benchmark

JVM options: -Xms8g -Xmx8g Insertions: 20 million (7 million duplicates) get(...) 13 million Contains key: 20 million Removals: 6 million containsValue(): 100


JVM 1.6

Map Memory usage (mb) get(...) (sec.) containsKey(...) (sec.) containsValue(...) (sec.) Removing keys and validating (sec.)
Java HashMap 2483 7.8
Colt HashMap 1709 7
Colt (Native) HashMap 1709 6.1
Colt (IntInt) HashMap 1304 4.7
Colt (Quick IntInt) HashMap 1304 4.8

JVM 1.7

Map Memory usage (mb) get(...) (sec.) containsKey(...) (sec.) containsValue(...) (sec.) Removing keys and validating (sec.)
Java HashMap 2483 7.8
Colt HashMap 1709 7
Colt (Native) HashMap 1709 6.1
Colt (IntInt) HashMap 1304 4.7
Colt (Quick IntInt) HashMap 1304 4.8