Merge pull request #56 from torokati44/master

Add option to compare by value when memoizing

java/src/main/java/net/razorvine/pickle/Pickler.java

@@ -82,6 +82,12 @@ public Memo(Object obj, int index) {
 	 */
 	protected boolean useMemo=true;
 
+	/**
+	 * When memoizing, compare objects by value. This saves pickle size, but can slow down pickling.
+	 * Alo, it should only be used if the object graph is immutable. Unused if useMemo is false.
+	 */
+	protected boolean valueCompare=true;
+
 	/**
 	 * The memoization cache.
 	 */
@@ -100,7 +106,17 @@ public Pickler() {
 	 * If you use a memo table, you can only pickle objects that are hashable.
 	 */
 	public Pickler(boolean useMemo) {
+		this(useMemo, false);
+	}
+
+	/**
+	 * Create a Pickler. Also specify if it is to compare objects by value.
+	 * If you compare objects by value, the object graph might be altered,
+	 * as different instances with the same value will be unified.
+	 */
+	public Pickler(boolean useMemo, boolean valueCompare) {
 		this.useMemo=useMemo;
+		this.valueCompare=valueCompare;
 	}
 
 	/**
@@ -184,15 +200,15 @@ public void save(Object o) throws PickleException, IOException {
 	 * Write the object to the memo table and output a memo write opcode
 	 * Only works for hashable objects
 	 */
-    protected void writeMemo( Object obj ) throws IOException
-    {
+	protected void writeMemo( Object obj ) throws IOException
+	{
 		if(!this.useMemo)
 			return;
-		int idHash = System.identityHashCode(obj);
-		if(!memo.containsKey(idHash))
+		int hash = valueCompare ? obj.hashCode() : System.identityHashCode(obj);
+		if(!memo.containsKey(hash))
 		{
 			int memo_index = memo.size();
-			memo.put(idHash, new Memo(obj, memo_index));
+			memo.put(hash, new Memo(obj, memo_index));
 			if(memo_index<=0xFF)
 			{
 				out.write(Opcodes.BINPUT);
@@ -213,10 +229,10 @@ protected void writeMemo( Object obj ) throws IOException
 	private boolean lookupMemo(Class<?> objectType, Object obj) throws IOException {
 		if(!this.useMemo)
 			return false;
-		int idHash = System.identityHashCode(obj);
 		if(!objectType.isPrimitive()) {
-			if(memo.containsKey(idHash) && memo.get(idHash).obj == obj) { // same object
-				int memo_index = memo.get(idHash).index;
+			int hash = valueCompare ? obj.hashCode() : System.identityHashCode(obj);
+			if(memo.containsKey(hash) && (valueCompare ? memo.get(hash).obj.equals(obj) : memo.get(hash).obj == obj)) { // same object or value
+				int memo_index = memo.get(hash).index;
 				if(memo_index <= 0xff) {
 					out.write(Opcodes.BINGET);
 					out.write((byte) memo_index);

java/src/test/java/net/razorvine/examples/ValueCompareExample.java

@@ -0,0 +1,54 @@
+package net.razorvine.examples;
+
+import java.util.*;
+import java.io.IOException;
+
+import net.razorvine.pickle.Pickler;
+import net.razorvine.pickle.Unpickler;
+
+public class ValueCompareExample {
+
+    public static void main(String[] args) throws IOException {
+
+        Random random = new Random(1337);
+        List<String> values = new ArrayList<String>();
+        for (int i = 0; i < 100000; ++i) {
+            values.add(("This is a string with a number in it: " + random.nextInt(100))
+                    // .intern() // You could also see what happens when the strings are interned
+            );
+        }
+
+        long t0 = System.nanoTime();
+        byte[] noValueCompare = new Pickler(true, false).dumps(values);
+        long t1 = System.nanoTime();
+        byte[] withValueCompare = new Pickler(true, true).dumps(values);
+        long t2 = System.nanoTime();
+
+        System.out.println("Pickle size without: " + noValueCompare.length + "B, with: " + withValueCompare.length + "B"
+                + " (" + (int) ((1 - withValueCompare.length / (double) noValueCompare.length) * 100.0) + "% smaller)");
+
+        Unpickler unpickler = new Unpickler();
+
+        long t3 = System.nanoTime();
+        List<String> without = (List<String>) unpickler.loads(noValueCompare);
+        long t4 = System.nanoTime();
+        List<String> with = (List<String>) unpickler.loads(withValueCompare);
+        long t5 = System.nanoTime();
+
+        System.out.println("Whole List equality: ==: " + (without == with) + ", .equals(): " + (without.equals(with)));
+
+        System.out.println(
+                "Two equal elements equality without valueCompare: ==: " + (without.get(107) == without.get(259))
+                        + ", .equals(): " + (without.get(107).equals(without.get(259))));
+        System.out.println("Two equal elements equality with valueCompare: ==: " + (with.get(107) == with.get(259))
+                + ", .equals(): " + (with.get(107).equals(with.get(259))));
+
+        System.out.println("Pickling without valueCompare took " + (t1 - t0) / 1000000.0 + "ms");
+        System.out.println("Pickling with valueCompare took " + (t2 - t1) / 1000000.0 + "ms");
+
+        System.out.println("Unpickling without valueCompare took " + (t4 - t3) / 1000000.0 + "ms");
+        System.out.println("Unpickling with valueCompare took " + (t5 - t4) / 1000000.0 + "ms");
+
+    }
+
+}