Merge pull request #62 from LukeButtersFunnelback/master

Stop Trie#removePartialMatches() from being expensive #61

src/main/java/org/ahocorasick/trie/Trie.java

@@ -5,6 +5,8 @@ import org.ahocorasick.interval.Intervalable;
 import org.ahocorasick.trie.handler.DefaultEmitHandler;
 import org.ahocorasick.trie.handler.EmitHandler;
 import org.ahocorasick.trie.handler.StatefulEmitHandler;
+import org.ahocorasick.util.ListElementRemoval;
+import org.ahocorasick.util.ListElementRemoval.RemoveElementPredicate;
 
 import java.util.ArrayList;
 import java.util.Collection;
@@ -207,17 +209,17 @@ public class Trie {
     }
 
     private void removePartialMatches(final CharSequence searchText, final List<Emit> collectedEmits) {
-        final List<Emit> removeEmits = new ArrayList<>();
+        
-
+        final RemoveElementPredicate<Emit> predicate = new RemoveElementPredicate<Emit>() {
-        for (final Emit emit : collectedEmits) {
+            
-            if (isPartialMatch(searchText, emit)) {
+            @Override
-                removeEmits.add(emit);
+            public boolean remove(Emit emit) {
+                return isPartialMatch(searchText, emit);
             }
-        }
+            
-
+        };
-        for (final Emit removeEmit : removeEmits) {
+        
-            collectedEmits.remove(removeEmit);
+        ListElementRemoval.removeIf(collectedEmits, predicate);
-        }
     }
 
     private void removePartialMatchesWhiteSpaceSeparated(final CharSequence searchText, final List<Emit> collectedEmits) {

src/main/java/org/ahocorasick/util/ListElementRemoval.java

@@ -0,0 +1,50 @@
+package org.ahocorasick.util;
+
+import java.util.ArrayList;
+import java.util.List;
+
+/**
+ * Helps removes elements from a list in a efficient way
+ * 
+ * <p>Removing elements from an ArrayList in a naive way can lead to O(n^3)
+ * running time. If the algorithm first creates a list of all the elements
+ * to remove, then we for each element in this list (assume n elements) we look 
+ * for the element in the original list (against n elements) and when found we need 
+ * to remove the element and move the elements to the right (of the removed element) 
+ * to the left by one, the size of this operation is at worst n hence O(n^3).</p>
+ * 
+ * <p>This instead makes a new list and copies over only elements we want to keep, 
+ * we then clear the original list and then add all of the elements to the original 
+ * list. This gives us (for ArrayList) a running time of O(n).</p>
+ * 
+ * <p>The performance of this has not been thoroughly tested for linked list.</p>
+ * 
+ * <p>This can be completely removed in java 8 as the List#removeIf() method can be used instead
+ * as this already is optimised for each list implementation.
+ *
+ */
+public class ListElementRemoval {
+
+    public static interface RemoveElementPredicate<T> {
+        public boolean remove(T t);
+    }
+    
+    /**
+     * Removes all elements from the list matching the given predicate.
+     * 
+     * @param list
+     * @param predicate
+     */
+    public static <T> void removeIf(final List<T> list, final RemoveElementPredicate<T> predicate) {
+        final List<T> newList = new ArrayList<>(list.size());
+        
+        for(final T element : list) { 
+            if (!predicate.remove(element)) {
+                newList.add(element);
+            }
+        }
+        
+        list.clear();
+        list.addAll(newList);
+    }
+}

src/test/java/org/ahocorasick/util/ListElementRemovalTest.java

@@ -0,0 +1,65 @@
+package org.ahocorasick.util;
+
+import java.util.ArrayList;
+import java.util.List;
+
+import org.ahocorasick.util.ListElementRemoval.RemoveElementPredicate;
+import org.junit.Test;
+
+import junit.framework.Assert;
+
+import static java.util.Arrays.asList;
+public class ListElementRemovalTest {
+
+    @Test
+    public void removeNone() {
+        List<String> list = new ArrayList<>(asList("a", "b", "c"));
+        RemoveElementPredicate<String> matchNothing = new RemoveElementPredicate<String>() {
+
+            @Override
+            public boolean remove(String t) {
+                return false;
+            }
+            
+        };
+        
+        ListElementRemoval.removeIf(list, matchNothing);
+        
+        Assert.assertEquals(3, list.size());
+    }
+    
+    @Test
+    public void removeAll() {
+        List<String> list = new ArrayList<>(asList("a", "b", "c"));
+        RemoveElementPredicate<String> matchNothing = new RemoveElementPredicate<String>() {
+
+            @Override
+            public boolean remove(String t) {
+                return true;
+            }
+            
+        };
+        
+        ListElementRemoval.removeIf(list, matchNothing);
+        
+        Assert.assertEquals(0, list.size());
+    }
+    
+    @Test
+    public void removeSome() {
+        List<String> list = new ArrayList<>(asList("a", "b", "c"));
+        RemoveElementPredicate<String> matchNothing = new RemoveElementPredicate<String>() {
+
+            @Override
+            public boolean remove(String t) {
+                return "a".equals(t) || "c".equals(t);
+            }
+            
+        };
+        
+        ListElementRemoval.removeIf(list, matchNothing);
+        
+        Assert.assertEquals(1, list.size());
+        Assert.assertEquals("b", list.get(0));
+    }
+}