解决上次优化索引分词带来的问题：hankcs#513

Author: hankcs

src/main/java/com/hankcs/hanlp/seg/Segment.java

@@ -274,7 +274,7 @@ protected static List<Vertex> combineByCustomDictionary(List<Vertex> vertexList,
         vertexList.toArray(wordNet);
         // DAT合并
         DoubleArrayTrie<CoreDictionary.Attribute> dat = CustomDictionary.dat;
-        for (int i = 0; i < wordNet.length; ++i)
+        for (int i = 0, line = 0; i < wordNet.length; ++i)
         {
             int state = 1;
             state = dat.transition(wordNet[i].realWord, state);
@@ -297,15 +297,24 @@ protected static List<Vertex> combineByCustomDictionary(List<Vertex> vertexList,
                 if (value != null)
                 {
                     combineWords(wordNet, i, end, value);
-                    wordNetAll.add(i, wordNet[i]);
+                    wordNetAll.add(line, wordNet[i]);
+                    line += wordNet[i].realWord.length();
                     i = end - 1;
                 }
+                else
+                {
+                    line += wordNet[i].realWord.length();
+                }
+            }
+            else
+            {
+                line += wordNet[i].realWord.length();
             }
         }
         // BinTrie合并
         if (CustomDictionary.trie != null)
         {
-            for (int i = 0; i < wordNet.length; ++i)
+            for (int i = 0, line = 0; i < wordNet.length; ++i)
             {
                 if (wordNet[i] == null) continue;
                 BaseNode<CoreDictionary.Attribute> state = CustomDictionary.trie.transition(wordNet[i].realWord.toCharArray(), 0);
@@ -328,9 +337,18 @@ protected static List<Vertex> combineByCustomDictionary(List<Vertex> vertexList,
                     if (value != null)
                     {
                         combineWords(wordNet, i, end, value);
-                        wordNetAll.add(i, wordNet[i]);
+                        wordNetAll.add(line, wordNet[i]);
+                        line += wordNet[i].realWord.length();
                         i = end - 1;
                     }
+                    else
+                    {
+                        line += wordNet[i].realWord.length();
+                    }
+                }
+                else
+                {
+                    line += wordNet[i].realWord.length();
                 }
             }
         }

src/main/java/com/hankcs/hanlp/seg/WordBasedGenerativeModelSegment.java

@@ -494,7 +494,9 @@ protected static List<Term> decorateResultForIndexMode(List<Vertex> vertexList,
                         if (
                                 ((termMain.nature == Nature.mq && smallVertex.hasNature(Nature.q)) ||
                                         smallVertex.realWord.length() > 1)
-                                        && smallVertex != vertex)
+                                        && smallVertex != vertex // 防止重复添加
+                                        && currentLine + smallVertex.realWord.length() <= line + vertex.realWord.length() // 防止超出边界
+                            )
                         {
                             listIterator.add(smallVertex);
                             Term termSub = convert(smallVertex);

src/test/java/com/hankcs/test/seg/TestSegment.java

@@ -406,4 +406,13 @@ public void testIssue496() throws Exception
         System.out.println(segment.seg("中医药"));
         System.out.println(segment.seg("中医药大学"));
     }
+
+    public void testIssue513() throws Exception
+    {
+        List<Term> termList = IndexTokenizer.segment("南京市长江大桥");
+        for (Term term : termList)
+        {
+            System.out.println(term + " [" + term.offset + ":" + (term.offset + term.word.length()) + "]");
+        }
+    }
 }