feat;添加ac自动机算法进行敏感词过滤

mallchat-common/pom.xml

@@ -114,6 +114,14 @@
             <artifactId>redisson-spring-boot-starter</artifactId>
             <version>3.17.1</version>
         </dependency>
+
+        <!-- Used for unit testing -->
+        <dependency>
+            <groupId>junit</groupId>
+            <artifactId>junit</artifactId>
+            <version>${junit.version}</version>
+            <scope>test</scope>
+        </dependency>
     </dependencies>
     <build>
         <plugins>

mallchat-common/src/main/java/com/abin/mallchat/common/common/algorithm/ac/ACTrie.java

@@ -0,0 +1,95 @@
+package com.abin.mallchat.common.common.algorithm.ac;
+
+import com.google.common.collect.Lists;
+
+import java.util.*;
+import java.util.stream.Collectors;
+
+/**
+ * aho-corasick算法（又称AC自动机算法）
+ * Created by berg on 2023/6/18.
+ */
+public class ACTrie {
+
+    // 根节点
+    private ACTrieNode root;
+
+    public ACTrie(List<String> words) {
+        words = words.stream().distinct().collect(Collectors.toList()); // 去重
+        root = new ACTrieNode();
+        for (String word : words) {
+            addWord(word);
+        }
+        initTrieFailover();
+    }
+
+    public void addWord(String word) {
+        ACTrieNode walkNode = root;
+        char[] chars = word.toCharArray();
+        for (int i = 0; i < word.length(); i++) {
+            walkNode.addChildrenIfAbsent(chars[i]);
+            walkNode = walkNode.childOf(chars[i]);
+            walkNode.setDepth(i + 1);
+        }
+        walkNode.setLeaf(true);
+    }
+
+    public void initTrieFailover() {
+        //第一层的fail指针指向root
+        Queue<ACTrieNode> queue = new LinkedList<>();
+        Map<Character, ACTrieNode> childrens = root.getChildren();
+        for (ACTrieNode node : childrens.values()) {
+            node.setFailover(root);
+            queue.offer(node);
+        }
+        //构建剩余层数节点的fail指针,利用层次遍历
+        while (!queue.isEmpty()) {
+            ACTrieNode parentNode = queue.poll();
+            for (Map.Entry<Character, ACTrieNode> entry : parentNode.getChildren().entrySet()) {
+                ACTrieNode childNode = entry.getValue();
+                ACTrieNode failover = parentNode.getFailover();
+                // 在树中找到以childNode为结尾的字符串的最长前缀匹配，failover指向了这个最长前缀匹配的父节点
+                while (failover != null && (!failover.hasChild(entry.getKey()))) {
+                    failover = failover.getFailover();
+                }
+                //回溯到了root节点
+                if (failover == null) {
+                    childNode.setFailover(root);
+                } else {
+                    // 更新当前节点的回退指针
+                    childNode.setFailover(failover.childOf(entry.getKey()));
+                }
+                queue.offer(childNode);
+            }
+        }
+    }
+
+    /**
+     * 查询句子中包含的敏感词的起始位置和结束位置
+     *
+     * @param text
+     */
+    public List<MatchResult> matches(String text) {
+        List<MatchResult> result = Lists.newArrayList();
+        ACTrieNode walkNode = root;
+        for (int i = 0; i < text.length(); i++) {
+            char c = text.charAt(i);
+            while (!walkNode.hasChild(c) && walkNode.getFailover() != null) {
+                walkNode = walkNode.getFailover();
+            }
+            //如果因为当前节点的孩子节点有这个字符，则将temp替换为下面的孩子节点
+            if (walkNode.hasChild(c)) {
+                walkNode = walkNode.childOf(c);
+                // 检索到了敏感词
+                if (walkNode.isLeaf()) {
+                    result.add(new MatchResult(i - walkNode.getDepth() + 1, i + 1));
+                    // 模式串回退到最长可匹配前缀位置并开启新一轮的匹配
+                    // 这种回退方式将一个不漏的匹配到所有的敏感词，匹配结果的区间可能会有重叠的部分
+                    walkNode = walkNode.getFailover();
+                }
+            }
+        }
+        return result;
+    }
+
+}

mallchat-common/src/main/java/com/abin/mallchat/common/common/algorithm/ac/ACTrieNode.java

@@ -0,0 +1,46 @@
+package com.abin.mallchat.common.common.algorithm.ac;
+
+import com.google.common.collect.Maps;
+import lombok.Getter;
+import lombok.Setter;
+
+import java.util.Map;
+
+/**
+ * Created by berg on 2023/6/18.
+ */
+@Getter
+@Setter
+public class ACTrieNode {
+
+    // 子节点
+    private Map<Character, ACTrieNode> children = Maps.newHashMap();
+
+    // 匹配过程中，如果模式串不匹配，模式串指针会回退到failover继续进行匹配
+    private ACTrieNode failover = null;
+
+    private int depth;
+
+    private boolean isLeaf = false;
+
+    public void addChildrenIfAbsent(char c) {
+        children.computeIfAbsent(c, (key) -> new ACTrieNode());
+    }
+
+    public ACTrieNode childOf(char c) {
+        return children.get(c);
+    }
+
+    public boolean hasChild(char c) {
+        return children.containsKey(c);
+    }
+
+    @Override
+    public String toString() {
+        return "ACTrieNode{" +
+                "failover=" + failover +
+                ", depth=" + depth +
+                ", isLeaf=" + isLeaf +
+                '}';
+    }
+}

mallchat-common/src/main/java/com/abin/mallchat/common/common/algorithm/ac/MatchResult.java

@@ -0,0 +1,24 @@
+package com.abin.mallchat.common.common.algorithm.ac;
+
+import lombok.*;
+
+/**
+ * Created by berg on 2023/6/18.
+ */
+@Getter
+@Setter
+@AllArgsConstructor
+public class MatchResult {
+
+    private int startIndex;
+
+    private int endIndex;
+
+    @Override
+    public String toString() {
+        return "MatchResult{" +
+                "startIndex=" + startIndex +
+                ", endIndex=" + endIndex +
+                '}';
+    }
+}

mallchat-common/src/main/java/com/abin/mallchat/common/common/algorithm/package-info.java

@@ -0,0 +1,4 @@
+/**
+ * Created by berg on 2023/6/18.
+ */
+package com.abin.mallchat.common.common.algorithm;

mallchat-common/src/main/java/com/abin/mallchat/common/common/utils/SensitiveWordUtils.java

@@ -1,14 +1,17 @@
 package com.abin.mallchat.common.common.utils;
 
+import com.abin.mallchat.common.common.algorithm.ac.ACTrie;
+import com.abin.mallchat.common.common.algorithm.ac.MatchResult;
 import org.apache.commons.lang3.StringUtils;
 
-import java.io.*;
+import java.io.BufferedReader;
+import java.io.File;
+import java.io.IOException;
+import java.io.InputStreamReader;
 import java.nio.file.Files;
 import java.util.ArrayList;
-import java.util.Collections;
 import java.util.List;
 import java.util.Objects;
-import java.util.stream.Collectors;
 
 
 /**
@@ -18,11 +21,9 @@ import java.util.stream.Collectors;
  * @since 2023/06/11
  */
 public final class SensitiveWordUtils {
-    private static SensitiveWordList wordList;
-    private final static char replace = '*'; // 替代字符
-    private final static char[] skip = new char[]{ // 遇到这些字符就会跳过
-            ' ', '!', '*', '-', '+', '_', '=', ',', '，', '.', '@', ';', ':', '；', '：'
-    };
+    private final static char mask_char = '*'; // 替代字符
+
+    private static ACTrie ac_trie = null;
 
     /**
      * 有敏感词
@@ -42,56 +43,20 @@ public final class SensitiveWordUtils {
      * @return 替换后的文本
      */
     public static String filter(String text) {
-        if (wordList == null || wordList.size() == 0 || StringUtils.isBlank(text)) return text;
-        char[] __char__ = text.toCharArray(); // 把String转化成char数组，便于遍历
-        int i, j;
-        Word word;
-        boolean flag; // 是否需要替换
-        for (i = 0; i < __char__.length; i++) { // 遍历所有字符
-            char c = __char__[i];
-            word = wordList.binaryGet(c); // 使用二分查找来寻找字符，提高效率
-            if (word != null) { // word != null说明找到了
-                flag = false;
-                j = i + 1;
-                while (j < __char__.length) { // 开始逐个比较后面的字符
-                    if (skip(__char__[j])) { // 跳过空格之类的无关字符
-                        j++;
-                        continue;
+        if (StringUtils.isBlank(text)) return text;
+        List<MatchResult> matchResults = ac_trie.matches(text);
+        StringBuffer result = new StringBuffer(text);
+        for (MatchResult matchResult : matchResults) {
+            replaceBetween(result, matchResult.getStartIndex(), matchResult.getEndIndex());
         }
-                    if (word.next != null) { // 字符串尚未结束，不确定是否存在敏感词
-                        /*
-                        以下代码并没有使用二分查找，因为以同一个字符开头的敏感词较少
-                        例如，wordList中记录了所有敏感词的开头第一个字，它的数量通常会有上千个
-                        假如现在锁定了字符“T”开头的敏感词，而“T”开头的敏感词只有10个，这时使用二分查找的效率反而低于顺序查找
-                         */
-                        word = word.next.get(__char__[j]);
-                        if (word == null) {
-                            break;
+        return result.toString();
     }
-                        j++;
-                    } else { // 字符串已结束，存在敏感词汇
-                        flag = true;
-                        break;
+
+    private static void replaceBetween(StringBuffer buffer, int startIndex, int endIndex) {
+        for (int i = startIndex; i < endIndex; i++) {
+            buffer.setCharAt(i, mask_char);
         }
     }
-                if (word != null && word.next == null) {
-                    flag = true;
-                }
-                if (flag) { // 如果flag==true，说明检测出敏感粗，需要替换
-                    while (i < j) {
-                        // if(skip(__char__[i])){ // 跳过空格之类的无关字符，如果要把空格也替换成'*'，则删除这个if语句
-                        //     i++;
-                        //     continue;
-                        // }
-                        __char__[i] = replace;
-                        i++;
-                    }
-                    i--;
-                }
-            }
-        }
-        return new String(__char__);
-    }
 
     /**
      * 加载敏感词列表
@@ -100,26 +65,7 @@ public final class SensitiveWordUtils {
      */
     public static void loadWord(List<String> words) {
         if (words == null) return;
-        words = words.stream().distinct().collect(Collectors.toList()); // 去重
-        char[] chars;
-        SensitiveWordList now;
-        Word word;
-        wordList = new SensitiveWordList();
-        for (String __word__ : words) {
-            if (__word__ == null) continue;
-            chars = __word__.toCharArray();
-            now = wordList;
-            word = null;
-            for (char c : chars) {
-                if (word != null) {
-                    if (word.next == null) word.next = new SensitiveWordList();
-                    now = word.next;
-                }
-                word = now.get(c);
-                if (word == null) word = now.add(c);
-            }
-        }
-        sort(wordList);
+        ac_trie = new ACTrie(words);
     }
 
     /**
@@ -150,102 +96,8 @@ public final class SensitiveWordUtils {
         }
     }
 
-    /**
-     * 对敏感词多叉树递增排序
-     *
-     * @param list 待排序List
-     */
-    private static void sort(SensitiveWordList list) {
-        if (list == null) return;
-        Collections.sort(list); // 递增排序
-        for (Word word : list) {
-            sort(word.next);
-        }
-    }
 
-    /**
-     * 判断是否跳过当前字符
-     *
-     * @param c 待检测字符
-     * @return true:需要跳过   false:不需要跳过
-     */
-    private static boolean skip(char c) {
-        for (char c1 : skip) {
-            if (c1 == c) return true;
-        }
-        return false;
-    }
-
-    /**
-     * 敏感词列表
-     *
-     * @author zhaoyuhang
-     * @since 2023/06/11
-     */
-    public static class SensitiveWordList extends ArrayList<Word> {
-        public Word get(char c) {
-            for (Word w : this) {
-                if (w.c == c) return w;
-            }
-            return null;
-        }
-
-        /**
-         * 二分查找，必须先升序排序
-         *
-         * @param c 需要查找的字符
-         * @return Word对象：如果找到   null:如果没找到
-         */
-        public Word binaryGet(char c) {
-            int left, right, key;
-            Word word;
-            left = 0;
-            right = this.size() - 1;
-            while (left <= right) {
-                key = (left + right) / 2;
-                word = get(key);
-                if (word.c == c) {
-                    return word;
-                } else if (word.c > c) {
-                    right = key - 1;
-                } else {
-                    left = key + 1;
-                }
-            }
-            return null;
-        }
-
-        public Word add(char c) {
-            Word word = new Word(c);
-            super.add(word);
-            return word;
-        }
 
 }
 
-    /**
-     * 敏感词
-     *
-     * @author zhaoyuhang
-     * @since 2023/06/11
-     */
-    public static class Word implements Comparable<Word> {
-        public char c;
-        public SensitiveWordList next = null;
-
-        public Word(char c) {
-            this.c = c;
-        }
-
-        @Override
-        public int compareTo(Word word) {
-            return c - word.c;
-        }
-
-        public String toString() {
-            return c + "(" + (next == null ? null : next.size()) + ")";
-        }
-    }
-}
-
 

mallchat-common/src/test/java/com/abin/mallchat/common/common/algorithm/ac/ACTrieTest.java

@@ -0,0 +1,52 @@
+package com.abin.mallchat.common.common.algorithm.ac;
+
+import com.google.common.collect.Lists;
+import org.junit.Test;
+
+import java.util.List;
+
+import static org.junit.Assert.assertEquals;
+
+/**
+ * Created by berg on 2023/6/18.
+ */
+public class ACTrieTest {
+
+    private final static List<String> ALPHABET = Lists.newArrayList("abc", "bcd", "cde");
+
+    private static ACTrie trie(List<String> keywords) {
+        return new ACTrie(keywords);
+    }
+
+    @Test
+    public void test_TextIsLongerThanKeyword() {
+        final ACTrie trie = trie(ALPHABET);
+        final String text = " " + ALPHABET.get(0);
+        List<MatchResult> matchResults = trie.matches(text);
+        checkResult(matchResults.get(0), 1, 4, ALPHABET.get(0), text);
+    }
+
+    @Test
+    public void test_VariousKeywordsOneMatch() {
+        final ACTrie trie = trie(ALPHABET);
+        final String text = "bcd";
+        List<MatchResult> matchResults = trie.matches(text);
+        checkResult(matchResults.get(0), 0, 3, ALPHABET.get(1), text);
+    }
+
+    @Test
+    public void test_VariousKeywordsMultiMatch() {
+        final ACTrie trie = trie(ALPHABET);
+        final String text = "abcd";
+        List<MatchResult> matchResults = trie.matches(text);
+        assertEquals(2, matchResults.size());
+        checkResult(matchResults.get(0), 0, 3, ALPHABET.get(0), text);
+        checkResult(matchResults.get(1), 1, 4, ALPHABET.get(1), text);
+    }
+
+    private void checkResult(MatchResult matchResult, int expectedStart, int expectedEnd, String expectedKeyword, String text) {
+        assertEquals("Start of match should have been " + expectedStart, expectedStart, matchResult.getStartIndex());
+        assertEquals("End of match should have been " + expectedEnd, expectedEnd, matchResult.getEndIndex());
+        assertEquals(expectedKeyword, text.substring(expectedStart, expectedEnd));
+    }
+}