1.异步构建前缀数,不要占用spring的启动流程

2.前缀树需要有结束节点,否则判断会出问题
3.前缀数用map构建,提高匹配速度,不要用arraylist。
This commit is contained in:
zhaoyuhang
2023-06-18 11:42:11 +08:00
parent 09ca3698c4
commit 77cafd1d5a
2 changed files with 103 additions and 163 deletions

View File

@@ -2,39 +2,38 @@ package com.abin.mallchat.common.common.utils;
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.*;
import java.util.stream.Collectors;
/**
* 敏感词过滤
*
* @author zhaoyuhang
* @since 2023/06/11
*/
public final class SensitiveWordUtils {
private static SensitiveWordList wordList;
private static Map<Character, Word> wordMap; // 敏感词Map
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
' ', '!', '*', '-', '+', '_', '=', ',', '', '.', '@', ';', ':', '', ''
};
/**
* 敏感词
* 判断文本中是否存在敏感词
*
* @param text 文本
* @return boolean
* @return true: 存在敏感词, false: 不存在敏感词
*/
public static boolean hasSensitiveWord(String text) {
if (StringUtils.isBlank(text)) return false;
return !Objects.equals(filter(text), text);
}
/**
* 过滤敏感词并替换为指定字符
*
* @param text 待替换文本
* @return 替换后的文本
*/
/**
* 敏感词替换
*
@@ -42,57 +41,49 @@ 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 (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;
}
j++;
} else { // 字符串已结束,存在敏感词汇
flag = true;
break;
}
if (wordMap == null || wordMap.isEmpty() || StringUtils.isBlank(text)) return text;
char[] chars = text.toCharArray(); // 将文本转换为字符数组
int length = chars.length; // 文本长度
StringBuilder result = new StringBuilder(length); // 存储替换后的结果
int i = 0; // 当前遍历的字符索引
while (i < length) {
char c = chars[i]; // 当前字符
if (skip(c)) { // 如果是需要跳过的字符,则直接追加到结果中
i++;
continue;
}
int startIndex = i; // 敏感词匹配的起始索引
Map<Character, Word> currentMap = wordMap; // 当前层级的敏感词字典
int matchLength = 0; // 匹配到的敏感词长度
for (int j = i; j < length; j++) {
char ch = chars[j]; // 当前遍历的字符
if (skip(ch)) { // 如果是需要跳过的字符,则直接追加到结果中
continue;
}
if (word != null && word.next == null) {
flag = true;
Word word = currentMap.get(ch); // 获取当前字符在当前层级的敏感词字典中对应的敏感词节点
if (word == null) { // 如果未匹配到敏感词节点,则终止循环
break;
}
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--;
if (word.end) { // 如果当前节点是敏感词的最后一个节点,则记录匹配长度
matchLength = j - startIndex + 1;
}
currentMap = word.next; // 进入下一层级的敏感词字典
if (word.next == null) { // 如果当前节点是敏感词的最后一个节点,则记录匹配长度
matchLength = j - startIndex + 1;
}
}
if (matchLength > 0) { // 如果匹配到敏感词,则将对应的字符替换为指定替代字符
for (int j = startIndex; j < startIndex + matchLength; j++) {
chars[j] = replace;
}
}
i += matchLength > 0 ? matchLength : 1; // 更新当前索引,跳过匹配到的敏感词
}
return new String(__char__);
result.append(chars); // 将匹配到的敏感词追加到结果中
return result.toString();
}
/**
* 加载敏感词列表
*
@@ -101,31 +92,33 @@ 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;
wordMap = new HashMap<>(); // 创建敏感词字典的根节点
for (String word : words) {
if (word == null) continue;
char[] chars = word.toCharArray();
Map<Character, Word> currentMap = wordMap; // 当前层级的敏感词字典
for (int i = 0; i < chars.length; i++) {
char c = chars[i];
Word currentWord = currentMap.get(c);
if (currentWord == null) {
Word newWord = new Word(c); // 创建新的敏感词节点
currentMap.put(c, newWord); // 将节点添加到当前层级的敏感词字典中
if (i == chars.length - 1) {
newWord.end = true; // 添加结束标志
}
currentMap = newWord.next = new HashMap<>(); // 进入下一层级
} else {
currentMap = currentWord.next; // 存在该字符的节点,则进入下一层级
}
word = now.get(c);
if (word == null) word = now.add(c);
}
}
sort(wordList);
}
/**
* 加载敏感词txt文件每个敏感词独占一行不可出现空格空行逗号等非文字内容,必须使用UTF-8编码
* 从文本文件中加载敏感词列表
*
* @param path txt文件的绝对地址
* @param path 文本文件的绝对路径
*/
public static void loadWordFromFile(String path) {
String encoding = "UTF-8";
@@ -151,101 +144,39 @@ 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
* 判断是否需要跳过当前字符
*
* @param c 待检测字符
* @return true:需要跳过 false:不需要跳过
* @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;
for (char skipChar : skip) {
if (skipChar == c) return true;
}
return false;
}
/**
* 敏感词列表
*
* @author zhaoyuhang
* @since 2023/06/11
* 敏感词
*/
public static class SensitiveWordList extends ArrayList<Word> {
public Word get(char c) {
for (Word w : this) {
if (w.c == c) return w;
}
return null;
}
private static class Word {
// 当前字符
private char c;
/**
* 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;
}
// 结束标识
private boolean end;
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;
// 下一层级的敏感词字典
private Map<Character, Word> next;
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()) + ")";
}
public static void main(String[] args) {
List<String> strings = Arrays.asList("白日梦", "白痴", "白痴是你","TMD");
loadWord(strings);
System.out.println(filter("TMD,白痴是你吗"));
}
}

View File

@@ -4,8 +4,10 @@ import com.abin.mallchat.common.common.utils.SensitiveWordUtils;
import com.abin.mallchat.common.sensitive.dao.SensitiveWordDao;
import com.abin.mallchat.common.sensitive.domain.SensitiveWord;
import com.abin.mallchat.common.sensitive.service.ISensitiveWordService;
import lombok.extern.slf4j.Slf4j;
import org.apache.commons.collections.CollectionUtils;
import org.springframework.beans.factory.annotation.Autowired;
import org.springframework.scheduling.concurrent.ThreadPoolTaskExecutor;
import org.springframework.stereotype.Service;
import javax.annotation.PostConstruct;
@@ -13,18 +15,25 @@ import java.util.List;
import java.util.stream.Collectors;
@Service
@Slf4j
public class SensitiveWordServiceImpl implements ISensitiveWordService {
@Autowired
private SensitiveWordDao sensitiveWordDao;
@Autowired
private ThreadPoolTaskExecutor threadPoolTaskExecutor;
@PostConstruct
public void initSensitiveWord() {
List<SensitiveWord> list = sensitiveWordDao.list();
if (!CollectionUtils.isEmpty(list)) {
List<String> wordList = list.stream()
.map(SensitiveWord::getWord)
.collect(Collectors.toList());
SensitiveWordUtils.loadWord(wordList);
}
threadPoolTaskExecutor.execute(() -> {
log.info("[initSensitiveWord] start");
List<SensitiveWord> list = sensitiveWordDao.list();
if (!CollectionUtils.isEmpty(list)) {
List<String> wordList = list.stream()
.map(SensitiveWord::getWord)
.collect(Collectors.toList());
SensitiveWordUtils.loadWord(wordList);
}
log.info("[initSensitiveWord] end; loading sensitiveWords num:{}", list.size());
});
}
}