前缀树算法实现敏感词过滤

---

基于前缀树的敏感词过滤

知识点：

1. @PostConstruct
   有点像静态代码块的作用，在方法构造完成后将字符串插入前缀树。
2. resource下的文件会被加载到类路径下，通过class.getClasssLoader().getResrouceAsStream()可读取
3. 跳过特殊字符的匹配算法。

@Component
@Slf4j
public class SensitiveFilter {

    private TireNode root = new TireNode();
    private static final String REPLACE_WORD = "***";

    @PostConstruct
    public void init() {
        //加载类路径下的敏感词文件，加入到前缀树
        try(
            final InputStream is = this.getClass().getClassLoader().getResourceAsStream("sensitive-words.txt");
            BufferedReader br = new BufferedReader(new InputStreamReader(is));
        ) {
            String keyWord;
            while ((keyWord = br.readLine()) != null) {
                this.addKeyWord(keyWord);
            }
        } catch (IOException e) {
        	log.error("加载错误:{}", e.getMessage());
            e.printStackTrace();
        }
    }

    public String filter(String text) {
        if (StringUtils.isBlank(text)) return "";
        final char[] chs = text.toCharArray();
        TireNode cur = root;
        int end = 0, start = 0;
        StringBuilder sb = new StringBuilder();
        while (end < chs.length) {
            if (isSymbol(chs[end])) {
                //当前字符是特殊字符，选择跳过检查
                if (cur == root) {
                    //当前前缀树指针没动，需要把两个指针全部右移一位
                    start++;
                }
                //TODO:通过注释这句话可以让特殊字符不显示
                sb.append(chs[end]);
                //无论怎么样，end都会+1
                end++;
            } else {
            cur = cur.getNode(chs[end]);
              if (cur == null) {
                  //若当前字符在前缀树找不到一条路径,添加开始字符到res
                  //abc babc 若start指向b，end指向a，那么不符合前缀树路径abc，start就会右移一位，同时end紧跟着右移一位
                  sb.append(chs[start]);
                  end = ++start;
                  cur = root;
              } else if (cur.isKeyWordEnd()) {
                  //当前start到end的位置满足一个敏感字符
                  sb.append(REPLACE_WORD);
                  //start和end到达下一个位置
                  start = ++end;
                  cur = root;
              } else {
                  //还没跑完前缀树一条路线
                  end++;
              }
            }
        }
        //将最后一批字符加入res
        // 如abc abcae 此时start = d, end = e
        sb.append(text.substring(start));
        return sb.toString();
    }

    //返回是否为特殊字符
    //后两个范围是东亚字符
    private boolean isSymbol(char ch) {
        return !CharUtils.isAscii(ch) && (ch < 0x2E80 || ch > 0x9FFF);
    }

    /**
     *将keyword加入到前缀树中
     */
    private void addKeyWord(String keyWord) {
        final char[] chars = keyWord.toCharArray();
        TireNode node = null, cur = root;
        for (char ch : chars) {
            if ((node = cur.getNode(ch)) == null) {
                node = new TireNode();
                //新的节点放到下轮位置
                cur.addNode(ch, node);
            }
            cur = node;
        }
        if (node != null) {
            node.setKeyWordEnd(true);
        }
    }

    private class TireNode {
        @Getter
        @Setter
        private boolean isKeyWordEnd = false;
        private Map<Character, TireNode> nexts = new HashMap<>();

        public void addNode(char c, TireNode node) {
            nexts.put(c,node);
        }

        public TireNode getNode(char c) {
            return nexts.get(c);
        }
    }
}