@nebulaaa/auditor 1.0.0 → 1.0.1

package/dist/auditor.cjs

@@ -16,6 +16,7 @@ class ACAutomaton {
         this._built = false;
         this._interferenceWords = new Set();
         this._ignoreCase = ignoreCase;
+        this._wordChars = new Map();
     }
     /**
      * 大小写归一（仅 ASCII 字母，逐码点等长替换，保证匹配位置与原始文本索引一致）
@@ -54,11 +55,13 @@ class ACAutomaton {
             }
             current = child;
         }
-        if (!current.Words.includes(word)) {
-            // 同词重复添加幂等：自身词独立存放，output 由 build 重新生成
+        // 归一后同一路径只保留首词：IgnoreCase 下 "abc"/"ABC" 共享路径，
+        // 变体重复入表会让 Match 对同一区间输出多条重复结果（下游命中计数虚高）
+        if (current.Words.length === 0) {
             current.Words.push(word);
+            this._wordChars.set(word, chars);
+            this._built = false; // 标记需要重新构建 fail 指针
         }
-        this._built = false; // 标记需要重新构建 fail 指针
     }
     /**
      * 构建 fail 指针（使用 BFS）
@@ -121,8 +124,8 @@ class ACAutomaton {
                 for (const word of current.Output) {
                     // 计算在原始文本中的起始位置（跳过干扰字符）
                     let j = i;
-                    // 从当前位置向前查找匹配的起始位置
-                    const wordChars = Array.from(this.normalize(word));
+                    // 从当前位置向前查找匹配的起始位置（AddWord 已缓存归一后的码点数组）
+                    const wordChars = this._wordChars.get(word);
                     let wordIndex = wordChars.length - 1;
                     // 从后向前匹配，找到原始文本中的位置
                     while (j >= 0 && wordIndex >= 0) {
@@ -135,6 +138,11 @@ class ACAutomaton {
                         }
                         j--;
                     }
+                    // 回扫必须完整命中：前向 AC 已保证该词是当前位置的后缀（两侧对称跳过干扰字符），
+                    // wordIndex >= 0 说明不变量被破坏（理论不可达），防御性跳过而非产出错误区间
+                    if (wordIndex >= 0) {
+                        continue;
+                    }
                     results.push({
                         Word: word,
                         Start: j + 1,
@@ -184,25 +192,29 @@ class ACAutomaton {
             Words: [],
             Output: [],
         };
+        this._wordChars.clear();
         this._built = false;
     }
     /**
      * 注入干扰词（仅单字符生效：匹配按单字符查表）
      * @param words 干扰词列表
+     * @returns 被忽略的多字符词（由调用方决定告警方式，保持算法类无日志副作用）
      */
     AddInterferenceWords(words) {
+        const ignored = [];
         for (const word of words) {
             if (!word || word.length === 0) {
                 continue;
             }
             const chars = Array.from(this.normalize(word));
             if (chars.length !== 1) {
-                // 多字符干扰词在按字符匹配的查表中永不命中——显式告警而非静默失效
-                console.warn(`addInterferenceWords: 干扰词须为单个字符，已忽略 "${word}"`);
+                // 多字符干扰词在按字符匹配的查表中永不命中——返回给调用方显式告警而非静默失效
+                ignored.push(word);
                 continue;
             }
             this._interferenceWords.add(chars[0]);
         }
+        return ignored;
     }
     /**
      * 清空干扰词
@@ -229,14 +241,14 @@ class AuditorPlugin extends plugin.Plugin {
             CacheSize: 1000,
         });
         super(deps, cfg);
+        // 缓存容量须为正整数：无论是否启用缓存都在构造期校验，配置错误不留到运行期
+        if (!Number.isInteger(cfg.CacheSize) || cfg.CacheSize <= 0) {
+            throw new Error(`AuditorPlugin CacheSize 须为正整数 >>> ${cfg.CacheSize}`);
+        }
         this._acAutomaton = new ACAutomaton(cfg.IgnoreCase);
         this._sensitiveWords = new Set();
         this._acAutomaton.AddInterferenceWords(' \t\r\n'.split(''));
         if (cfg.EnableCache) {
-            // 缓存容量须为正整数：非法配置构造期显式失败，错误信息带插件上下文便于定位
-            if (!Number.isInteger(cfg.CacheSize) || cfg.CacheSize <= 0) {
-                throw new Error(`AuditorPlugin CacheSize 须为正整数 >>> ${cfg.CacheSize}`);
-            }
             this._testCache = new lru.LRU(cfg.CacheSize);
             this._findCache = new lru.LRU(cfg.CacheSize);
             this._replaceCache = new lru.LRU(cfg.CacheSize);
@@ -353,16 +365,29 @@ class AuditorPlugin extends plugin.Plugin {
         return replacedText;
     }
     InjectSensitiveWords(words) {
+        let added = false;
         for (const word of words) {
-            if (word && word.length > 0) {
+            if (word && word.length > 0 && !this._sensitiveWords.has(word)) {
                 this._sensitiveWords.add(word);
+                added = true;
             }
         }
+        // 无新增词时词库未变：跳过自动机重建与缓存清理，重复注入不打断既有缓存
+        if (!added) {
+            return;
+        }
         this.buildAutomaton();
         this.clearCache();
     }
     InjectInterferenceWords(words) {
-        this._acAutomaton.AddInterferenceWords(words);
+        const ignored = this._acAutomaton.AddInterferenceWords(words);
+        // 多字符干扰词永不命中：经插件日志显式告警（词库内容不落 console）
+        if (ignored.length > 0) {
+            const logger = this.Logger;
+            for (const word of ignored) {
+                logger.Warn(`干扰词须为单个字符，已忽略 >>> ${word}`);
+            }
+        }
         this.clearCache();
     }
     GetSensitiveWordCount() {

package/dist/auditor.d.ts

@@ -18,7 +18,7 @@ interface AuditorCfg extends PluginCfg {
     /**
      * 缓存大小
      * @default 1000
-     * @tip 须为正整数，非法值构造期显式报错
+     * @tip 须为正整数，非法值构造期显式报错（无论是否启用缓存均校验）
      */
     readonly CacheSize?: number;
     /**
@@ -33,9 +33,15 @@ interface AuditorCfg extends PluginCfg {
 interface IAuditorMatchPosition {
     /** 匹配到的敏感词 */
     Word: string;
-    /** 起始位置（字符索引） */
+    /**
+     * 起始位置（**码点索引**，`Array.from` 语义）
+     * @tip 非 UTF-16 下标：文本含 emoji 等代理对时，直接 `str.slice(Start, End)` 会切错位置
+     */
     Start: number;
-    /** 结束位置（字符索引，不包含） */
+    /**
+     * 结束位置（码点索引，不包含）
+     * @tip 同上，切分须按码点进行（如 `Array.from(str).slice(Start, End)`）
+     */
     End: number;
 }
 /** 匹配结果 */
@@ -58,15 +64,8 @@ interface IACNode {
     /** 匹配输出（自身词 + fail 链词，构建时由 words 重新生成） */
     Output: string[];
 }
-/** AC 自动机匹配结果 */
-interface IACMatchResult {
-    /** 匹配到的敏感词 */
-    Word: string;
-    /** 起始位置 */
-    Start: number;
-    /** 结束位置 */
-    End: number;
-}
+/** AC 自动机匹配结果（与外部匹配位置结构一致，复用同一形状避免重复定义） */
+type IACMatchResult = IAuditorMatchPosition;
 /** 敏感词审计员插件接口 */
 interface IAuditorPlugin extends IPlugin<AuditorDeps, AuditorCfg> {
     /**
@@ -78,7 +77,7 @@ interface IAuditorPlugin extends IPlugin<AuditorDeps, AuditorCfg> {
     /**
      * 查找文本中的所有敏感词
      * @param text 待查找的文本
-     * @returns 匹配结果
+     * @returns 匹配结果（匹配位置为码点索引）
      */
     Find(text: string): IAuditorMatchResult;
     /**
@@ -94,6 +93,7 @@ interface IAuditorPlugin extends IPlugin<AuditorDeps, AuditorCfg> {
     Replace(text: string, replacement?: string): string;
     /**
      * 注入敏感词
+     * @tip 重复注入已存在的词为幂等操作，不触发自动机重建与缓存清理
      * @param words 敏感词列表
      */
     InjectSensitiveWords(words: string[]): void;
@@ -102,6 +102,8 @@ interface IAuditorPlugin extends IPlugin<AuditorDeps, AuditorCfg> {
      *
      * - 干扰词在匹配时被跳过（如 "敏 感 词" 仍命中），仅单字符
      *   生效，多字符词将被忽略并告警
+     * - 构造期默认注入空白字符（空格/制表符/回车/换行），故全新实例的
+     *   GetInterferenceWordCount 返回 4
      * - 敏感词本身包含干扰字符时该词将无法命中，注入时须避开词库字符
      * @param words 干扰词列表
      */
@@ -113,12 +115,15 @@ interface IAuditorPlugin extends IPlugin<AuditorDeps, AuditorCfg> {
     GetSensitiveWordCount(): number;
     /**
      * 获取当前干扰词数量
-     * @returns 干扰词数量
+     * @returns 干扰词数量（含构造期默认注入的空白字符）
      */
     GetInterferenceWordCount(): number;
     /** 清空所有敏感词 */
     ClearSensitiveWords(): void;
-    /** 清空所有干扰词 */
+    /**
+     * 清空所有干扰词（含构造期默认注入的空白字符）
+     * @tip 清空后 "敏 感 词" 这类夹空白的变体不再命中
+     */
     ClearInterferenceWords(): void;
 }
 

package/dist/auditor.min.cjs

@@ -1 +1 @@
-"use strict";var t=require("@nebulaaa/dict"),e=require("@nebulaaa/lru"),r=require("@nebulaaa/plugin");class n{constructor(t=!1){this._root={Children:new Map,Fail:null,Words:[],Output:[]},this._built=!1,this._interferenceWords=new Set,this._ignoreCase=t}normalize(t){if(!this._ignoreCase)return t;let e="";for(const r of t){const t=r.codePointAt(0);e+=t>=65&&t<=90?String.fromCharCode(t+32):r}return e}AddWord(t){if(!t||0===t.length)return;let e=this._root;const r=Array.from(this.normalize(t));for(const t of r){let r=e.Children.get(t);r||(r={Children:new Map,Fail:null,Words:[],Output:[]},e.Children.set(t,r)),e=r}e.Words.includes(t)||e.Words.push(t),this._built=!1}Build(){if(this._built)return;const t=[];for(const e of this._root.Children.values())e.Fail=this._root,e.Output=[...e.Words],t.push(e);let e=0;for(;e<t.length;){const r=t[e++];for(const[e,n]of r.Children){t.push(n);let s=r.Fail;for(;s&&!s.Children.has(e);)s=s.Fail;n.Fail=s&&s.Children.get(e)||this._root,n.Output=[...n.Words,...n.Fail.Output]}}this._built=!0}Match(t){this._built||this.Build();const e=[];let r=this._root;const n=Array.from(this.normalize(t));for(let t=0;t<n.length;t++){const s=n[t];if(!this._interferenceWords.has(s)){for(;r!==this._root&&!r.Children.has(s);)r=r.Fail;if(r.Children.has(s)){r=r.Children.get(s);for(const s of r.Output){let r=t;const i=Array.from(this.normalize(s));let o=i.length-1;for(;r>=0&&o>=0;)this._interferenceWords.has(n[r])||n[r]===i[o]&&o--,r--;e.push({Word:s,Start:r+1,End:t+1})}}}}return e}Test(t){this._built||this.Build();let e=this._root;const r=Array.from(this.normalize(t));for(let t=0;t<r.length;t++){const n=r[t];if(!this._interferenceWords.has(n)){for(;e!==this._root&&!e.Children.has(n);)e=e.Fail;if(e.Children.has(n)&&(e=e.Children.get(n),e.Output.length>0))return!0}}return!1}Clear(){this._root={Children:new Map,Fail:null,Words:[],Output:[]},this._built=!1}AddInterferenceWords(t){for(const e of t){if(!e||0===e.length)continue;const t=Array.from(this.normalize(e));1===t.length?this._interferenceWords.add(t[0]):console.warn(`addInterferenceWords: 干扰词须为单个字符，已忽略 "${e}"`)}}ClearInterferenceWords(){this._interferenceWords.clear()}GetInterferenceWordCount(){return this._interferenceWords.size}}class s extends r.Plugin{constructor(r,s){if(t.Override(s,{Replacement:"█",IgnoreCase:!1,CacheSize:1e3}),super(r,s),this._acAutomaton=new n(s.IgnoreCase),this._sensitiveWords=new Set,this._acAutomaton.AddInterferenceWords(" \t\r\n".split("")),s.EnableCache){if(!Number.isInteger(s.CacheSize)||s.CacheSize<=0)throw new Error(`AuditorPlugin CacheSize 须为正整数 >>> ${s.CacheSize}`);this._testCache=new e.LRU(s.CacheSize),this._findCache=new e.LRU(s.CacheSize),this._replaceCache=new e.LRU(s.CacheSize)}else this._testCache=null,this._findCache=null,this._replaceCache=null;this.buildAutomaton()}Test(t){if(!t||0===t.length)return!1;if(this._testCache){const e=this._testCache.Get(t);if(null!==e)return e}const e=this._acAutomaton.Test(t);return this._testCache&&this._testCache.Set(t,e),e}Find(t){if(!t||0===t.length)return{Matched:!1,Words:[],Matches:[]};if(this._findCache){const e=this._findCache.Get(t);if(null!==e)return{Matched:e.Matched,Words:[...e.Words],Matches:e.Matches.map(t=>Object.assign({},t))}}const e=this._acAutomaton.Match(t),r=new Set;for(const t of e)r.add(t.Word);const n={Matched:e.length>0,Words:Array.from(r),Matches:e};return this._findCache&&this._findCache.Set(t,{Matched:n.Matched,Words:[...n.Words],Matches:n.Matches.map(t=>Object.assign({},t))}),n}Replace(t,e){var r;if(!t||0===t.length)return t;const n=JSON.stringify([t,null!=e?e:this.Cfg.Replacement]);if(this._replaceCache){const t=this._replaceCache.Get(n);if(null!==t)return t}const s=this.Find(t);if(!s.Matched)return null===(r=this._replaceCache)||void 0===r||r.Set(n,t),t;const i=Array.from(t),o=[...s.Matches].sort((t,e)=>t.Start-e.Start),h=[];for(const t of o){const e=h[h.length-1];e&&t.Start<e.end?e.end=Math.max(e.end,t.End):h.push({start:t.Start,end:t.End})}const a=null!=e?e:this.Cfg.Replacement;for(let t=h.length-1;t>=0;t--){const{start:e,end:r}=h[t];i.splice(e,r-e,a)}const c=i.join("");return this._replaceCache&&this._replaceCache.Set(n,c),c}InjectSensitiveWords(t){for(const e of t)e&&e.length>0&&this._sensitiveWords.add(e);this.buildAutomaton(),this.clearCache()}InjectInterferenceWords(t){this._acAutomaton.AddInterferenceWords(t),this.clearCache()}GetSensitiveWordCount(){return this._sensitiveWords.size}GetInterferenceWordCount(){return this._acAutomaton.GetInterferenceWordCount()}ClearSensitiveWords(){this._sensitiveWords.clear(),this.buildAutomaton(),this.clearCache()}ClearInterferenceWords(){this._acAutomaton.ClearInterferenceWords(),this.clearCache()}buildAutomaton(){this._acAutomaton.Clear();const t=this._sensitiveWords;for(const e of t)this._acAutomaton.AddWord(e);this._acAutomaton.Build()}clearCache(){var t,e,r;null===(t=this._testCache)||void 0===t||t.Clear(),null===(e=this._findCache)||void 0===e||e.Clear(),null===(r=this._replaceCache)||void 0===r||r.Clear()}}exports.AuditorPlugin=s;
+"use strict";var t=require("@nebulaaa/dict"),e=require("@nebulaaa/lru"),r=require("@nebulaaa/plugin");class s{constructor(t=!1){this._root={Children:new Map,Fail:null,Words:[],Output:[]},this._built=!1,this._interferenceWords=new Set,this._ignoreCase=t,this._wordChars=new Map}normalize(t){if(!this._ignoreCase)return t;let e="";for(const r of t){const t=r.codePointAt(0);e+=t>=65&&t<=90?String.fromCharCode(t+32):r}return e}AddWord(t){if(!t||0===t.length)return;let e=this._root;const r=Array.from(this.normalize(t));for(const t of r){let r=e.Children.get(t);r||(r={Children:new Map,Fail:null,Words:[],Output:[]},e.Children.set(t,r)),e=r}0===e.Words.length&&(e.Words.push(t),this._wordChars.set(t,r),this._built=!1)}Build(){if(this._built)return;const t=[];for(const e of this._root.Children.values())e.Fail=this._root,e.Output=[...e.Words],t.push(e);let e=0;for(;e<t.length;){const r=t[e++];for(const[e,s]of r.Children){t.push(s);let n=r.Fail;for(;n&&!n.Children.has(e);)n=n.Fail;s.Fail=n&&n.Children.get(e)||this._root,s.Output=[...s.Words,...s.Fail.Output]}}this._built=!0}Match(t){this._built||this.Build();const e=[];let r=this._root;const s=Array.from(this.normalize(t));for(let t=0;t<s.length;t++){const n=s[t];if(!this._interferenceWords.has(n)){for(;r!==this._root&&!r.Children.has(n);)r=r.Fail;if(r.Children.has(n)){r=r.Children.get(n);for(const n of r.Output){let r=t;const i=this._wordChars.get(n);let o=i.length-1;for(;r>=0&&o>=0;)this._interferenceWords.has(s[r])||s[r]===i[o]&&o--,r--;o>=0||e.push({Word:n,Start:r+1,End:t+1})}}}}return e}Test(t){this._built||this.Build();let e=this._root;const r=Array.from(this.normalize(t));for(let t=0;t<r.length;t++){const s=r[t];if(!this._interferenceWords.has(s)){for(;e!==this._root&&!e.Children.has(s);)e=e.Fail;if(e.Children.has(s)&&(e=e.Children.get(s),e.Output.length>0))return!0}}return!1}Clear(){this._root={Children:new Map,Fail:null,Words:[],Output:[]},this._wordChars.clear(),this._built=!1}AddInterferenceWords(t){const e=[];for(const r of t){if(!r||0===r.length)continue;const t=Array.from(this.normalize(r));1===t.length?this._interferenceWords.add(t[0]):e.push(r)}return e}ClearInterferenceWords(){this._interferenceWords.clear()}GetInterferenceWordCount(){return this._interferenceWords.size}}class n extends r.Plugin{constructor(r,n){if(t.Override(n,{Replacement:"█",IgnoreCase:!1,CacheSize:1e3}),super(r,n),!Number.isInteger(n.CacheSize)||n.CacheSize<=0)throw new Error(`AuditorPlugin CacheSize 须为正整数 >>> ${n.CacheSize}`);this._acAutomaton=new s(n.IgnoreCase),this._sensitiveWords=new Set,this._acAutomaton.AddInterferenceWords(" \t\r\n".split("")),n.EnableCache?(this._testCache=new e.LRU(n.CacheSize),this._findCache=new e.LRU(n.CacheSize),this._replaceCache=new e.LRU(n.CacheSize)):(this._testCache=null,this._findCache=null,this._replaceCache=null),this.buildAutomaton()}Test(t){if(!t||0===t.length)return!1;if(this._testCache){const e=this._testCache.Get(t);if(null!==e)return e}const e=this._acAutomaton.Test(t);return this._testCache&&this._testCache.Set(t,e),e}Find(t){if(!t||0===t.length)return{Matched:!1,Words:[],Matches:[]};if(this._findCache){const e=this._findCache.Get(t);if(null!==e)return{Matched:e.Matched,Words:[...e.Words],Matches:e.Matches.map(t=>Object.assign({},t))}}const e=this._acAutomaton.Match(t),r=new Set;for(const t of e)r.add(t.Word);const s={Matched:e.length>0,Words:Array.from(r),Matches:e};return this._findCache&&this._findCache.Set(t,{Matched:s.Matched,Words:[...s.Words],Matches:s.Matches.map(t=>Object.assign({},t))}),s}Replace(t,e){var r;if(!t||0===t.length)return t;const s=JSON.stringify([t,null!=e?e:this.Cfg.Replacement]);if(this._replaceCache){const t=this._replaceCache.Get(s);if(null!==t)return t}const n=this.Find(t);if(!n.Matched)return null===(r=this._replaceCache)||void 0===r||r.Set(s,t),t;const i=Array.from(t),o=[...n.Matches].sort((t,e)=>t.Start-e.Start),h=[];for(const t of o){const e=h[h.length-1];e&&t.Start<e.end?e.end=Math.max(e.end,t.End):h.push({start:t.Start,end:t.End})}const a=null!=e?e:this.Cfg.Replacement;for(let t=h.length-1;t>=0;t--){const{start:e,end:r}=h[t];i.splice(e,r-e,a)}const c=i.join("");return this._replaceCache&&this._replaceCache.Set(s,c),c}InjectSensitiveWords(t){let e=!1;for(const r of t)r&&r.length>0&&!this._sensitiveWords.has(r)&&(this._sensitiveWords.add(r),e=!0);e&&(this.buildAutomaton(),this.clearCache())}InjectInterferenceWords(t){const e=this._acAutomaton.AddInterferenceWords(t);if(e.length>0){const t=this.Logger;for(const r of e)t.Warn(`干扰词须为单个字符，已忽略 >>> ${r}`)}this.clearCache()}GetSensitiveWordCount(){return this._sensitiveWords.size}GetInterferenceWordCount(){return this._acAutomaton.GetInterferenceWordCount()}ClearSensitiveWords(){this._sensitiveWords.clear(),this.buildAutomaton(),this.clearCache()}ClearInterferenceWords(){this._acAutomaton.ClearInterferenceWords(),this.clearCache()}buildAutomaton(){this._acAutomaton.Clear();const t=this._sensitiveWords;for(const e of t)this._acAutomaton.AddWord(e);this._acAutomaton.Build()}clearCache(){var t,e,r;null===(t=this._testCache)||void 0===t||t.Clear(),null===(e=this._findCache)||void 0===e||e.Clear(),null===(r=this._replaceCache)||void 0===r||r.Clear()}}exports.AuditorPlugin=n;

package/dist/auditor.min.mjs

@@ -1 +1 @@
-import{Override as t}from"@nebulaaa/dict";import{LRU as e}from"@nebulaaa/lru";import{Plugin as r}from"@nebulaaa/plugin";class n{constructor(t=!1){this._root={Children:new Map,Fail:null,Words:[],Output:[]},this._built=!1,this._interferenceWords=new Set,this._ignoreCase=t}normalize(t){if(!this._ignoreCase)return t;let e="";for(const r of t){const t=r.codePointAt(0);e+=t>=65&&t<=90?String.fromCharCode(t+32):r}return e}AddWord(t){if(!t||0===t.length)return;let e=this._root;const r=Array.from(this.normalize(t));for(const t of r){let r=e.Children.get(t);r||(r={Children:new Map,Fail:null,Words:[],Output:[]},e.Children.set(t,r)),e=r}e.Words.includes(t)||e.Words.push(t),this._built=!1}Build(){if(this._built)return;const t=[];for(const e of this._root.Children.values())e.Fail=this._root,e.Output=[...e.Words],t.push(e);let e=0;for(;e<t.length;){const r=t[e++];for(const[e,n]of r.Children){t.push(n);let s=r.Fail;for(;s&&!s.Children.has(e);)s=s.Fail;n.Fail=s&&s.Children.get(e)||this._root,n.Output=[...n.Words,...n.Fail.Output]}}this._built=!0}Match(t){this._built||this.Build();const e=[];let r=this._root;const n=Array.from(this.normalize(t));for(let t=0;t<n.length;t++){const s=n[t];if(!this._interferenceWords.has(s)){for(;r!==this._root&&!r.Children.has(s);)r=r.Fail;if(r.Children.has(s)){r=r.Children.get(s);for(const s of r.Output){let r=t;const i=Array.from(this.normalize(s));let o=i.length-1;for(;r>=0&&o>=0;)this._interferenceWords.has(n[r])||n[r]===i[o]&&o--,r--;e.push({Word:s,Start:r+1,End:t+1})}}}}return e}Test(t){this._built||this.Build();let e=this._root;const r=Array.from(this.normalize(t));for(let t=0;t<r.length;t++){const n=r[t];if(!this._interferenceWords.has(n)){for(;e!==this._root&&!e.Children.has(n);)e=e.Fail;if(e.Children.has(n)&&(e=e.Children.get(n),e.Output.length>0))return!0}}return!1}Clear(){this._root={Children:new Map,Fail:null,Words:[],Output:[]},this._built=!1}AddInterferenceWords(t){for(const e of t){if(!e||0===e.length)continue;const t=Array.from(this.normalize(e));1===t.length?this._interferenceWords.add(t[0]):console.warn(`addInterferenceWords: 干扰词须为单个字符，已忽略 "${e}"`)}}ClearInterferenceWords(){this._interferenceWords.clear()}GetInterferenceWordCount(){return this._interferenceWords.size}}class s extends r{constructor(r,s){if(t(s,{Replacement:"█",IgnoreCase:!1,CacheSize:1e3}),super(r,s),this._acAutomaton=new n(s.IgnoreCase),this._sensitiveWords=new Set,this._acAutomaton.AddInterferenceWords(" \t\r\n".split("")),s.EnableCache){if(!Number.isInteger(s.CacheSize)||s.CacheSize<=0)throw new Error(`AuditorPlugin CacheSize 须为正整数 >>> ${s.CacheSize}`);this._testCache=new e(s.CacheSize),this._findCache=new e(s.CacheSize),this._replaceCache=new e(s.CacheSize)}else this._testCache=null,this._findCache=null,this._replaceCache=null;this.buildAutomaton()}Test(t){if(!t||0===t.length)return!1;if(this._testCache){const e=this._testCache.Get(t);if(null!==e)return e}const e=this._acAutomaton.Test(t);return this._testCache&&this._testCache.Set(t,e),e}Find(t){if(!t||0===t.length)return{Matched:!1,Words:[],Matches:[]};if(this._findCache){const e=this._findCache.Get(t);if(null!==e)return{Matched:e.Matched,Words:[...e.Words],Matches:e.Matches.map(t=>Object.assign({},t))}}const e=this._acAutomaton.Match(t),r=new Set;for(const t of e)r.add(t.Word);const n={Matched:e.length>0,Words:Array.from(r),Matches:e};return this._findCache&&this._findCache.Set(t,{Matched:n.Matched,Words:[...n.Words],Matches:n.Matches.map(t=>Object.assign({},t))}),n}Replace(t,e){var r;if(!t||0===t.length)return t;const n=JSON.stringify([t,null!=e?e:this.Cfg.Replacement]);if(this._replaceCache){const t=this._replaceCache.Get(n);if(null!==t)return t}const s=this.Find(t);if(!s.Matched)return null===(r=this._replaceCache)||void 0===r||r.Set(n,t),t;const i=Array.from(t),o=[...s.Matches].sort((t,e)=>t.Start-e.Start),h=[];for(const t of o){const e=h[h.length-1];e&&t.Start<e.end?e.end=Math.max(e.end,t.End):h.push({start:t.Start,end:t.End})}const a=null!=e?e:this.Cfg.Replacement;for(let t=h.length-1;t>=0;t--){const{start:e,end:r}=h[t];i.splice(e,r-e,a)}const c=i.join("");return this._replaceCache&&this._replaceCache.Set(n,c),c}InjectSensitiveWords(t){for(const e of t)e&&e.length>0&&this._sensitiveWords.add(e);this.buildAutomaton(),this.clearCache()}InjectInterferenceWords(t){this._acAutomaton.AddInterferenceWords(t),this.clearCache()}GetSensitiveWordCount(){return this._sensitiveWords.size}GetInterferenceWordCount(){return this._acAutomaton.GetInterferenceWordCount()}ClearSensitiveWords(){this._sensitiveWords.clear(),this.buildAutomaton(),this.clearCache()}ClearInterferenceWords(){this._acAutomaton.ClearInterferenceWords(),this.clearCache()}buildAutomaton(){this._acAutomaton.Clear();const t=this._sensitiveWords;for(const e of t)this._acAutomaton.AddWord(e);this._acAutomaton.Build()}clearCache(){var t,e,r;null===(t=this._testCache)||void 0===t||t.Clear(),null===(e=this._findCache)||void 0===e||e.Clear(),null===(r=this._replaceCache)||void 0===r||r.Clear()}}export{s as AuditorPlugin};
+import{Override as t}from"@nebulaaa/dict";import{LRU as e}from"@nebulaaa/lru";import{Plugin as r}from"@nebulaaa/plugin";class s{constructor(t=!1){this._root={Children:new Map,Fail:null,Words:[],Output:[]},this._built=!1,this._interferenceWords=new Set,this._ignoreCase=t,this._wordChars=new Map}normalize(t){if(!this._ignoreCase)return t;let e="";for(const r of t){const t=r.codePointAt(0);e+=t>=65&&t<=90?String.fromCharCode(t+32):r}return e}AddWord(t){if(!t||0===t.length)return;let e=this._root;const r=Array.from(this.normalize(t));for(const t of r){let r=e.Children.get(t);r||(r={Children:new Map,Fail:null,Words:[],Output:[]},e.Children.set(t,r)),e=r}0===e.Words.length&&(e.Words.push(t),this._wordChars.set(t,r),this._built=!1)}Build(){if(this._built)return;const t=[];for(const e of this._root.Children.values())e.Fail=this._root,e.Output=[...e.Words],t.push(e);let e=0;for(;e<t.length;){const r=t[e++];for(const[e,s]of r.Children){t.push(s);let n=r.Fail;for(;n&&!n.Children.has(e);)n=n.Fail;s.Fail=n&&n.Children.get(e)||this._root,s.Output=[...s.Words,...s.Fail.Output]}}this._built=!0}Match(t){this._built||this.Build();const e=[];let r=this._root;const s=Array.from(this.normalize(t));for(let t=0;t<s.length;t++){const n=s[t];if(!this._interferenceWords.has(n)){for(;r!==this._root&&!r.Children.has(n);)r=r.Fail;if(r.Children.has(n)){r=r.Children.get(n);for(const n of r.Output){let r=t;const i=this._wordChars.get(n);let o=i.length-1;for(;r>=0&&o>=0;)this._interferenceWords.has(s[r])||s[r]===i[o]&&o--,r--;o>=0||e.push({Word:n,Start:r+1,End:t+1})}}}}return e}Test(t){this._built||this.Build();let e=this._root;const r=Array.from(this.normalize(t));for(let t=0;t<r.length;t++){const s=r[t];if(!this._interferenceWords.has(s)){for(;e!==this._root&&!e.Children.has(s);)e=e.Fail;if(e.Children.has(s)&&(e=e.Children.get(s),e.Output.length>0))return!0}}return!1}Clear(){this._root={Children:new Map,Fail:null,Words:[],Output:[]},this._wordChars.clear(),this._built=!1}AddInterferenceWords(t){const e=[];for(const r of t){if(!r||0===r.length)continue;const t=Array.from(this.normalize(r));1===t.length?this._interferenceWords.add(t[0]):e.push(r)}return e}ClearInterferenceWords(){this._interferenceWords.clear()}GetInterferenceWordCount(){return this._interferenceWords.size}}class n extends r{constructor(r,n){if(t(n,{Replacement:"█",IgnoreCase:!1,CacheSize:1e3}),super(r,n),!Number.isInteger(n.CacheSize)||n.CacheSize<=0)throw new Error(`AuditorPlugin CacheSize 须为正整数 >>> ${n.CacheSize}`);this._acAutomaton=new s(n.IgnoreCase),this._sensitiveWords=new Set,this._acAutomaton.AddInterferenceWords(" \t\r\n".split("")),n.EnableCache?(this._testCache=new e(n.CacheSize),this._findCache=new e(n.CacheSize),this._replaceCache=new e(n.CacheSize)):(this._testCache=null,this._findCache=null,this._replaceCache=null),this.buildAutomaton()}Test(t){if(!t||0===t.length)return!1;if(this._testCache){const e=this._testCache.Get(t);if(null!==e)return e}const e=this._acAutomaton.Test(t);return this._testCache&&this._testCache.Set(t,e),e}Find(t){if(!t||0===t.length)return{Matched:!1,Words:[],Matches:[]};if(this._findCache){const e=this._findCache.Get(t);if(null!==e)return{Matched:e.Matched,Words:[...e.Words],Matches:e.Matches.map(t=>Object.assign({},t))}}const e=this._acAutomaton.Match(t),r=new Set;for(const t of e)r.add(t.Word);const s={Matched:e.length>0,Words:Array.from(r),Matches:e};return this._findCache&&this._findCache.Set(t,{Matched:s.Matched,Words:[...s.Words],Matches:s.Matches.map(t=>Object.assign({},t))}),s}Replace(t,e){var r;if(!t||0===t.length)return t;const s=JSON.stringify([t,null!=e?e:this.Cfg.Replacement]);if(this._replaceCache){const t=this._replaceCache.Get(s);if(null!==t)return t}const n=this.Find(t);if(!n.Matched)return null===(r=this._replaceCache)||void 0===r||r.Set(s,t),t;const i=Array.from(t),o=[...n.Matches].sort((t,e)=>t.Start-e.Start),h=[];for(const t of o){const e=h[h.length-1];e&&t.Start<e.end?e.end=Math.max(e.end,t.End):h.push({start:t.Start,end:t.End})}const a=null!=e?e:this.Cfg.Replacement;for(let t=h.length-1;t>=0;t--){const{start:e,end:r}=h[t];i.splice(e,r-e,a)}const c=i.join("");return this._replaceCache&&this._replaceCache.Set(s,c),c}InjectSensitiveWords(t){let e=!1;for(const r of t)r&&r.length>0&&!this._sensitiveWords.has(r)&&(this._sensitiveWords.add(r),e=!0);e&&(this.buildAutomaton(),this.clearCache())}InjectInterferenceWords(t){const e=this._acAutomaton.AddInterferenceWords(t);if(e.length>0){const t=this.Logger;for(const r of e)t.Warn(`干扰词须为单个字符，已忽略 >>> ${r}`)}this.clearCache()}GetSensitiveWordCount(){return this._sensitiveWords.size}GetInterferenceWordCount(){return this._acAutomaton.GetInterferenceWordCount()}ClearSensitiveWords(){this._sensitiveWords.clear(),this.buildAutomaton(),this.clearCache()}ClearInterferenceWords(){this._acAutomaton.ClearInterferenceWords(),this.clearCache()}buildAutomaton(){this._acAutomaton.Clear();const t=this._sensitiveWords;for(const e of t)this._acAutomaton.AddWord(e);this._acAutomaton.Build()}clearCache(){var t,e,r;null===(t=this._testCache)||void 0===t||t.Clear(),null===(e=this._findCache)||void 0===e||e.Clear(),null===(r=this._replaceCache)||void 0===r||r.Clear()}}export{n as AuditorPlugin};

package/dist/auditor.mjs

@@ -14,6 +14,7 @@ class ACAutomaton {
         this._built = false;
         this._interferenceWords = new Set();
         this._ignoreCase = ignoreCase;
+        this._wordChars = new Map();
     }
     /**
      * 大小写归一（仅 ASCII 字母，逐码点等长替换，保证匹配位置与原始文本索引一致）
@@ -52,11 +53,13 @@ class ACAutomaton {
             }
             current = child;
         }
-        if (!current.Words.includes(word)) {
-            // 同词重复添加幂等：自身词独立存放，output 由 build 重新生成
+        // 归一后同一路径只保留首词：IgnoreCase 下 "abc"/"ABC" 共享路径，
+        // 变体重复入表会让 Match 对同一区间输出多条重复结果（下游命中计数虚高）
+        if (current.Words.length === 0) {
             current.Words.push(word);
+            this._wordChars.set(word, chars);
+            this._built = false; // 标记需要重新构建 fail 指针
         }
-        this._built = false; // 标记需要重新构建 fail 指针
     }
     /**
      * 构建 fail 指针（使用 BFS）
@@ -119,8 +122,8 @@ class ACAutomaton {
                 for (const word of current.Output) {
                     // 计算在原始文本中的起始位置（跳过干扰字符）
                     let j = i;
-                    // 从当前位置向前查找匹配的起始位置
-                    const wordChars = Array.from(this.normalize(word));
+                    // 从当前位置向前查找匹配的起始位置（AddWord 已缓存归一后的码点数组）
+                    const wordChars = this._wordChars.get(word);
                     let wordIndex = wordChars.length - 1;
                     // 从后向前匹配，找到原始文本中的位置
                     while (j >= 0 && wordIndex >= 0) {
@@ -133,6 +136,11 @@ class ACAutomaton {
                         }
                         j--;
                     }
+                    // 回扫必须完整命中：前向 AC 已保证该词是当前位置的后缀（两侧对称跳过干扰字符），
+                    // wordIndex >= 0 说明不变量被破坏（理论不可达），防御性跳过而非产出错误区间
+                    if (wordIndex >= 0) {
+                        continue;
+                    }
                     results.push({
                         Word: word,
                         Start: j + 1,
@@ -182,25 +190,29 @@ class ACAutomaton {
             Words: [],
             Output: [],
         };
+        this._wordChars.clear();
         this._built = false;
     }
     /**
      * 注入干扰词（仅单字符生效：匹配按单字符查表）
      * @param words 干扰词列表
+     * @returns 被忽略的多字符词（由调用方决定告警方式，保持算法类无日志副作用）
      */
     AddInterferenceWords(words) {
+        const ignored = [];
         for (const word of words) {
             if (!word || word.length === 0) {
                 continue;
             }
             const chars = Array.from(this.normalize(word));
             if (chars.length !== 1) {
-                // 多字符干扰词在按字符匹配的查表中永不命中——显式告警而非静默失效
-                console.warn(`addInterferenceWords: 干扰词须为单个字符，已忽略 "${word}"`);
+                // 多字符干扰词在按字符匹配的查表中永不命中——返回给调用方显式告警而非静默失效
+                ignored.push(word);
                 continue;
             }
             this._interferenceWords.add(chars[0]);
         }
+        return ignored;
     }
     /**
      * 清空干扰词
@@ -227,14 +239,14 @@ class AuditorPlugin extends Plugin {
             CacheSize: 1000,
         });
         super(deps, cfg);
+        // 缓存容量须为正整数：无论是否启用缓存都在构造期校验，配置错误不留到运行期
+        if (!Number.isInteger(cfg.CacheSize) || cfg.CacheSize <= 0) {
+            throw new Error(`AuditorPlugin CacheSize 须为正整数 >>> ${cfg.CacheSize}`);
+        }
         this._acAutomaton = new ACAutomaton(cfg.IgnoreCase);
         this._sensitiveWords = new Set();
         this._acAutomaton.AddInterferenceWords(' \t\r\n'.split(''));
         if (cfg.EnableCache) {
-            // 缓存容量须为正整数：非法配置构造期显式失败，错误信息带插件上下文便于定位
-            if (!Number.isInteger(cfg.CacheSize) || cfg.CacheSize <= 0) {
-                throw new Error(`AuditorPlugin CacheSize 须为正整数 >>> ${cfg.CacheSize}`);
-            }
             this._testCache = new LRU(cfg.CacheSize);
             this._findCache = new LRU(cfg.CacheSize);
             this._replaceCache = new LRU(cfg.CacheSize);
@@ -351,16 +363,29 @@ class AuditorPlugin extends Plugin {
         return replacedText;
     }
     InjectSensitiveWords(words) {
+        let added = false;
         for (const word of words) {
-            if (word && word.length > 0) {
+            if (word && word.length > 0 && !this._sensitiveWords.has(word)) {
                 this._sensitiveWords.add(word);
+                added = true;
             }
         }
+        // 无新增词时词库未变：跳过自动机重建与缓存清理，重复注入不打断既有缓存
+        if (!added) {
+            return;
+        }
         this.buildAutomaton();
         this.clearCache();
     }
     InjectInterferenceWords(words) {
-        this._acAutomaton.AddInterferenceWords(words);
+        const ignored = this._acAutomaton.AddInterferenceWords(words);
+        // 多字符干扰词永不命中：经插件日志显式告警（词库内容不落 console）
+        if (ignored.length > 0) {
+            const logger = this.Logger;
+            for (const word of ignored) {
+                logger.Warn(`干扰词须为单个字符，已忽略 >>> ${word}`);
+            }
+        }
         this.clearCache();
     }
     GetSensitiveWordCount() {

package/dist/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nebulaaa/auditor",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "敏感词审计员插件",
   "author": "doooreyn",
   "license": "MIT",

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@nebulaaa/auditor",
-  "version": "1.0.0",
+  "version": "1.0.1",
   "description": "敏感词审计员插件",
   "main": "./dist/auditor.cjs",
   "module": "./dist/auditor.mjs",