terlik.js 2.2.0 → 2.3.0

package/README.md

@@ -451,6 +451,37 @@ See [CONTRIBUTING.md](./CONTRIBUTING.md) for contribution guidelines.
 
 ## Changelog
 
+### 2026-02-28 (v2.3.0) — 40x Faster Cold Start: V8 JIT Regex Optimization
+
+**Replaces `\p{L}`/`\p{N}` Unicode property escapes with explicit Latin ranges, eliminating V8 JIT bottleneck.**
+
+- **40x faster cold start** — First `containsProfanity()` call: 16,494ms → 404ms.
+- **356x faster multi-language warmup** — 4-language warmup: 19,234ms → 54ms.
+- **13x less memory** — Heap usage: 492MB → 38MB.
+- **Static pattern cache** — Same-language instances share compiled patterns via `Detector.patternCache`.
+- **Background warmup** — Dev server starts instantly, warms up in background.
+
+| Change | File |
+|---|---|
+| Replace `\p{L}\p{N}` with `[a-zA-Z0-9À-ɏ]` | `src/patterns.ts` |
+| Static pattern cache + explicit range in getSurroundingWord | `src/detector.ts` |
+| Explicit range in number expander + punctuation removal | `src/normalizer.ts` |
+| Pass cacheKey to Detector | `src/terlik.ts` |
+| Background warmup, lazy instance cache | `live_test_server/server.ts` |
+| NODE_OPTIONS heap safety net | `.github/workflows/ci.yml` |
+
+### 2026-02-28 (v2.2.1) — CI Fix: Timeout Race Condition + İ Platform Compatibility
+
+**Fixes detection failures on slow runners and cross-platform İ (U+0130) handling.**
+
+- **Timeout race condition fix** — `REGEX_TIMEOUT_MS` check moved from _before_ match processing to _after_. Previously, V8 JIT compilation on first `exec()` call (triggered by lazy compilation) could exceed 250ms, causing the timeout to discard a valid match before it was recorded. Now the current match is always processed; the timeout only prevents scanning for additional matches.
+- **İ (U+0130) cross-platform fix** — First regex pass now runs on `text.toLocaleLowerCase(locale)` instead of raw text. Turkish İ→i mapping is performed explicitly before regex matching, avoiding inconsistent V8/ICU case-folding behavior across platforms (Ubuntu vs macOS). The `mapNormalizedToOriginal()` mapper recovers original-cased words for result output.
+
+| Change | File |
+|---|---|
+| Timeout check moved after match processing | `src/detector.ts` (`runPatterns`) |
+| Locale-lower first pass for İ safety | `src/detector.ts` (`detectPattern`) |
+
 ### 2026-02-28 (v2.2) — Lazy Compilation + Linguistic Patch
 
 **Zero-cost construction. Background warmup. Turkish agglutination hardening.**

package/dist/index.js

@@ -147,7 +147,10 @@ var Dictionary = class {
 };
 
 // src/patterns.ts
-var SEPARATOR = "[^\\p{L}\\p{N}]{0,3}";
+var WORD_CHAR = "a-zA-Z0-9\xC0-\u024F";
+var SEPARATOR = `[^${WORD_CHAR}]{0,3}`;
+var WORD_BOUNDARY_BEHIND = `(?<![${WORD_CHAR}])`;
+var WORD_BOUNDARY_AHEAD = `(?![${WORD_CHAR}])`;
 var MAX_PATTERN_LENGTH = 1e4;
 var MAX_SUFFIX_CHAIN = 2;
 var REGEX_TIMEOUT_MS = 250;
@@ -185,15 +188,15 @@ function compilePatterns(entries, suffixes, charClasses, normalizeFn) {
     const useSuffix = entry.suffixable && suffixGroup.length > 0;
     let pattern;
     if (useSuffix) {
-      pattern = `(?<![\\p{L}\\p{N}])(?:${combined})${suffixGroup}{0,${MAX_SUFFIX_CHAIN}}(?![\\p{L}\\p{N}])`;
+      pattern = `${WORD_BOUNDARY_BEHIND}(?:${combined})${suffixGroup}{0,${MAX_SUFFIX_CHAIN}}${WORD_BOUNDARY_AHEAD}`;
     } else {
-      pattern = `(?<![\\p{L}\\p{N}])(?:${combined})(?![\\p{L}\\p{N}])`;
+      pattern = `${WORD_BOUNDARY_BEHIND}(?:${combined})${WORD_BOUNDARY_AHEAD}`;
     }
     if (pattern.length > MAX_PATTERN_LENGTH && useSuffix) {
-      pattern = `(?<![\\p{L}\\p{N}])(?:${combined})(?![\\p{L}\\p{N}])`;
+      pattern = `${WORD_BOUNDARY_BEHIND}(?:${combined})${WORD_BOUNDARY_AHEAD}`;
     }
     try {
-      const regex = new RegExp(pattern, "giu");
+      const regex = new RegExp(pattern, "gi");
       patterns.push({
         root: entry.root,
         severity: entry.severity,
@@ -203,8 +206,8 @@ function compilePatterns(entries, suffixes, charClasses, normalizeFn) {
     } catch (err) {
       if (useSuffix) {
         try {
-          const fallbackPattern = `(?<![\\p{L}\\p{N}])(?:${combined})(?![\\p{L}\\p{N}])`;
-          const regex = new RegExp(fallbackPattern, "giu");
+          const fallbackPattern = `${WORD_BOUNDARY_BEHIND}(?:${combined})${WORD_BOUNDARY_AHEAD}`;
+          const regex = new RegExp(fallbackPattern, "gi");
           patterns.push({
             root: entry.root,
             severity: entry.severity,
@@ -278,31 +281,45 @@ function getFuzzyMatcher(algorithm) {
 }
 
 // src/detector.ts
-var Detector = class {
+var Detector = class _Detector {
+  /** Static cache: shares compiled patterns across instances with identical dictionaries. */
+  static patternCache = /* @__PURE__ */ new Map();
   dictionary;
   _patterns = null;
+  cacheKey;
   normalizedWordSet;
   normalizedWordToRoot;
   normalizeFn;
   locale;
   charClasses;
-  constructor(dictionary, normalizeFn, locale, charClasses) {
+  constructor(dictionary, normalizeFn, locale, charClasses, cacheKey) {
     this.dictionary = dictionary;
     this.normalizeFn = normalizeFn;
     this.locale = locale;
     this.charClasses = charClasses;
+    this.cacheKey = cacheKey ?? null;
     this.normalizedWordSet = /* @__PURE__ */ new Set();
     this.normalizedWordToRoot = /* @__PURE__ */ new Map();
     this.buildNormalizedLookup();
   }
   ensureCompiled() {
     if (this._patterns === null) {
+      if (this.cacheKey) {
+        const cached = _Detector.patternCache.get(this.cacheKey);
+        if (cached) {
+          this._patterns = cached;
+          return this._patterns;
+        }
+      }
       this._patterns = compilePatterns(
         this.dictionary.getEntries(),
         this.dictionary.getSuffixes(),
         this.charClasses,
         this.normalizeFn
       );
+      if (this.cacheKey) {
+        _Detector.patternCache.set(this.cacheKey, this._patterns);
+      }
     }
     return this._patterns;
   }
@@ -310,6 +327,7 @@ var Detector = class {
     this.ensureCompiled();
   }
   recompile() {
+    this.cacheKey = null;
     this._patterns = compilePatterns(
       this.dictionary.getEntries(),
       this.dictionary.getSuffixes(),
@@ -383,9 +401,9 @@ var Detector = class {
     }
   }
   detectPattern(text, whitelist, results) {
-    this.runPatterns(text, text, whitelist, results, false);
-    const normalizedText = this.normalizeFn(text);
     const lowerText = text.toLocaleLowerCase(this.locale);
+    this.runPatterns(lowerText, text, whitelist, results, lowerText !== text);
+    const normalizedText = this.normalizeFn(text);
     if (normalizedText !== lowerText && normalizedText.length > 0) {
       this.runPatterns(normalizedText, text, whitelist, results, true);
     }
@@ -398,7 +416,6 @@ var Detector = class {
       pattern.regex.lastIndex = 0;
       let match;
       while ((match = pattern.regex.exec(searchText)) !== null) {
-        if (Date.now() - patternStart > REGEX_TIMEOUT_MS) break;
         const matchedText = match[0];
         const matchIndex = match.index;
         const normalizedMatch = this.normalizeFn(matchedText);
@@ -433,6 +450,7 @@ var Detector = class {
         if (matchedText.length === 0) {
           pattern.regex.lastIndex++;
         }
+        if (Date.now() - patternStart > REGEX_TIMEOUT_MS) break;
       }
     }
   }
@@ -499,8 +517,8 @@ var Detector = class {
   getSurroundingWord(text, index, length) {
     let start = index;
     let end = index + length;
-    while (start > 0 && /\p{L}/u.test(text[start - 1])) start--;
-    while (end < text.length && /\p{L}/u.test(text[end])) end++;
+    while (start > 0 && /[a-zA-ZÀ-ɏ]/.test(text[start - 1])) start--;
+    while (end < text.length && /[a-zA-ZÀ-ɏ]/.test(text[end])) end++;
     return text.slice(start, end);
   }
   deduplicateResults(results) {
@@ -1793,15 +1811,15 @@ function buildNumberExpander(expansions) {
   const regex = new RegExp(
     expansions.map(([num]) => {
       const escaped = num.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
-      return `(?<=\\p{L})${escaped}(?=\\p{L})`;
+      return `(?<=[a-zA-Z\xC0-\u024F])${escaped}(?=[a-zA-Z\xC0-\u024F])`;
     }).join("|"),
-    "gu"
+    "g"
   );
   const lookup = Object.fromEntries(expansions);
   return (text) => text.replace(regex, (match) => lookup[match] ?? match);
 }
 function removePunctuation(text) {
-  return text.replace(/(?<=\p{L})[.\-_*,;:!?]+(?=\p{L})/gu, "");
+  return text.replace(/(?<=[a-zA-ZÀ-ɏ])[.\-_*,;:!?]+(?=[a-zA-ZÀ-ɏ])/g, "");
 }
 function collapseRepeats(text) {
   return text.replace(/(.)\1{2,}/g, "$1");
@@ -1917,11 +1935,13 @@ var Terlik = class _Terlik {
       options?.customList,
       options?.whitelist
     );
+    const hasCustomDict = !!(options?.customList?.length || options?.whitelist?.length);
     this.detector = new Detector(
       this.dictionary,
       normalizeFn,
       langConfig.locale,
-      langConfig.charClasses
+      langConfig.charClasses,
+      hasCustomDict ? null : this.language
     );
     if (options?.backgroundWarmup) {
       setTimeout(() => {

package/dist/index.mjs

@@ -114,7 +114,10 @@ var Dictionary = class {
 };
 
 // src/patterns.ts
-var SEPARATOR = "[^\\p{L}\\p{N}]{0,3}";
+var WORD_CHAR = "a-zA-Z0-9\xC0-\u024F";
+var SEPARATOR = `[^${WORD_CHAR}]{0,3}`;
+var WORD_BOUNDARY_BEHIND = `(?<![${WORD_CHAR}])`;
+var WORD_BOUNDARY_AHEAD = `(?![${WORD_CHAR}])`;
 var MAX_PATTERN_LENGTH = 1e4;
 var MAX_SUFFIX_CHAIN = 2;
 var REGEX_TIMEOUT_MS = 250;
@@ -152,15 +155,15 @@ function compilePatterns(entries, suffixes, charClasses, normalizeFn) {
     const useSuffix = entry.suffixable && suffixGroup.length > 0;
     let pattern;
     if (useSuffix) {
-      pattern = `(?<![\\p{L}\\p{N}])(?:${combined})${suffixGroup}{0,${MAX_SUFFIX_CHAIN}}(?![\\p{L}\\p{N}])`;
+      pattern = `${WORD_BOUNDARY_BEHIND}(?:${combined})${suffixGroup}{0,${MAX_SUFFIX_CHAIN}}${WORD_BOUNDARY_AHEAD}`;
     } else {
-      pattern = `(?<![\\p{L}\\p{N}])(?:${combined})(?![\\p{L}\\p{N}])`;
+      pattern = `${WORD_BOUNDARY_BEHIND}(?:${combined})${WORD_BOUNDARY_AHEAD}`;
     }
     if (pattern.length > MAX_PATTERN_LENGTH && useSuffix) {
-      pattern = `(?<![\\p{L}\\p{N}])(?:${combined})(?![\\p{L}\\p{N}])`;
+      pattern = `${WORD_BOUNDARY_BEHIND}(?:${combined})${WORD_BOUNDARY_AHEAD}`;
     }
     try {
-      const regex = new RegExp(pattern, "giu");
+      const regex = new RegExp(pattern, "gi");
       patterns.push({
         root: entry.root,
         severity: entry.severity,
@@ -170,8 +173,8 @@ function compilePatterns(entries, suffixes, charClasses, normalizeFn) {
     } catch (err) {
       if (useSuffix) {
         try {
-          const fallbackPattern = `(?<![\\p{L}\\p{N}])(?:${combined})(?![\\p{L}\\p{N}])`;
-          const regex = new RegExp(fallbackPattern, "giu");
+          const fallbackPattern = `${WORD_BOUNDARY_BEHIND}(?:${combined})${WORD_BOUNDARY_AHEAD}`;
+          const regex = new RegExp(fallbackPattern, "gi");
           patterns.push({
             root: entry.root,
             severity: entry.severity,
@@ -245,31 +248,45 @@ function getFuzzyMatcher(algorithm) {
 }
 
 // src/detector.ts
-var Detector = class {
+var Detector = class _Detector {
+  /** Static cache: shares compiled patterns across instances with identical dictionaries. */
+  static patternCache = /* @__PURE__ */ new Map();
   dictionary;
   _patterns = null;
+  cacheKey;
   normalizedWordSet;
   normalizedWordToRoot;
   normalizeFn;
   locale;
   charClasses;
-  constructor(dictionary, normalizeFn, locale, charClasses) {
+  constructor(dictionary, normalizeFn, locale, charClasses, cacheKey) {
     this.dictionary = dictionary;
     this.normalizeFn = normalizeFn;
     this.locale = locale;
     this.charClasses = charClasses;
+    this.cacheKey = cacheKey ?? null;
     this.normalizedWordSet = /* @__PURE__ */ new Set();
     this.normalizedWordToRoot = /* @__PURE__ */ new Map();
     this.buildNormalizedLookup();
   }
   ensureCompiled() {
     if (this._patterns === null) {
+      if (this.cacheKey) {
+        const cached = _Detector.patternCache.get(this.cacheKey);
+        if (cached) {
+          this._patterns = cached;
+          return this._patterns;
+        }
+      }
       this._patterns = compilePatterns(
         this.dictionary.getEntries(),
         this.dictionary.getSuffixes(),
         this.charClasses,
         this.normalizeFn
       );
+      if (this.cacheKey) {
+        _Detector.patternCache.set(this.cacheKey, this._patterns);
+      }
     }
     return this._patterns;
   }
@@ -277,6 +294,7 @@ var Detector = class {
     this.ensureCompiled();
   }
   recompile() {
+    this.cacheKey = null;
     this._patterns = compilePatterns(
       this.dictionary.getEntries(),
       this.dictionary.getSuffixes(),
@@ -350,9 +368,9 @@ var Detector = class {
     }
   }
   detectPattern(text, whitelist, results) {
-    this.runPatterns(text, text, whitelist, results, false);
-    const normalizedText = this.normalizeFn(text);
     const lowerText = text.toLocaleLowerCase(this.locale);
+    this.runPatterns(lowerText, text, whitelist, results, lowerText !== text);
+    const normalizedText = this.normalizeFn(text);
     if (normalizedText !== lowerText && normalizedText.length > 0) {
       this.runPatterns(normalizedText, text, whitelist, results, true);
     }
@@ -365,7 +383,6 @@ var Detector = class {
       pattern.regex.lastIndex = 0;
       let match;
       while ((match = pattern.regex.exec(searchText)) !== null) {
-        if (Date.now() - patternStart > REGEX_TIMEOUT_MS) break;
         const matchedText = match[0];
         const matchIndex = match.index;
         const normalizedMatch = this.normalizeFn(matchedText);
@@ -400,6 +417,7 @@ var Detector = class {
         if (matchedText.length === 0) {
           pattern.regex.lastIndex++;
         }
+        if (Date.now() - patternStart > REGEX_TIMEOUT_MS) break;
       }
     }
   }
@@ -466,8 +484,8 @@ var Detector = class {
   getSurroundingWord(text, index, length) {
     let start = index;
     let end = index + length;
-    while (start > 0 && /\p{L}/u.test(text[start - 1])) start--;
-    while (end < text.length && /\p{L}/u.test(text[end])) end++;
+    while (start > 0 && /[a-zA-ZÀ-ɏ]/.test(text[start - 1])) start--;
+    while (end < text.length && /[a-zA-ZÀ-ɏ]/.test(text[end])) end++;
     return text.slice(start, end);
   }
   deduplicateResults(results) {
@@ -1760,15 +1778,15 @@ function buildNumberExpander(expansions) {
   const regex = new RegExp(
     expansions.map(([num]) => {
       const escaped = num.replace(/[.*+?^${}()|[\]\\]/g, "\\$&");
-      return `(?<=\\p{L})${escaped}(?=\\p{L})`;
+      return `(?<=[a-zA-Z\xC0-\u024F])${escaped}(?=[a-zA-Z\xC0-\u024F])`;
     }).join("|"),
-    "gu"
+    "g"
   );
   const lookup = Object.fromEntries(expansions);
   return (text) => text.replace(regex, (match) => lookup[match] ?? match);
 }
 function removePunctuation(text) {
-  return text.replace(/(?<=\p{L})[.\-_*,;:!?]+(?=\p{L})/gu, "");
+  return text.replace(/(?<=[a-zA-ZÀ-ɏ])[.\-_*,;:!?]+(?=[a-zA-ZÀ-ɏ])/g, "");
 }
 function collapseRepeats(text) {
   return text.replace(/(.)\1{2,}/g, "$1");
@@ -1884,11 +1902,13 @@ var Terlik = class _Terlik {
       options?.customList,
       options?.whitelist
     );
+    const hasCustomDict = !!(options?.customList?.length || options?.whitelist?.length);
     this.detector = new Detector(
       this.dictionary,
       normalizeFn,
       langConfig.locale,
-      langConfig.charClasses
+      langConfig.charClasses,
+      hasCustomDict ? null : this.language
     );
     if (options?.backgroundWarmup) {
       setTimeout(() => {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "terlik.js",
-  "version": "2.2.0",
+  "version": "2.3.0",
   "description": "Ultra-fast, zero-dependency multi-language profanity detection engine for Turkish, English, Spanish, and German with lazy compilation, deep agglutination support, and ReDoS-safe regex patterns",
   "main": "./dist/index.js",
   "module": "./dist/index.mjs",