eyeling 1.25.0 → 1.25.2

package/lib/lexer.js

@@ -30,7 +30,26 @@ class N3SyntaxError extends SyntaxError {
 }
 
 function isWs(c) {
-  return /\s/.test(c);
+  if (c === null || c === undefined) return false;
+  const code = c.charCodeAt(0);
+  // Fast path for the whitespace used by N3/Turtle inputs.
+  return code === 0x20 || code === 0x09 || code === 0x0a || code === 0x0d || code === 0x0c;
+}
+
+function isAsciiAlphaCode(code) {
+  return (code >= 65 && code <= 90) || (code >= 97 && code <= 122);
+}
+
+function isAsciiDigitCode(code) {
+  return code >= 48 && code <= 57;
+}
+
+function isAsciiAlpha(c) {
+  return c !== null && c !== undefined && isAsciiAlphaCode(c.charCodeAt(0));
+}
+
+function isAsciiDigit(c) {
+  return c !== null && c !== undefined && isAsciiDigitCode(c.charCodeAt(0));
 }
 
 // Turtle/N3 prefixed names (PNAME_*) allow many Unicode letters and certain
@@ -43,13 +62,18 @@ function isWs(c) {
 //
 // We implement a grammar-aligned matcher for PN_CHARS* and PLX fragments.
 function isHexDigit(c) {
-  return c !== null && /^[0-9A-Fa-f]$/.test(c);
+  if (c === null || c === undefined) return false;
+  const code = c.charCodeAt(0);
+  return (code >= 48 && code <= 57) || (code >= 65 && code <= 70) || (code >= 97 && code <= 102);
 }
 
 function isPnCharsBase(c) {
   // Approximation of PN_CHARS_BASE from the N3 grammar using Unicode properties.
   // Covers most letters used in practice (including ñ) and common scripts.
-  return c !== null && /[A-Za-z]|\p{L}|\p{Nl}/u.test(c);
+  if (c === null || c === undefined) return false;
+  const code = c.charCodeAt(0);
+  if (isAsciiAlphaCode(code)) return true;
+  return /\p{L}|\p{Nl}/u.test(c);
 }
 
 function isPnCharsU(c) {
@@ -59,9 +83,11 @@ function isPnCharsU(c) {
 
 function isPnChars(c) {
   // PN_CHARS ::= PN_CHARS_U | '-' | [0-9] | U+00B7 | [U+0300-U+036F] | [U+203F-U+2040]
-  if (c === null) return false;
+  if (c === null || c === undefined) return false;
+  const code = c.charCodeAt(0);
+  if (isAsciiAlphaCode(code) || isAsciiDigitCode(code) || code === 95 || code === 45) return true;
   if (isPnCharsU(c)) return true;
-  if (c === '-' || /[0-9]/.test(c) || c === '\u00B7') return true;
+  if (c === '\u00B7') return true;
   const cp = c.codePointAt(0);
   return (cp >= 0x0300 && cp <= 0x036f) || (cp >= 0x203f && cp <= 0x2040);
 }
@@ -1163,10 +1189,23 @@ function normalizeRdfCompatibility(inputText) {
   return text;
 }
 
+
+function isNumericLikeIdentifier(word) {
+  if (typeof word !== 'string' || word.length === 0) return false;
+  for (let j = 0; j < word.length; j++) {
+    const code = word.charCodeAt(j);
+    if (!((code >= 48 && code <= 57) || code === 46 || code === 45)) return false;
+  }
+  return true;
+}
+
 function lex(inputText, opts = {}) {
   const rdf = !!(opts && opts.rdf);
   if (rdf) inputText = normalizeRdfCompatibility(inputText);
-  const chars = Array.from(inputText);
+  // Avoid copying large ASCII/BMP inputs into an Array.  Array.from() is
+  // only needed when the text contains surrogate pairs and we want the old
+  // code-point iteration behavior for non-BMP characters.
+  const chars = /[\uD800-\uDFFF]/.test(inputText) ? Array.from(inputText) : inputText;
   const n = chars.length;
   let i = 0;
   const tokens = [];
@@ -1182,19 +1221,29 @@ function lex(inputText, opts = {}) {
   // - Accepts percent escapes (%HH) as PLX fragments.
   // - Accepts PN_LOCAL_ESC backslash escapes and decodes them ("\\." -> ".").
   // - Accepts '.' inside a name only when it is not terminal.
+  function sliceChars(start, end) {
+    return typeof chars === 'string' ? chars.slice(start, end) : chars.slice(start, end).join('');
+  }
+
   function readIdentText(startOffsetForErrors) {
-    const out = [];
+    const start = i;
+    let out = null;
+
+    function appendRawUntilHere() {
+      if (out === null) out = [sliceChars(start, i)];
+    }
+
     while (i < n) {
-      const cc = peek();
-      if (cc === null || isWs(cc)) break;
+      const cc = chars[i];
+      if (cc === null || cc === undefined || isWs(cc)) break;
 
       // Hard stops: delimiters cannot appear unescaped inside PNAME tokens.
-      if ('{}()[];,'.includes(cc)) break;
+      if (cc === '{' || cc === '}' || cc === '(' || cc === ')' || cc === '[' || cc === ']' || cc === ';' || cc === ',') break;
 
       // Dot is allowed inside PN_LOCAL, but not at the end.
       if (cc === '.') {
         if (!canContinueAfterDot(peek(1))) break;
-        out.push('.');
+        if (out !== null) out.push('.');
         i++;
         continue;
       }
@@ -1209,6 +1258,7 @@ function lex(inputText, opts = {}) {
             typeof startOffsetForErrors === 'number' ? startOffsetForErrors : i,
           );
         }
+        appendRawUntilHere();
         out.push('%', h1, h2);
         i += 3;
         continue;
@@ -1218,6 +1268,7 @@ function lex(inputText, opts = {}) {
       if (cc === '\\') {
         const esc = peek(1);
         if (esc !== null && PN_LOCAL_ESC_SET.has(esc)) {
+          appendRawUntilHere();
           out.push(esc); // decoded form
           i += 2;
           continue;
@@ -1229,14 +1280,14 @@ function lex(inputText, opts = {}) {
       }
 
       if (isIdentChar(cc)) {
-        out.push(cc);
+        if (out !== null) out.push(cc);
         i++;
         continue;
       }
 
       break;
     }
-    return out.join('');
+    return out === null ? sliceChars(start, i) : out.join('');
   }
 
   while (i < n) {
@@ -1315,22 +1366,47 @@ function lex(inputText, opts = {}) {
       continue;
     }
 
-    // 5) Single-character punctuation
-    if ('{}()[];,.'.includes(c)) {
-      const mapping = {
-        '{': 'LBrace',
-        '}': 'RBrace',
-        '(': 'LParen',
-        ')': 'RParen',
-        '[': 'LBracket',
-        ']': 'RBracket',
-        ';': 'Semicolon',
-        ',': 'Comma',
-        '.': 'Dot',
-      };
-      tokens.push(new Token(mapping[c], null, i));
-      i++;
-      continue;
+    // 5) Single-character punctuation.  Use a switch rather than allocating a
+    // mapping object for every punctuation token in large inputs.
+    switch (c) {
+      case '{':
+        tokens.push(new Token('LBrace', null, i));
+        i++;
+        continue;
+      case '}':
+        tokens.push(new Token('RBrace', null, i));
+        i++;
+        continue;
+      case '(':
+        tokens.push(new Token('LParen', null, i));
+        i++;
+        continue;
+      case ')':
+        tokens.push(new Token('RParen', null, i));
+        i++;
+        continue;
+      case '[':
+        tokens.push(new Token('LBracket', null, i));
+        i++;
+        continue;
+      case ']':
+        tokens.push(new Token('RBracket', null, i));
+        i++;
+        continue;
+      case ';':
+        tokens.push(new Token('Semicolon', null, i));
+        i++;
+        continue;
+      case ',':
+        tokens.push(new Token('Comma', null, i));
+        i++;
+        continue;
+      case '.':
+        tokens.push(new Token('Dot', null, i));
+        i++;
+        continue;
+      default:
+        break;
     }
 
     // String literal: short "..." or long """..."""
@@ -1389,26 +1465,36 @@ function lex(inputText, opts = {}) {
         continue;
       }
 
-      // Short string literal " ... "
+      // Short string literal " ... ".  Most data files contain plain
+      // unescaped labels; keep that path slice-based and avoid building an
+      // intermediate character array + raw quoted string.
       i++; // consume opening "
-      const sChars = [];
+      const contentStart = i;
+      let sChars = null;
+      let closed = false;
       while (i < n) {
         const cc = chars[i];
         i++;
         if (cc === '\\') {
+          if (sChars === null) sChars = [sliceChars(contentStart, i - 1)];
           if (i < n) {
             const esc = chars[i];
             i++;
             sChars.push('\\');
             sChars.push(esc);
+          } else {
+            sChars.push('\\');
           }
           continue;
         }
-        if (cc === '"') break;
-        sChars.push(cc);
+        if (cc === '"') {
+          closed = true;
+          break;
+        }
+        if (sChars !== null) sChars.push(cc);
       }
-      const raw = '"' + sChars.join('') + '"';
-      const decoded = decodeN3StringEscapes(stripQuotes(raw), start);
+      const rawContent = sChars === null ? sliceChars(contentStart, closed ? i - 1 : i) : sChars.join('');
+      const decoded = sChars === null ? rawContent : decodeN3StringEscapes(rawContent, start);
       assertValidStringLiteralValue(decoded, start);
       const s = JSON.stringify(decoded); // canonical short quoted form
       tokens.push(new Token('Literal', s, start));
@@ -1473,24 +1559,32 @@ function lex(inputText, opts = {}) {
 
       // Short string literal ' ... '
       i++; // consume opening '
-      const sChars = [];
+      const contentStart = i;
+      let sChars = null;
+      let closed = false;
       while (i < n) {
         const cc = chars[i];
         i++;
         if (cc === '\\') {
+          if (sChars === null) sChars = [sliceChars(contentStart, i - 1)];
           if (i < n) {
             const esc = chars[i];
             i++;
             sChars.push('\\');
             sChars.push(esc);
+          } else {
+            sChars.push('\\');
           }
           continue;
         }
-        if (cc === "'") break;
-        sChars.push(cc);
+        if (cc === "'") {
+          closed = true;
+          break;
+        }
+        if (sChars !== null) sChars.push(cc);
       }
-      const raw = "'" + sChars.join('') + "'";
-      const decoded = decodeN3StringEscapes(stripQuotes(raw), start);
+      const rawContent = sChars === null ? sliceChars(contentStart, closed ? i - 1 : i) : sChars.join('');
+      const decoded = sChars === null ? rawContent : decodeN3StringEscapes(rawContent, start);
       assertValidStringLiteralValue(decoded, start);
       const s = JSON.stringify(decoded); // canonical short quoted form
       tokens.push(new Token('Literal', s, start));
@@ -1523,10 +1617,10 @@ function lex(inputText, opts = {}) {
         //   "@" [a-zA-Z]+ ("-" [a-zA-Z0-9]+)*
         const tagChars = [];
         let cc = peek();
-        if (cc === null || !/[A-Za-z]/.test(cc)) {
+        if (cc === null || !isAsciiAlpha(cc)) {
           throw new N3SyntaxError("Invalid language tag (expected [A-Za-z] after '@')", start);
         }
-        while ((cc = peek()) !== null && /[A-Za-z]/.test(cc)) {
+        while ((cc = peek()) !== null && isAsciiAlpha(cc)) {
           tagChars.push(cc);
           i++;
         }
@@ -1550,7 +1644,7 @@ function lex(inputText, opts = {}) {
       // Otherwise, treat as a directive (@prefix, @base)
       const wordChars = [];
       let cc;
-      while ((cc = peek()) !== null && /[A-Za-z]/.test(cc)) {
+      while ((cc = peek()) !== null && isAsciiAlpha(cc)) {
         wordChars.push(cc);
         i++;
       }
@@ -1562,19 +1656,19 @@ function lex(inputText, opts = {}) {
     }
 
     // 6) Numeric literal (integer or float)
-    if (/[0-9]/.test(c) || (c === '-' && peek(1) !== null && /[0-9]/.test(peek(1)))) {
+    if (isAsciiDigit(c) || (c === '-' && peek(1) !== null && isAsciiDigit(peek(1)))) {
       const start = i;
       const numChars = [c];
       i++;
       while (i < n) {
         const cc = chars[i];
-        if (/[0-9]/.test(cc)) {
+        if (isAsciiDigit(cc)) {
           numChars.push(cc);
           i++;
           continue;
         }
         if (cc === '.') {
-          if (i + 1 < n && /[0-9]/.test(chars[i + 1])) {
+          if (i + 1 < n && isAsciiDigit(chars[i + 1])) {
             numChars.push('.');
             i++;
             continue;
@@ -1589,14 +1683,14 @@ function lex(inputText, opts = {}) {
       if (i < n && (chars[i] === 'e' || chars[i] === 'E')) {
         let j = i + 1;
         if (j < n && (chars[j] === '+' || chars[j] === '-')) j++;
-        if (j < n && /[0-9]/.test(chars[j])) {
+        if (j < n && isAsciiDigit(chars[j])) {
           numChars.push(chars[i]); // e/E
           i++;
           if (i < n && (chars[i] === '+' || chars[i] === '-')) {
             numChars.push(chars[i]);
             i++;
           }
-          while (i < n && /[0-9]/.test(chars[i])) {
+          while (i < n && isAsciiDigit(chars[i])) {
             numChars.push(chars[i]);
             i++;
           }
@@ -1615,7 +1709,7 @@ function lex(inputText, opts = {}) {
     }
     if (word === 'true' || word === 'false') {
       tokens.push(new Token('Literal', word, start));
-    } else if ([...word].every((ch) => /[0-9.-]/.test(ch))) {
+    } else if (isNumericLikeIdentifier(word)) {
       tokens.push(new Token('Literal', word, start));
     } else {
       tokens.push(new Token('Ident', word, start));

package/lib/parser.js

@@ -86,7 +86,15 @@ class Parser {
   }
 
   isIdentKeyword(tok, keyword) {
-    return tok && tok.typ === 'Ident' && typeof tok.value === 'string' && tok.value.toLowerCase() === keyword;
+    if (!tok || tok.typ !== 'Ident' || typeof tok.value !== 'string') return false;
+    const v = tok.value;
+    if (v.length !== keyword.length) return false;
+    for (let i = 0; i < keyword.length; i++) {
+      const code = v.charCodeAt(i);
+      const lower = code >= 65 && code <= 90 ? code + 32 : code;
+      if (lower !== keyword.charCodeAt(i)) return false;
+    }
+    return true;
   }
 
   canStartSparqlPrefixDirective() {
@@ -233,7 +241,7 @@ class Parser {
     } else if (tok2.typ === 'Ident') {
       const qn = tok2.value || '';
       if (!qn.includes(':')) failInvalidKeywordLikeIdent(this.fail.bind(this), tok2, qn);
-      assertValidQNamePrefix(qn.split(':', 1)[0], this.fail.bind(this), tok2, '@prefix directive IRI');
+      assertValidQNamePrefix(qn.slice(0, qn.indexOf(':')), this.fail.bind(this), tok2, '@prefix directive IRI');
       iri = this.prefixes.expandQName(qn);
     } else {
       this.fail(`Expected IRI after @prefix, got ${tok2.toString()}`, tok2);
@@ -250,7 +258,7 @@ class Parser {
     } else if (tok.typ === 'Ident') {
       const qn = tok.value || '';
       if (!qn.includes(':')) failInvalidKeywordLikeIdent(this.fail.bind(this), tok, qn);
-      assertValidQNamePrefix(qn.split(':', 1)[0], this.fail.bind(this), tok, '@base directive IRI');
+      assertValidQNamePrefix(qn.slice(0, qn.indexOf(':')), this.fail.bind(this), tok, '@base directive IRI');
       iri = this.prefixes.expandQName(qn);
     } else {
       this.fail(`Expected IRI after @base, got ${tok.toString()}`, tok);
@@ -279,7 +287,7 @@ class Parser {
     } else if (tok2.typ === 'Ident') {
       const qn = tok2.value || '';
       if (!qn.includes(':')) failInvalidKeywordLikeIdent(this.fail.bind(this), tok2, qn);
-      assertValidQNamePrefix(qn.split(':', 1)[0], this.fail.bind(this), tok2, '@prefix directive IRI');
+      assertValidQNamePrefix(qn.slice(0, qn.indexOf(':')), this.fail.bind(this), tok2, '@prefix directive IRI');
       iri = this.prefixes.expandQName(qn);
     } else {
       this.fail(`Expected IRI after PREFIX, got ${tok2.toString()}`, tok2);
@@ -300,7 +308,7 @@ class Parser {
     } else if (tok.typ === 'Ident') {
       const qn = tok.value || '';
       if (!qn.includes(':')) failInvalidKeywordLikeIdent(this.fail.bind(this), tok, qn);
-      assertValidQNamePrefix(qn.split(':', 1)[0], this.fail.bind(this), tok, 'BASE directive IRI');
+      assertValidQNamePrefix(qn.slice(0, qn.indexOf(':')), this.fail.bind(this), tok, 'BASE directive IRI');
       iri = this.prefixes.expandQName(qn);
     } else {
       this.fail(`Expected IRI after BASE, got ${tok.toString()}`, tok);
@@ -347,14 +355,18 @@ class Parser {
       const name = val || '';
       if (name === 'a') {
         return internIri(RDF_NS + 'type');
-      } else if (name.startsWith('_:')) {
+      }
+      const sep = name.indexOf(':');
+      if (sep === 1 && name.charCodeAt(0) === 95) {
         return new Blank(name);
-      } else if (name.includes(':')) {
-        assertValidQNamePrefix(name.split(':', 1)[0], this.fail.bind(this), tok);
-        return internIri(this.prefixes.expandQName(name));
-      } else {
-        failInvalidKeywordLikeIdent(this.fail.bind(this), tok, name);
       }
+      if (sep >= 0) {
+        const prefixName = name.slice(0, sep);
+        assertValidQNamePrefix(prefixName, this.fail.bind(this), tok);
+        const base = this.prefixes.map[prefixName] || '';
+        return internIri(base ? base + name.slice(sep + 1) : name);
+      }
+      failInvalidKeywordLikeIdent(this.fail.bind(this), tok, name);
     }
 
     if (typ === 'Literal') {
@@ -385,7 +397,7 @@ class Parser {
         } else if (dtTok.typ === 'Ident') {
           const qn = dtTok.value || '';
           if (!qn.includes(':')) failInvalidKeywordLikeIdent(this.fail.bind(this), dtTok, qn);
-          assertValidQNamePrefix(qn.split(':', 1)[0], this.fail.bind(this), dtTok, 'datatype prefixed name');
+          assertValidQNamePrefix(qn.slice(0, qn.indexOf(':')), this.fail.bind(this), dtTok, 'datatype prefixed name');
           dtIri = this.prefixes.expandQName(qn);
         } else {
           this.fail(`Expected datatype after ^^, got ${dtTok.toString()}`, dtTok);

package/lib/prelude.js

@@ -250,21 +250,40 @@ function literalParts(lit) {
 // equality fast-paths than repeated string key construction.
 
 let __nextTid = 1;
-const __tidIntern = new Map(); // string key -> number
+const __tidIntern = new Map(); // legacy generic key -> number
+const __iriTidIntern = new Map(); // IRI value -> number
+const __blankTidIntern = new Map(); // blank label -> number
+const __literalTidIntern = new Map(); // normalized literal lexical form -> number
 
 // Avoid storing extremely large literal keys in the global term-id intern map.
 // For huge literals we still assign a unique __tid, but we do not intern the key.
 const MAX_LITERAL_TID_LEN = 1024;
 
-function __getTid(key) {
-  let id = __tidIntern.get(key);
+function __getTidFromMap(map, key) {
+  let id = map.get(key);
   if (!id) {
     id = __nextTid++;
-    __tidIntern.set(key, id);
+    map.set(key, id);
   }
   return id;
 }
 
+function __getTid(key) {
+  return __getTidFromMap(__tidIntern, key);
+}
+
+function __getIriTid(value) {
+  return __getTidFromMap(__iriTidIntern, value);
+}
+
+function __getBlankTid(label) {
+  return __getTidFromMap(__blankTidIntern, label);
+}
+
+function __getLiteralTid(norm) {
+  return __getTidFromMap(__literalTidIntern, norm);
+}
+
 function __isQuotedLexical(lit) {
   if (typeof lit !== 'string') return false;
   if (lit.length >= 6) {
@@ -310,6 +329,14 @@ function __isPlainStringLiteralValue(lit) {
 function normalizeLiteralForTid(lit) {
   // Canonicalize so that plain string and explicit xsd:string share the same id.
   if (typeof lit !== 'string') return lit;
+
+  // Fast path for the overwhelmingly common lexer output for plain string
+  // literals: a canonical JSON-style quoted lexical form with no suffix.
+  // This avoids literalParts()/language-tag parsing for large fact tables.
+  if (lit.length >= 2 && lit.charCodeAt(0) === 34 && lit.charCodeAt(lit.length - 1) === 34 && lit.indexOf('^^') < 0) {
+    return `${lit}^^<${XSD_NS}string>`;
+  }
+
   const [lex, dt] = literalParts(lit);
   if (dt === XSD_NS + 'string') return `${lex}^^<${XSD_NS}string>`;
   if (dt === null && __isPlainStringLiteralValue(lit)) return `${lex}^^<${XSD_NS}string>`;
@@ -327,7 +354,7 @@ class Iri extends Term {
     super();
     this.value = value;
     Object.defineProperty(this, '__tid', {
-      value: __getTid('I:' + value),
+      value: __getIriTid(value),
       enumerable: false,
     });
   }
@@ -339,7 +366,7 @@ class Literal extends Term {
     this.value = value; // raw lexical form, e.g. "foo", 12, true, or "\"1944-08-21\"^^..."
     const norm = normalizeLiteralForTid(value);
     const useIntern = typeof norm === 'string' && norm.length <= MAX_LITERAL_TID_LEN;
-    const tid = useIntern ? __getTid('L:' + norm) : __nextTid++;
+    const tid = useIntern ? __getLiteralTid(norm) : __nextTid++;
     Object.defineProperty(this, '__tid', {
       value: tid,
       enumerable: false,
@@ -359,7 +386,7 @@ class Blank extends Term {
     super();
     this.label = label; // _:b1, etc.
     Object.defineProperty(this, '__tid', {
-      value: __getTid('B:' + label),
+      value: __getBlankTid(label),
       enumerable: false,
     });
   }

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "eyeling",
-  "version": "1.25.0",
+  "version": "1.25.2",
   "description": "A minimal Notation3 (N3) reasoner in JavaScript.",
   "main": "./index.js",
   "keywords": [