eyeprolog 1.2.22 → 1.2.23

package/package.json

@@ -3,7 +3,7 @@
   "publishConfig": {
     "access": "public"
   },
-  "version": "1.2.22",
+  "version": "1.2.23",
   "description": "EyeProlog turns facts and rules into answers and proofs.",
   "type": "module",
   "main": "./index.js",

package/src/iso.js

@@ -5,7 +5,8 @@ import {
   isDecimalInteger, listFromItems, numberTerm, numberTextFromDouble,
   properListItems, termIsGround, termToString, unify, variable, variantTerms,
 } from './term.js';
-import { createParserOperatorState, parseClauses, parseGoalText } from './parser.js';
+import { sameNumberValue } from './number-value.js';
+import { createParserOperatorState, parseClauses, parseGoalText, parseNumberTokenText } from './parser.js';
 import { formatTermForWrite } from './write.js';
 import { emptyTerminalSequence, expandDcgBody, isListOrPartialList, validateDcgEmbeddedGoals } from './dcg.js';
 
@@ -220,7 +221,8 @@ function subsumesTerm(general, specific, env) {
       }
       continue;
     }
-    if (left.type !== right.type || left.name !== right.name || left.arity !== right.arity) return false;
+    if (left.type !== right.type || left.arity !== right.arity) return false;
+    if (left.type === NUMBER ? !sameNumberValue(left.name, right.name) : left.name !== right.name) return false;
     for (let i = left.arity - 1; i >= 0; i--) pending.push([left.args[i], right.args[i]]);
   }
   return true;
@@ -239,7 +241,8 @@ function* nonIdentity({ goal, env }) {
 function identical(left, right, env) {
   left = deref(left, env);
   right = deref(right, env);
-  if (left.type !== right.type || left.name !== right.name || left.arity !== right.arity) return false;
+  if (left.type !== right.type || left.arity !== right.arity) return false;
+  if (left.type === NUMBER ? !sameNumberValue(left.name, right.name) : left.name !== right.name) return false;
   if (left.type === VAR) return left.name === right.name;
   for (let i = 0; i < left.arity; i++) if (!identical(left.args[i], right.args[i], env)) return false;
   return true;
@@ -1549,9 +1552,7 @@ function quotedNumberSign(text, start) {
 }
 
 function parseIsoNumber(text) {
-  // number_chars/2 accepts leading layout, but its character list must end
-  // with the numeric token itself rather than trailing layout text.
-  if (text.length === 0 || /[\u0009-\u000d\u0020]$/.test(text)) return null;
+  if (text.length === 0) return null;
   let position = skipNumberLayout(text, 0);
   let sign = '';
 
@@ -1587,10 +1588,7 @@ function parseIsoNumber(text) {
   // ISO floating-point syntax requires a decimal fraction before an exponent.
   if (/^-?\d+[eE][+-]?\d+$/.test(numericText)) return null;
   try {
-    const parsed = parseGoalText(`number_chars_value(${numericText})`);
-    if (parsed.type !== COMPOUND || parsed.name !== 'number_chars_value' ||
-        parsed.arity !== 1 || parsed.args[0].type !== NUMBER) return null;
-    const value = parsed.args[0];
+    const value = parseNumberTokenText(numericText);
     if (isDecimalInteger(value.name)) return numberTerm(BigInt(value.name).toString());
     const finite = Number(value.name);
     if (!Number.isFinite(finite)) return null;
@@ -1601,14 +1599,18 @@ function parseIsoNumber(text) {
 }
 
 function sameNumber(left, right) {
-  const leftInteger = isDecimalInteger(left.name);
-  const rightInteger = isDecimalInteger(right.name);
-  if (leftInteger || rightInteger) {
-    return leftInteger && rightInteger && BigInt(left.name) === BigInt(right.name);
+  return sameNumberValue(left.name, right.name);
+}
+
+function canonicalNumberText(value) {
+  if (isDecimalInteger(value.name)) return BigInt(value.name).toString();
+  const finite = Number(value.name);
+  if (Number.isFinite(finite) && /^-?\d+\.\d+(?:[eE][+-]?\d+)?$/.test(value.name)) {
+    return value.name;
   }
-  const leftValue = Number(left.name);
-  const rightValue = Number(right.name);
-  return Number.isFinite(leftValue) && Number.isFinite(rightValue) && leftValue === rightValue;
+  const text = numberTextFromDouble(finite);
+  if (text == null) throw new PrologError('representation_error(max_float)');
+  return text;
 }
 
 function numberListText(list, env, kind, valueIsBound) {
@@ -1650,7 +1652,7 @@ function numberListBuiltin(kind) {
         if (sameNumber(value, parsed)) yield next;
         return;
       }
-      const items = characters(value.name).map((ch) =>
+      const items = characters(canonicalNumberText(value)).map((ch) =>
         kind === 'chars' ? atom(ch) : numberTerm(ch.codePointAt(0)));
       if (unify(goal.args[1], listFromItems(items), next)) yield next;
       return;

package/src/number-value.js

@@ -0,0 +1,28 @@
+// Numeric value identity shared by term semantics and scalar indexes. Keep
+// integer and float terms distinct while ignoring insignificant spelling
+// differences within either ISO numeric type.
+const decimalInteger = (text) => /^-?\d+$/.test(text ?? '');
+
+const finiteFloat = (text) => {
+  if (!/^[+-]?(?:\d+(?:\.\d*)?|\.\d+)(?:[eE][+-]?\d+)?$/.test(text ?? '')) return null;
+  const value = Number(text);
+  return Number.isFinite(value) ? value : null;
+};
+
+export function sameNumberValue(left, right) {
+  const leftInteger = decimalInteger(left);
+  const rightInteger = decimalInteger(right);
+  if (leftInteger || rightInteger) {
+    return leftInteger && rightInteger && BigInt(left) === BigInt(right);
+  }
+  const leftValue = finiteFloat(left);
+  const rightValue = finiteFloat(right);
+  return leftValue != null && rightValue != null && leftValue === rightValue;
+}
+
+export function numberValueKey(text) {
+  if (decimalInteger(text)) return `integer:${BigInt(text)}`;
+  const value = finiteFloat(text);
+  if (value == null) return `invalid:${text}`;
+  return `float:${Object.is(value, -0) ? 0 : value}`;
+}

package/src/parser.js

@@ -433,6 +433,18 @@ class Parser {
         this.take();
         this.take();
         let value = this.take();
+        if (value) {
+          const firstCode = value.charCodeAt(0);
+          if (firstCode >= 0xd800 && firstCode <= 0xdbff) {
+            const secondCode = this.peek().charCodeAt(0);
+            if (secondCode < 0xdc00 || secondCode > 0xdfff) {
+              throw new Error(`parse line ${line}: bad character code constant`);
+            }
+            value += this.take();
+          } else if (firstCode >= 0xdc00 && firstCode <= 0xdfff) {
+            throw new Error(`parse line ${line}: bad character code constant`);
+          }
+        }
         if (!value || (value !== ' ' && isWhitespaceCode(value.charCodeAt(0)))) {
           throw new Error(`parse line ${line}: bad character code constant`);
         }
@@ -1294,6 +1306,22 @@ export function parseProgramText(source, options = {}) {
   return parseClauses(source, options);
 }
 
+export function parseNumberTokenText(text) {
+  const source = String(text ?? '');
+  const parser = new Parser(source, {
+    includeDefaultOperators: false,
+    sourceMetadata: false,
+  });
+  // Inspect the tokenizer position before requesting another token: advancing
+  // would skip trailing layout and make `3 ` or `3/**/` look complete. A
+  // literal space consumed by the character-code token `0' ` is already part
+  // of parser.pos and is therefore correctly accepted.
+  if (parser.token.type !== TOK.NUMBER || parser.pos !== source.length) {
+    throw new Error('not exactly one number token');
+  }
+  return numberTerm(parser.token.text);
+}
+
 export function parseGoalText(text, options = {}) {
   const clauses = parseClauses(`zz_goal((${text})).`, options);
   const head = clauses[0]?.head;

package/src/program.js

@@ -1,6 +1,7 @@
 // Program representation and clause indexing.
 // Indexes are deliberately conservative: they speed up common scalar arguments but never replace unification as the final check.
 import { ATOM, COMPOUND, VAR, Env, atom, compound, deref, flattenConjunction, isScalar, numberTerm, properListItems, termToString, variable } from './term.js';
+import { numberValueKey } from './number-value.js';
 import { formatTermForWrite } from './write.js';
 import {
   ISO_OPERATOR_DEFINITIONS,
@@ -1238,15 +1239,19 @@ function scalarBuckets(index, term) {
 }
 
 function argumentBucket(index, term) {
-  return scalarBuckets(index, term).get(term.name) ?? null;
+  return scalarBuckets(index, term).get(scalarBucketKey(term.type, term.name)) ?? null;
 }
 
 function addArgumentBucket(index, term, clause) {
-  addClauseBucket(scalarBuckets(index, term), term.name, clause);
+  addClauseBucket(scalarBuckets(index, term), scalarBucketKey(term.type, term.name), clause);
 }
 
 function scalarIndexKey(term) {
-  return `${term.type}\u0000${term.name}`;
+  return `${term.type}\u0000${scalarBucketKey(term.type, term.name)}`;
+}
+
+function scalarBucketKey(type, name) {
+  return type === 'number' ? numberValueKey(name) : name;
 }
 
 function addClauseBucket(buckets, key, clause) {
@@ -1276,7 +1281,7 @@ function indexCompactOne(index, type, name, clause, clauses = null, clausePositi
       index.sawScalar = true;
       if (clauses && clausePosition > 0) index.fallback = clauses.slice(0, clausePosition);
     }
-    addClauseBucket(compactScalarBuckets(index, type), name, clause);
+    addClauseBucket(compactScalarBuckets(index, type), scalarBucketKey(type, name), clause);
   } else if (index.sawScalar) {
     index.fallback.push(clause);
   }

package/src/solver.js

@@ -4,6 +4,7 @@ import {
   COMPOUND, Env, compound, copyResolved, deref, emptyList, flattenConjunction, freshTerm,
   numberTerm, numberTextFromDouble, termIsGround, termToString, unify, variantTerms,
 } from './term.js';
+import { sameNumberValue } from './number-value.js';
 import { PrologError, getStrictIsoRegistry } from './iso.js';
 import { getEyePrologRegistry } from './standard-library.js';
 import { selectClauseCandidates, selectClauseCandidatesForValues, selectGroundClauseCandidates } from './program.js';
@@ -1089,11 +1090,13 @@ function isScalarTerm(term) {
 }
 
 function sameScalarTerm(left, right) {
-  return isScalarTerm(left) && isScalarTerm(right) && left.type === right.type && left.name === right.name;
+  return isScalarTerm(left) && isScalarTerm(right) && left.type === right.type &&
+    (left.type === 'number' ? sameNumberValue(left.name, right.name) : left.name === right.name);
 }
 
 function sameGroundTerm(left, right) {
-  if (left?.type !== right?.type || left?.name !== right?.name) return false;
+  if (left?.type !== right?.type) return false;
+  if (left?.type === 'number' ? !sameNumberValue(left.name, right.name) : left?.name !== right?.name) return false;
   const arity = left.args?.length ?? 0;
   if (arity !== (right.args?.length ?? 0)) return false;
   for (let i = 0; i < arity; i++) if (!sameGroundTerm(left.args[i], right.args[i])) return false;

package/src/term.js

@@ -1,5 +1,7 @@
 // Term model, environments, unification, readback, and ordering helpers.
-// This file is intentionally dependency-free because nearly every other module imports it.
+// Keep dependencies minimal because nearly every other module imports this file.
+import { sameNumberValue } from './number-value.js';
+
 export const VAR = 'var';
 export const ATOM = 'atom';
 export const STRING = 'string';
@@ -247,7 +249,7 @@ export function unify(left, right, env, options = {}) {
     }
 
     if (isScalar(a)) {
-      if (a.name !== b.name) return false;
+      if (a.type === NUMBER ? !sameNumberValue(a.name, b.name) : a.name !== b.name) return false;
       continue;
     }
 
@@ -497,7 +499,8 @@ export function variantTerms(left, leftEnv, right, rightEnv, pairs = new Map(),
     reverse.set(right.name, left.name);
     return true;
   }
-  if (left.type !== right.type || left.name !== right.name || left.arity !== right.arity) return false;
+  if (left.type !== right.type || left.arity !== right.arity) return false;
+  if (left.type === NUMBER ? !sameNumberValue(left.name, right.name) : left.name !== right.name) return false;
   for (let i = 0; i < left.arity; i++) {
     if (!variantTerms(left.args[i], leftEnv, right.args[i], rightEnv, pairs, reverse)) return false;
   }

package/test/run-regression.mjs

@@ -567,6 +567,63 @@ c4 ?- call((!;1)).
         }
       },
     },
+    {
+      name: 'number conversion accepts ISO space and Unicode character-code constants exactly',
+      run: () => {
+        const source = String.raw`
+?- N = 0' .
+   N = 32.
+?- number_chars(N,"0' ").
+   N = 32.
+?- number_codes(N,[48,39,32]).
+   N = 32.
+?- N = 0'😀.
+   N = 128512.
+?- number_chars(N,"0'😀").
+   N = 128512.
+?- number_codes(N,[48,39,128512]).
+   N = 128512.
+?- number_codes(N,[48,39,34]).
+   N = 34.
+?- number_codes(N,[48,39,96]).
+   N = 96.
+?- number_codes(N,[48,39,92,92]).
+   N = 92.
+?- number_chars(01,Chars).
+   Chars = "1".
+?- number_codes(01,Codes).
+   Codes = [49].
+?- 1.2 = 1.20.
+   true.
+?- 1.2 == 1.20.
+   true.
+?- 1 = 1.0.
+   false.
+?- number_chars(1.20,C), number_chars(Y,C), 1.20 == Y.
+   C = "1.20", Y = 1.2.
+`;
+        const result = publicApi.runQuads(source);
+        assertEqual(result.total, 15, 'quad total');
+        assertEqual(result.passed, 15, 'quad passed');
+        assertEqual(result.stdout, 'quads: 15 run, 15 passed, 0 failed.\n', 'quad report');
+
+        for (const goal of [
+          'number_chars(N,"3/**/")',
+          'number_codes(N,[51,47,42,42,47])',
+          'number_chars(N,"0\'  ")',
+          'number_codes(N,[48,39,32,32])',
+        ]) {
+          let caught = null;
+          try {
+            publicApi.run('', { goal });
+          } catch (error) {
+            caught = error;
+          }
+          if (caught == null) throw new Error(`${goal} should reject trailing layout`);
+          assertIncludes(String(caught?.message ?? caught), 'syntax_error(number)', goal);
+        }
+      },
+    },
     {
       name: 'number conversion rejects parenthesized numeric terms',
       run: () => {

package/the-art-of-eyeprolog.md

@@ -5777,7 +5777,13 @@ therefore follow `-` directly because `%` cannot continue a graphic token; an
 adjacent bracketed comment in `-/**/1` remains a syntax error under the eager
 token-consumer rule. Decimal fractions and decimal exponents are supported;
 the apostrophe character code is written with a doubled apostrophe as `0'''`
-and has value 39. Trailing material and non-finite values are rejected. The
+and has value 39, while a literal space character code is `0' ` and has value
+32. Character-code constants consume one Unicode scalar, including a
+non-BMP character. Trailing layout, comments, and other material are rejected;
+bound integers are converted to their canonical decimal spelling, and
+non-finite values are rejected. Equivalent spellings of the same numeric type
+compare by value, preserving the standard conversion round trip, while integer
+and floating-point terms remain distinct. The
 regression gate vendors all 74 numbered cases from Ulrich Neumerkel's contemporary
 `number_chars/2` comparison, including the Cor.2 error-precedence cases;
 `number_codes/2` shares the same numeric parser and has mirrored coverage for