Merge pull request #26 from hildjj/fix-25

Correctly handle unicode ranges

.gitignore

@@ -5,3 +5,4 @@ examples/*.peggy
 examples/*.pest
 node_modules
 t.abnf
+.cert/

bin/abnf_gen.js

@@ -112,6 +112,6 @@ program
   })
   .parseAsync()
   .catch(er => {
-    console.error(er.message);
+    console.error(er);
     process.exit(1);
   });

lib/ast.js

@@ -25,6 +25,22 @@ const PEST_DELIMS = {
   rangeSep: "..",
 };
 
+const BEFORE_SURROGATES = 0xd7ff;
+const FIRST_HIGH = 0xd800;
+const FIRST_LOW = 0xdc00;
+const LAST_LOW = 0xdfff;
+const AFTER_SURROGATES = 0xe000;
+const LAST_BMP = 0xffff;
+const FIRST_ASTRAL = 0x10000;
+const LAST_ASTRAL = 0x10ffff;
+const SURROGATE_PAGE_SIZE = 0x400;
+const UTF16_RANGES = [
+  [0, BEFORE_SURROGATES],
+  // [FIRST_HIGH, LAST_LOW],  // Ignore this range
+  [AFTER_SURROGATES, LAST_BMP],
+  [FIRST_ASTRAL, LAST_ASTRAL],
+];
+
 /**
  * @typedef {object} FormatOptions
  * @prop {string} [format='peggy']
@@ -42,15 +58,33 @@ function slug(s) {
   return s.replace(/-/g, "_");
 }
 
-function str(s) {
-  return `"${s.replace(/[\\"\x00-\x19\x7f-\xff]/g, c => `\\${{
+function str(s, opts) {
+  s = `"${s.replace(/[\\"\x00-\x19\x7f-\xff]/g, c => `\\${{
     "\r": "r",
     "\n": "n",
     '"': '"',
     "\t": "t",
     "\v": "x0B",
     "\\": "\\",
   }[c] || `x${c.charCodeAt(0).toString(16).padStart(2, "0")}`}`)}"`;
+
+  if (opts?.format === "peggy") {
+    s = s.replace(
+      /[\u0100-\uffff]/,
+      c => `\\u${c.codePointAt(0).toString(16).padStart(4, 0)}`
+    );
+    s = s.replace(
+      /[\u{10000}-\u{10ffff}]/u,
+      c => `\\u{${c.codePointAt(0).toString(16)}}`
+    );
+  } else if (opts?.format === "pest") {
+    s = s.replace(
+      /[\u{ff}-\u{10ffff}]/u,
+      c => `\\u{${c.codePointAt(0).toString(16)}}`
+    );
+  }
+
+  return s;
 }
 
 function fromArray(opts, a, joiner, needed, parent) {
@@ -81,6 +115,34 @@ function delims(opts) {
   return badFormat(opts);
 }
 
+function surrogates(codePoint) {
+  return [
+    Math.floor((codePoint - FIRST_ASTRAL) / SURROGATE_PAGE_SIZE) + FIRST_HIGH,
+    ((codePoint - FIRST_ASTRAL) % SURROGATE_PAGE_SIZE) + FIRST_LOW,
+  ];
+}
+
+/**
+ * Partition the inclusive range [first, last] into chunks that are in one of
+ * the ranges specified in parts.  If you want to ignore a range, do not
+ * include it in parts.
+ *
+ * @param {number} first
+ * @param {number} last
+ * @param {[start: number, end: number][]} parts Ranges to check. The start
+ *   and end points are both inclusive.
+ * @returns {[start: number, end: number][]}
+ */
+function partitionRange(first, last, parts) {
+  const res = [];
+  for (const [start, end] of parts) {
+    if ((first <= end) && (last >= start)) { // Overlap with range?
+      res.push([Math.max(first, start), Math.min(last, end)]);
+    }
+  }
+  return res;
+}
+
 // Only exported for testing
 export class Base {
   constructor(type, loc, simple = true) {
@@ -132,7 +194,7 @@ export class CaseInsensitiveString extends Base {
   }
 
   toFormat(opts) {
-    const s = str(this.str);
+    const s = str(this.str, opts);
     if (this.str.match(/[a-z]/i)) {
       switch (opts.format) {
         case "peggy":
@@ -155,8 +217,8 @@ export class CaseSensitiveString extends Base {
     this.base = base;
   }
 
-  toFormat(_opts) {
-    return str(this.str);
+  toFormat(opts) {
+    return str(this.str, opts);
   }
 }
 
@@ -299,22 +361,78 @@ export class Range extends Base {
   }
 
   static create(base, first, last, loc, utf16 = true) {
-    if (utf16 && (first <= 0xffff) && (last === 0x10ffff)) {
-      // Special case "all high Unicode" since it shows up a lot
-      // This should be generalized.
+    if (first > last) {
+      throw new Error(`Range out of order ${first.toString(16)}-${last.toString(16)}`);
+    }
+    if (first === last) {
+      return new CaseSensitiveString(String.fromCodePoint(first), base, loc);
+    }
+    if (utf16 && (last > BEFORE_SURROGATES)) {
       const alts = [];
-      if (first < 0xd800) {
-        alts.push(new Range(base, first, 0xd7ff, loc));
-        alts.push(new Range(base, 0xe000, 0xffff, loc));
-      } else {
-        alts.push(new Range(base, first, 0xffff, loc));
+      for (const [start, end] of partitionRange(first, last, UTF16_RANGES)) {
+        if (start < FIRST_ASTRAL) {
+          // No surrogates needed
+          alts.push(Range.create(base, start, end, loc, false));
+        } else {
+          // Pure astral range.
+
+          // This code follows the logic in regenerate:
+          // https://github.com/mathiasbynens/regenerate/blob/11567339f40fd262435934d544885bc047cb4220/regenerate.js#L996
+          // I didn't use regenerate directly because:
+          // a) I only needed a small part of it
+          // b) Regenerate will only generate a string, which I then would have
+          //    to parse to get the info I needed out.
+          // I believe this use is within the spirit of the MIT license.
+          const [startH, startL] = surrogates(start);
+          const [endH, endL] = surrogates(end);
+          let complete = false;
+          if (
+            (startH === endH) || ((startL === FIRST_LOW) && (endL === LAST_LOW))
+          ) {
+            alts.push(new Concatenation([
+              Range.create(base, startH, endH, loc, false),
+              Range.create(base, startL, endL, loc, false),
+            ], loc));
+            complete = true;
+          } else {
+            // First part of range, where startL might be greater than FIRST_LOW
+            // May one day be combined with below if startL === FIRST_LOW
+            alts.push(new Concatenation([
+              new CaseSensitiveString(String.fromCodePoint(startH), base, loc),
+              Range.create(base, startL, LAST_LOW, loc, false),
+            ], loc));
+          }
+          if (!complete && (startH + 1 < endH)) {
+            if (endL === LAST_LOW) {
+              alts.push(new Concatenation([
+                Range.create(base, startH + 1, endH + 1, loc, false),
+                Range.create(base, FIRST_LOW, endL, loc, false),
+              ], loc));
+              complete = true;
+            } else {
+              alts.push(new Concatenation([
+                Range.create(base, startH + 1, endH - 1, loc, false),
+                Range.create(base, FIRST_LOW, LAST_LOW, loc, false),
+              ], loc));
+            }
+          }
+          if (!complete) {
+            alts.push(new Concatenation([
+              new CaseSensitiveString(String.fromCodePoint(endH), base, loc),
+              Range.create(base, FIRST_LOW, endL, loc, false),
+            ], loc));
+          }
+        }
       }
-      alts.push(new Concatenation([
-        new Range(base, 0xd800, 0xdbff, loc),
-        new Range(base, 0xdc00, 0xdfff, loc),
-      ], loc));
 
-      return new Alternation(alts, loc);
+      switch (alts.length) {
+        case 0:
+          throw new Error(`Range consists of all surrogates ${first.toString(16)}-${last.toString(16)}`);
+        case 1:
+          return alts[0];
+        default:
+          return new Alternation(alts, loc);
+      }
     }
     return new Range(base, first, last, loc);
   }
@@ -323,7 +441,7 @@ export class Range extends Base {
     if (opts.format === "peggy") {
       if (num <= 0xff) {
         return "\\x" + num.toString(16).padStart(2, 0);
-      } else if (num <= 0xffff) {
+      } else if (num <= LAST_BMP) {
         return "\\u" + num.toString(16).padStart(4, 0);
       } else {
         throw new Error(`0x${num.toString(16)} does not fit in UTF-16`);
@@ -339,6 +457,9 @@ export class Range extends Base {
     const { rangeBefore, rangeAfter, rangeSep } = delims(opts);
     const first = Range.escape(opts, this.first);
     const last = Range.escape(opts, this.last);
+    if ((opts.format === "peggy") && (this.first + 1 === this.last)) {
+      return `${rangeBefore}${first}${last}${rangeAfter}`;
+    }
     return `${rangeBefore}${first}${rangeSep}${last}${rangeAfter}`;
   }
 }

package.json

@@ -50,26 +50,20 @@
     "ci": "npm run coverage && npm run lint"
   },
   "dependencies": {
-    "commander": "^12.1.0",
+    "commander": "^13.0.0",
     "peggy": "^4.2.0"
   },
   "devDependencies": {
     "@peggyjs/coverage": "1.3.2",
-    "@peggyjs/eslint-config": "^5.0.1",
-    "@typescript-eslint/eslint-plugin": "^8.18.0",
-    "@typescript-eslint/parser": "^8.18.0",
+    "@peggyjs/eslint-config": "^5.0.3",
+    "@typescript-eslint/eslint-plugin": "^8.19.0",
+    "@typescript-eslint/parser": "^8.19.0",
     "ava": "6.2.0",
     "c8": "10.1.3",
-    "eslint": "^9.16.0",
+    "eslint": "^9.17.0",
     "typescript": "^5.7.2"
   },
-  "packageManager": "pnpm@9.15.0",
-  "pnpm": {
-    "overrides": {
-      "cross-spawn": "^7.0.6",
-      "@eslint/plugin-kit": "^0.2.4"
-    }
-  },
+  "packageManager": "pnpm@9.15.2",
   "engines": {
     "node": ">=18"
   }