Error for unsupported grapheme boundary

README.md

@@ -384,7 +384,7 @@ Notice that nearly every feature below has at least subtle differences from Java
     <td>
       ✔ Allows 1 hex digit<br>
       ✔ Above <code>7F</code>, is UTF-8 encoded byte (unlike JS)<br>
-      ✔ Error for invalid encoded bytes<br>
+      ✔ Error for invalid encoded byte/sequence<br>
     </td>
   </tr>
   <tr valign="top">
@@ -602,7 +602,7 @@ Notice that nearly every feature below has at least subtle differences from Java
   </tr>
 
   <tr valign="top">
-    <th align="left" rowspan="6">Assertions</th>
+    <th align="left" rowspan="7">Assertions</th>
     <td>Line start, end</td>
     <td><code>^</code>, <code>$</code></td>
     <td align="middle">✅</td>
@@ -639,15 +639,6 @@ Notice that nearly every feature below has at least subtle differences from Java
       ● Common uses supported<br>
     </td>
   </tr>
-  <tr valign="top">
-    <td>Word boundary</td>
-    <td><code>\b</code>, <code>\B</code></td>
-    <td align="middle">✅</td>
-    <td align="middle">✅</td>
-    <td>
-      ✔ Unicode based (unlike JS)<br>
-    </td>
-  </tr>
   <tr valign="top">
     <td>Lookaround</td>
     <td>
@@ -663,6 +654,24 @@ Notice that nearly every feature below has at least subtle differences from Java
       ✔ Allows variable-length quantifiers and alternation within lookbehind<br>
     </td>
   </tr>
+  <tr valign="top">
+    <td>Word boundary</td>
+    <td><code>\b</code>, <code>\B</code></td>
+    <td align="middle">✅</td>
+    <td align="middle">✅</td>
+    <td>
+      ✔ Unicode based (unlike JS)<br>
+    </td>
+  </tr>
+  <tr valign="top">
+    <td>Grapheme boundary (extremely rare)</td>
+    <td><code>\y</code>, <code>\Y</code></td>
+    <td align="middle">❌</td>
+    <td align="middle">❌</td>
+    <td>
+      ● Not yet supported<br>
+    </td>
+  </tr>
 
   <tr valign="top">
     <th align="left" rowspan="3">Quantifiers</th>

spec/match-assertion.spec.js

@@ -8,7 +8,7 @@ beforeEach(() => {
 });
 
 describe('Assertion', () => {
-  // [Note] For kinds `lookahead` and `lookbehind`, see `match-lookaround.spec.js`
+  // [NOTE] For kinds `lookahead` and `lookbehind`, see `match-lookaround.spec.js`
 
   describe('line_end', () => {
     it('should match at the end of the string', () => {
@@ -341,4 +341,12 @@ describe('Assertion', () => {
       });
     });
   });
+
+  describe('grapheme_boundary', () => {
+    // Supportable with close approximation, but extremely rare and not many use cases
+    it('should throw as unsupported', () => {
+      expect(() => toDetails(r`\y`)).toThrow();
+      expect(() => toDetails(r`\Y`)).toThrow();
+    });
+  });
 });

spec/match-char-class.spec.js

@@ -41,7 +41,7 @@ describe('CharacterClass', () => {
         expect(cp(0o177)).toExactlyMatch(r`[\177]`);
       });
 
-      it(r`should throw for UTF-8 encoded byte sequence in octal (above \177)`, () => {
+      it(r`should throw for octal UTF-8 encoded byte above \177`, () => {
         expect(() => toDetails(r`[\200]`)).toThrow();
         expect(() => toDetails(r`[\777]`)).toThrow();
       });
@@ -72,4 +72,5 @@ describe('CharacterClass', () => {
 
   // TODO: Add remaining
   // TODO: Test that nested negated classes throw for target ES2018
+  // TODO: Test assertions/var-length escapes are identity escapes
 });

spec/match-char.spec.js

@@ -136,7 +136,7 @@ describe('Character', () => {
       expect(cp(0o177)).toExactlyMatch(r`\177`);
     });
 
-    it(r`should throw for UTF-8 encoded byte sequence in octal (above \177)`, () => {
+    it(r`should throw for octal UTF-8 encoded byte above \177`, () => {
       expect(() => toDetails(r`\200`)).toThrow();
       expect(() => toDetails(r`\777`)).toThrow();
     });

src/tokenize.js

@@ -260,6 +260,10 @@ function getTokenWithDetails(context, pattern, m, lastIndex) {
         }),
       };
     }
+    // Grapheme boundaries not yet unsupported; avoid treating as an identity escape
+    if ('yY'.includes(m1)) {
+      throw new Error(`Unsupported grapheme boundary "${m}"`);
+    }
     // Run last since it assumes an identity escape as final condition
     const result = createTokenForSharedEscape(m, {inCharClass: false});
     return Array.isArray(result) ? {tokens: result} : {token: result};
@@ -523,7 +527,7 @@ function createTokenForSharedEscape(raw, {inCharClass}) {
   }
   // Meta `\M-x` and `\M-\C-x` are unsupported; avoid treating as an identity escape
   if (char1 === 'M') {
-    // [TODO] Supportable; see <github.com/kkos/oniguruma/blob/master/doc/SYNTAX.md#12-onig_syn_op2_esc_capital_m_bar_meta-enable-m-x>, <github.com/kkos/oniguruma/blob/43a8c3f3daf263091f3a74019d4b32ebb6417093/src/regparse.c#L4695>
+    // [TODO] Supportable; see <github.com/kkos/oniguruma/blob/master/doc/SYNTAX.md#12-onig_syn_op2_esc_capital_m_bar_meta-enable-m-x>, <github.com/kkos/oniguruma/blob/43a8c3f3daf263091f3a74019d4b32ebb6417093/src/regparse.c#L4695>, <https://github.com/ammar/regexp_parser/blob/8851030feda68223d74f502335fb254a20d77016/lib/regexp_parser/expression/classes/escape_sequence.rb#L75>
     throw new Error(`Unsupported meta "${raw}"`);
   }
   // Identity escape; count code point length
@@ -663,11 +667,11 @@ function getFlagPropsForToken(flags) {
 function getValidatedHexCharCode(raw) {
   // Note: Onig (tested 6.9.8) has a bug where bare `\u` and `\x` are identity escapes if they
   // appear at the very end of the pattern, so e.g. `\u` matches `u`, but `\u0`, `\u.`, and `[\u]`
-  // are all errors, and `\x.` and `[\x]` are not errors but fail to match anything. Don't emulate
-  // these bugs, and just treat these cases as errors. Also, Onig treats incomplete `\x{` (with the
-  // brace and not immediately followed by a hex digit) as an identity escape, so e.g. `\x{`
-  // matches `x{` and `^\x{,2}$` matches `xx`, but `\x{2,}` and `\x{0,2}` are errors. Don't emulate
-  // this crazy and pointless ambiguity, and just treat incomplete `\x{` as an error
+  // are all errors, and `\x.` and `[\x]` are not errors but seemingly fail to match anything.
+  // Don't emulate these bugs, and just treat these cases as errors. Also, Onig treats incomplete
+  // `\x{` (with the brace and not immediately followed by a hex digit) as an identity escape, so
+  // e.g. `\x{` matches `x{` and `^\x{,2}$` matches `xx`, but `\x{2,}` and `\x{0,2}` are errors.
+  // Don't emulate this nasty/pointless ambiguity; just treat incomplete `\x{` as an error
   if (/^(?:\\u(?!\p{AHex}{4})|\\x(?!\p{AHex}{1,2}|\{\p{AHex}{1,8}\}))/u.test(raw)) {
     throw new Error(`Incomplete or invalid escape "${raw}"`);
   }