Improve unicode escape in regex

Author: HenryRLee

src/com/google/javascript/jscomp/regex/RegExpTree.java

@@ -399,7 +399,7 @@ private RegExpTree parseCharset() {
         CharRanges ieExplicits = CharRanges.EMPTY;
         while (pos < limit && pattern.charAt(pos) != ']') {
           char ch = pattern.charAt(pos);
-          char start;
+          int start;
           if (ch == '\\') {
             ++pos;
             char possibleGroupName = pattern.charAt(pos);
@@ -414,7 +414,7 @@ private RegExpTree parseCharset() {
             start = ch;
             ++pos;
           }
-          char end = start;
+          int end = start;
           if (pos + 1 < limit && pattern.charAt(pos) == '-'
               && pattern.charAt(pos + 1) != ']') {
             ++pos;
@@ -464,15 +464,20 @@ private RegExpTree parseCharset() {
        * contexts, so contexts must filter those instead.
        * E.g. '\b' means a different thing inside a charset than without.
        */
-      private char parseEscapeChar() {
+      private int parseEscapeChar() {
         char ch = pattern.charAt(pos++);
         switch (ch) {
           case 'b': return '\b';
           case 'f': return '\f';
           case 'n': return '\n';
           case 'r': return '\r';
           case 't': return '\t';
-          case 'u': return parseHex(4);
+          case 'u':
+            if (flags.contains("u") && pos < limit && pattern.charAt(pos) == '{') {
+              return parseUnicodeEscape();
+            } else {
+              return parseHex(4);
+            }
           case 'v': return '\u000b';
           case 'x': return parseHex(2);
           default:
@@ -599,7 +604,7 @@ private RegExpTree parseEscape() {
             ++pos;
             return new Charset(charGroup, CharRanges.EMPTY);
           }
-          return new Text("" + parseEscapeChar());
+          return new Text(new String(Character.toChars(parseEscapeChar())));
         }
       }
 
@@ -630,6 +635,42 @@ private char parseHex(int n) {
         return (char) result;
       }
 
+      private int parseUnicodeEscape() {
+        checkState(pattern.charAt(pos) == '{');
+        int start = pos++;
+        int result = 0;
+        char ch = pattern.charAt(pos);
+        if (ch == '}') {
+          throw new IllegalArgumentException("Invalid unicode escape: "
+              + pattern.substring(start, ++pos));
+        }
+        while (pos < limit) {
+          int digit;
+          ch = pattern.charAt(pos++);
+          if ('0' <= ch && ch <= '9') {
+            digit = ch - '0';
+          } else if ('a' <= ch && ch <= 'f') {
+            digit = ch + (10 - 'a');
+          } else if ('A' <= ch && ch <= 'F') {
+            digit = ch + (10 - 'A');
+          } else if (ch == '}') {
+            break;
+          } else {
+            throw new IllegalArgumentException("Invalid character in unicode escape: " + ch);
+          }
+          result = (result << 4) | digit;
+        }
+        if (ch != '}') {
+          throw new IllegalArgumentException("Malformed unicode escape: expected '}' after "
+              + pattern.substring(start, pos));
+        }
+        if (result > 0x10FFFF) {
+          throw new IllegalArgumentException("Unicode must not be greater than 0x10FFFF: "
+              + pattern.substring(start, pos));
+        }
+        return result;
+      }
+
       private boolean isRepetitionStart(char ch) {
         switch (ch) {
           case '?':

test/com/google/javascript/jscomp/parsing/ParserTest.java

@@ -4894,6 +4894,14 @@ public void testRegExpError() {
     parseError("/\b.\\/", "Expected '/' in regular expression literal");
   }
 
+  @Test
+  public void testRegExpUnicode() {
+    assertNodeEquality(parse("/\\u10fA/"), script(expr(regex("\\u10fA"))));
+    assertNodeEquality(parse("/\\u{10fA}/u"), script(expr(regex("\\u{10fA}", "u"))));
+    assertNodeEquality(parse("/\\u{1fA}/u"), script(expr(regex("\\u{1fA}", "u"))));
+    assertNodeEquality(parse("/\\u{10FFFF}/u"), script(expr(regex("\\u{10FFFF}", "u"))));
+  }
+
   @Test
   public void testRegExpFlags() {
     // Various valid combinations.
@@ -6556,6 +6564,10 @@ private static Node regex(String regex) {
     return new Node(Token.REGEXP, Node.newString(regex));
   }
 
+  private static Node regex(String regex, String flag) {
+    return new Node(Token.REGEXP, Node.newString(regex), Node.newString(flag));
+  }
+
   /**
    * Verify that the given code has the given parse errors.
    * @return If in IDE mode, returns a partial tree.

test/com/google/javascript/jscomp/regex/RegExpTreeTest.java

@@ -193,4 +193,30 @@ public void testBackreferencingTreatedAsStringIfNoGroup() {
     // (?: ) in expected output serves same purpose as above test
     assertRegexCompilesTo("[(?<foo>)]\\k<foo>", "", "(?:[()<>?fo]k)<foo>");
   }
+
+  @Test
+  public void testValidUnicodeEscape() {
+    assertRegexCompilesTo("\\u0061", "", "a");
+    assertRegexCompilesTo("\\u10b1", "u", "\\u10b1");
+    assertRegexCompilesTo("\\u{61}", "u", "a");
+    assertRegexCompilesTo("\\u{10b1}", "u", "\\u10b1");
+    assertRegexCompilesTo("\\u{1bc}", "u", "\\u01bc");
+    assertRegexCompilesTo("\\u{100A3}", "u", "\\ud800\\udca3");
+  }
+
+  @Test
+  public void testInvalidUnicodeEscape() {
+    assertRegexThrowsExceptionThat("\\u{a012", "u")
+        .hasMessageThat()
+        .isEqualTo("Malformed unicode escape: expected '}' after {a012");
+    assertRegexThrowsExceptionThat("\\u{}", "u")
+        .hasMessageThat()
+        .isEqualTo("Invalid unicode escape: {}");
+    assertRegexThrowsExceptionThat("\\u{10za}", "u")
+        .hasMessageThat()
+        .isEqualTo("Invalid character in unicode escape: z");
+    assertRegexThrowsExceptionThat("\\u{FFFFFF}", "u")
+        .hasMessageThat()
+        .isEqualTo("Unicode must not be greater than 0x10FFFF: {FFFFFF}");
+  }
 }