Fix valid character sets in Annex-B

Cargo.toml

@@ -29,6 +29,9 @@ index-positions = []
 # Prohibits all uses of unsafe code, for the paranoid.
 prohibit-unsafe = []
 
+# Enable EcmaScript's Annex-B RegExp syntax.
+annex-b = []
+
 [dependencies]
 hashbrown = "0.13.2"
 memchr = { version = "2.4.0", default-features = false }

src/parse.rs

@@ -416,39 +416,46 @@ where
             }
 
             // Parse a code point or character class.
-            let first = self.try_consume_bracket_class_atom()?;
-            if first.is_none() {
+            let Some(first) = self.try_consume_bracket_class_atom()? else {
                 continue;
-            }
+            };
 
             // Check for a dash; we may have a range.
             if !self.try_consume('-') {
-                add_class_atom(&mut result, first.unwrap());
+                add_class_atom(&mut result, first);
                 continue;
             }
 
-            let second = self.try_consume_bracket_class_atom()?;
-            if second.is_none() {
+            let Some(second) = self.try_consume_bracket_class_atom()? else {
                 // No second atom. For example: [a-].
-                add_class_atom(&mut result, first.unwrap());
+                add_class_atom(&mut result, first);
                 add_class_atom(&mut result, ClassAtom::CodePoint(u32::from('-')));
                 continue;
-            }
+            };
 
-            // Ranges can't contain character classes: [\d-z] is invalid.
             // Ranges must also be in order: z-a is invalid.
             // ES6 21.2.2.15.1 "If i > j, throw a SyntaxError exception"
-            match (first.unwrap(), second.unwrap()) {
-                (ClassAtom::CodePoint(c1), ClassAtom::CodePoint(c2)) if c1 <= c2 => {
-                    result.cps.add(Interval {
-                        first: c1 as u32,
-                        last: c2 as u32,
-                    })
-                }
-                _ => {
-                    return error("Invalid character range");
+            if let (ClassAtom::CodePoint(c1), ClassAtom::CodePoint(c2)) = (&first, &second) {
+                if c1 > c2 {
+                    return error(
+                        "Range values reversed, start char code is greater than end char code.",
+                    );
                 }
+                result.cps.add(Interval {
+                    first: *c1,
+                    last: *c2,
+                });
+                continue;
             }
+
+            if cfg!(not(feature = "annex-b")) || self.flags.unicode {
+                return error("Invalid character range");
+            }
+
+            // If it does not match a range treat as any match single characters.
+            add_class_atom(&mut result, first);
+            add_class_atom(&mut result, ClassAtom::CodePoint(u32::from('-')));
+            add_class_atom(&mut result, second);
         }
     }
 

tests/common/mod.rs

@@ -82,6 +82,7 @@ impl TestCompiledRegex {
     }
 
     /// Match against a string, returning the first formatted match.
+    #[track_caller]
     pub fn match1f(&self, input: &str) -> String {
         match self.find(input) {
             Some(m) => format_match(&m, input),
@@ -116,11 +117,13 @@ impl TestCompiledRegex {
     }
 
     /// Test that matching against \p input fails.
+    #[track_caller]
     pub fn test_fails(&self, input: &str) {
         assert!(self.find(input).is_none(), "Should not have matched")
     }
 
     /// Test that matching against \p input succeeds.
+    #[track_caller]
     pub fn test_succeeds(&self, input: &str) {
         assert!(self.find(input).is_some(), "Should have matched")
     }
@@ -184,6 +187,7 @@ impl TestConfig {
     }
 
     /// Compile a pattern to a regex, with given flags.
+    #[track_caller]
     pub fn compilef(&self, pattern: &str, flags_str: &str) -> TestCompiledRegex {
         let mut flags = regress::Flags::from(flags_str);
         flags.no_opt = !self.optimize;
@@ -204,6 +208,7 @@ impl TestConfig {
 
     /// Test that \p pattern and \p flags successfully parses, and matches
     /// \p input.
+    #[track_caller]
     pub fn test_match_succeeds(&self, pattern: &str, flags_str: &str, input: &str) {
         let cr = self.compilef(pattern, flags_str);
         cr.test_succeeds(input)

tests/syntax_error_tests.rs

@@ -1,3 +1,4 @@
+#[track_caller]
 fn test_1_error(pattern: &str, expected_err: &str) {
     let res = regress::Regex::with_flags(pattern, "u");
     assert!(res.is_err(), "Pattern should not have parsed: {}", pattern);
@@ -39,8 +40,13 @@ fn test_syntax_errors() {
     test_1_error(r"(?!", "Unbalanced parenthesis");
     test_1_error(r"abc)", "Unbalanced parenthesis");
 
-    test_1_error(r"[z-a]", "Invalid character range");
-    test_1_error(r"[\d-z]", "Invalid character range");
+    test_1_error(
+        r"[z-a]",
+        "Range values reversed, start char code is greater than end char code.",
+    );
+
+    // In unicode mode this is not allowed.
+    test_1_error(r"[a-\s]", "Invalid character range");
 
     test_1_error("\\", "Incomplete escape");
 

tests/tests.rs

@@ -1411,3 +1411,30 @@ fn unicode_escape_property_binary_assigned_tc(tc: TestConfig) {
         }
     }
 }
+
+#[test]
+#[cfg(feature = "annex-b")]
+fn test_valid_character_sets() {
+    test_with_configs(test_valid_character_sets_tc)
+}
+
+#[rustfmt::skip]
+#[cfg(feature = "annex-b")]
+fn test_valid_character_sets_tc(tc: TestConfig) {
+    // From: https://github.com/boa-dev/boa/issues/2794
+    let regexp = r"[a-\s]";
+    tc.test_match_succeeds(regexp, "", "a");
+    tc.test_match_succeeds(regexp, "", "-");
+    tc.test_match_succeeds(regexp, "", " ");
+    tc.test_match_fails(regexp, "", "s");
+    tc.test_match_fails(regexp, "", "$");
+
+    let regexp = r"[\d-z]";
+    tc.test_match_succeeds(regexp, "", "z");
+    tc.test_match_succeeds(regexp, "", "1");
+    tc.test_match_succeeds(regexp, "", "7");
+    tc.test_match_succeeds(regexp, "", "9");
+    tc.test_match_fails(regexp, "", "a");
+    tc.test_match_fails(regexp, "", "f");
+    tc.test_match_fails(regexp, "", " ");
+}