globset-0.4.13

We currently implement globs by converting them to regexes, and in doing
so, sometimes use grouping. In all but one case, we used non-capturing
groups. But for alternations, we used capturing groups, which was likely
just an oversight. We don't make use of capture groups at all, and while
they usually don't have any overhead, they lead to weird cases like this
one: https://github.com/rust-lang/regex/issues/1059

That particular issue is also a bug in the regex crate itself, which is
fixed in https://github.com/rust-lang/regex/pull/1062. Note though that
the bug fix in the regex crate is required. Even with this patch to
globset, memory usage is reduced (by about half in rust-lang/regex#1059)
but is not returned to where it was prior to the regex 1.9 release.

CHANGELOG.md

@@ -38,6 +38,8 @@ Bug fixes:
   Fix bug when using inline regex flags with `-e/--regexp`.
 * [BUG #2523](https://github.com/BurntSushi/ripgrep/issues/2523):
   Make executable searching take `.com` into account on Windows.
+* [BUG #2574](https://github.com/BurntSushi/ripgrep/issues/2574):
+  Fix bug in `-w/--word-regexp` that would result in incorrect match offsets.
 
 
 13.0.0 (2021-06-12)

Cargo.lock

@@ -119,7 +119,7 @@ checksum = "d2fabcfbdc87f4758337ca535fb41a6d701b65693ce38287d856d1674551ec9b"
 
 [[package]]
 name = "globset"
-version = "0.4.12"
+version = "0.4.13"
 dependencies = [
  "aho-corasick",
  "bstr",

crates/globset/Cargo.toml

@@ -1,6 +1,6 @@
 [package]
 name = "globset"
-version = "0.4.12"  #:version
+version = "0.4.13"  #:version
 authors = ["Andrew Gallant <jamslam@gmail.com>"]
 description = """
 Cross platform single glob and glob set matching. Glob set matching is the

crates/globset/src/glob.rs

@@ -736,7 +736,7 @@ impl Tokens {
                     // It is possible to have an empty set in which case the
                     // resulting alternation '()' would be an error.
                     if !parts.is_empty() {
-                        re.push('(');
+                        re.push_str("(?:");
                         re.push_str(&parts.join("|"));
                         re.push(')');
                     }
@@ -1276,6 +1276,7 @@ mod tests {
     toregex!(re32, "/a**", r"^/a.*.*$");
     toregex!(re33, "/**a", r"^/.*.*a$");
     toregex!(re34, "/a**b", r"^/a.*.*b$");
+    toregex!(re35, "{a,b}", r"^(?:b|a)$");
 
     matches!(match1, "a", "a");
     matches!(match2, "a*b", "a_b");

crates/ignore/src/default_types.rs

@@ -55,6 +55,7 @@ pub const DEFAULT_TYPES: &[(&[&str], &[&str])] = &[
     (&["cs"], &["*.cs"]),
     (&["csharp"], &["*.cs"]),
     (&["cshtml"], &["*.cshtml"]),
+    (&["csproj"], &["*.csproj"]),
     (&["css"], &["*.css", "*.scss"]),
     (&["csv"], &["*.csv"]),
     (&["cuda"], &["*.cu", "*.cuh"]),

crates/regex/src/word.rs

@@ -128,6 +128,9 @@ impl WordMatcher {
         // The reason why we cannot handle the ^/$ cases here is because we
         // can't assume anything about the original pattern. (Try commenting
         // out the checks for ^/$ below and run the tests to see examples.)
+        //
+        // NOTE(2023-07-31): After fixing #2574, this logic honestly still
+        // doesn't seem correct. Regex composition is hard.
         let input = Input::new(haystack).span(at..haystack.len());
         let mut cand = match self.regex.find(input) {
             None => return Ok(None),
@@ -136,8 +139,17 @@ impl WordMatcher {
         if cand.start() == 0 || cand.end() == haystack.len() {
             return Err(());
         }
-        let (_, slen) = bstr::decode_utf8(&haystack[cand]);
-        let (_, elen) = bstr::decode_last_utf8(&haystack[cand]);
+        // We decode the chars on either side of the match. If either char is
+        // a word character, then that means the ^/$ matched and not \W. In
+        // that case, we defer to the slower engine.
+        let (ch, slen) = bstr::decode_utf8(&haystack[cand]);
+        if ch.map_or(true, regex_syntax::is_word_character) {
+            return Err(());
+        }
+        let (ch, elen) = bstr::decode_last_utf8(&haystack[cand]);
+        if ch.map_or(true, regex_syntax::is_word_character) {
+            return Err(());
+        }
         let new_start = cand.start() + slen;
         let new_end = cand.end() - elen;
         // This occurs the original regex can match the empty string. In this

tests/regression.rs

@@ -1173,3 +1173,18 @@ rgtest!(r2480, |dir: Dir, mut cmd: TestCommand| {
     cmd.args(&["--only-matching", "-e", "(?i)notfoo", "-e", "bar", "file"]);
     cmd.assert_err();
 });
+
+// See: https://github.com/BurntSushi/ripgrep/issues/2574
+rgtest!(r2574, |dir: Dir, mut cmd: TestCommand| {
+    dir.create("haystack", "some.domain.com\nsome.domain.com/x\n");
+    let got = cmd
+        .args(&[
+            "--no-filename",
+            "--no-unicode",
+            "-w",
+            "-o",
+            r"(\w+\.)*domain\.(\w+)",
+        ])
+        .stdout();
+    eqnice!("some.domain.com\nsome.domain.com\n", got);
+});