Fix ripgrep output parsing on Windows drive-letter paths

coworker/tools/search.py

@@ -134,16 +134,24 @@ def _rel(path: str, root: Path) -> str:
         return path
 
 
+# ripgrep --no-heading --line-number emits "<path>:<line>:<text>". A positional
+# split(":", 2) mis-parses on Windows, where the absolute path starts with a
+# drive letter ("C:\...") whose colon looks like the field separator. Anchor on
+# the numeric ":<line>:" instead, so both the drive-letter colon and any colons
+# inside the matched text stay intact.
+_RG_LINE = re.compile(r"(.+?):(\d+):(.*)")
+
+
 def _parse_rg(stdout: str, root: Path, n: int) -> dict[str, Any]:
     matches: list[dict[str, Any]] = []
     for line in stdout.splitlines():
-        parts = line.split(":", 2)
-        if len(parts) == 3:
-            f, ln, txt = parts
+        m = _RG_LINE.match(line)
+        if m:
+            f, ln, txt = m.group(1), m.group(2), m.group(3)
             matches.append(
                 {
                     "file": _rel(f, root),
-                    "line": int(ln) if ln.isdigit() else 0,
+                    "line": int(ln),
                     "text": txt[:300],
                 }
             )

tests/test_code_tools.py

@@ -13,7 +13,7 @@
 
 from coworker.tools.files import file_tools
 from coworker.tools.git import git_tools
-from coworker.tools.search import _py_grep, search_tools
+from coworker.tools.search import _parse_rg, _py_grep, search_tools
 from coworker.web.fetch import _html_to_text, make_web_fetch_tool
 
 
@@ -61,6 +61,23 @@ def test_ripgrep_uses_the_same_ignored_dirs_as_the_python_fallback(tmp_path, mon
         assert commands[0].index(f"!**/{ignored}/**") > user_glob
 
 
+def test_parse_rg_handles_windows_drive_letter_and_colons_in_text():
+    # ripgrep prints absolute paths; on Windows they begin with a drive letter
+    # ("C:\\..."), whose colon must not be mistaken for the path:line:text
+    # separator. The matched text can also contain colons. Hardcode a Windows
+    # line so this guards the parser on every platform, not just Windows CI.
+    from pathlib import Path
+
+    stdout = "C:\\Users\\me\\ws\\a.py:12:x = {'k': 'v'}\n"
+    out = _parse_rg(stdout, Path("C:\\Users\\me\\ws"), 100)
+
+    assert out["count"] == 1
+    m = out["matches"][0]
+    assert m["line"] == 12  # was 0 with the old split(":", 2)
+    assert m["text"] == "x = {'k': 'v'}"  # line number not leaked, colons kept
+    assert m["file"] != "C"  # the drive letter is not the filename
+
+
 def test_grep_rejects_path_escape(tmp_path):
     grep = search_tools(str(tmp_path))[0]
     assert "escapes" in grep(pattern="x", path="../..")["error"]