Skip to content
Python

fix: handle Windows drive-letter paths in grep (ripgrep) output

Mohamed Abdeltawab0622e93bd8f8
diff --git a/coworker/tools/search.py b/coworker/tools/search.py
index ad489e1f..0a50984b 100644
--- a/coworker/tools/search.py
+++ b/coworker/tools/search.py
@@ -91,6 +91,14 @@ def search_tools(workspace: str) -> list:
"--line-number",
"--no-heading",
"--color=never",
+ # Always print the filename (even for a single-file target) and
+ # NUL-separate it from `line:text`, so parsing never has to guess
+ # where the path ends — a Windows drive-letter colon (C:\ws\a.py),
+ # or a colon inside the matched text, would otherwise be taken as a
+ # field separator (issue #17). Without --with-filename a single-file
+ # search emits no path and no NUL, and the parser would drop it.
+ "--with-filename",
+ "--null",
"--max-count",
str(n),
"-e",
@@ -135,18 +143,23 @@ def _rel(path: str, root: Path) -> str:
def _parse_rg(stdout: str, root: Path, n: int) -> dict[str, Any]:
+ # `rg --null` emits each match as `<path>\0<line>:<text>`. Splitting the path off
+ # on the NUL byte first means the colon inside a Windows drive letter (C:\...) is
+ # never confused with the line/text separators — the old `split(":", 2)` parsed
+ # `C:\ws\a.py:12:def f()` as file="C", line="\ws\a.py", text="12:def f()".
matches: list[dict[str, Any]] = []
for line in stdout.splitlines():
- parts = line.split(":", 2)
- if len(parts) == 3:
- f, ln, txt = parts
- matches.append(
- {
- "file": _rel(f, root),
- "line": int(ln) if ln.isdigit() else 0,
- "text": txt[:300],
- }
- )
+ path, sep, rest = line.partition("\0")
+ if not sep:
+ continue
+ ln, _, txt = rest.partition(":")
+ matches.append(
+ {
+ "file": _rel(path, root),
+ "line": int(ln) if ln.isdigit() else 0,
+ "text": txt[:300],
+ }
+ )
if len(matches) >= n:
break
return {"count": len(matches), "matches": matches}
diff --git a/tests/test_code_tools.py b/tests/test_code_tools.py
index 0edc83b2..cb9e14e9 100644
--- a/tests/test_code_tools.py
+++ b/tests/test_code_tools.py
@@ -13,7 +13,7 @@ import pytest
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
@@ -41,6 +41,24 @@ def test_grep_finds_matches_and_respects_glob(tmp_path):
assert only_py["matches"][0]["line"] == 1
+def test_parse_rg_handles_windows_drive_letter_paths():
+ # `rg --null` emits `<path>\0<line>:<text>`. A Windows absolute path leads with a
+ # drive-letter colon (C:\...); the old `split(":", 2)` mis-parsed it as file="C",
+ # line=0, text="12:...". Regression for issue #17 — asserts hold on every platform.
+ from pathlib import Path
+
+ stdout = "C:\\ws\\a.py\x0012:def hello():\n"
+ out = _parse_rg(stdout, Path("C:\\ws"), 100)
+ assert out["count"] == 1
+ m = out["matches"][0]
+ assert m["line"] == 12
+ assert m["text"] == "def hello():"
+ # the whole path survives (not truncated to the drive letter "C"); .endswith
+ # stays platform-independent — on Linux _rel can't relativize a `C:\` path, so it
+ # returns the raw string, which still ends with the filename.
+ assert m["file"].endswith("a.py")
+
+
def test_ripgrep_uses_the_same_ignored_dirs_as_the_python_fallback(tmp_path, monkeypatch):
import coworker.tools.search as search