From 496f77a1e0d7417c479eb176ba713e6063924b69 Mon Sep 17 00:00:00 2001
From: Luan Taraschi <130802253+luantaraschi@users.noreply.github.com>
Date: Thu, 20 Aug 2026 19:54:56 -0300
Subject: [PATCH] Fix re.findall reporting a group that did not match

findall returns the matched text rather than a match object, so a group that
took no part in the match is reported as an empty value. With one group it
was reported as `None`:

    >>> re.findall(r"(a)?b", "b ab")
    [None, 'a']          # CPython: ['', 'a']

The branch for two or more groups was already right, since it passes `""` to
`Match.groups` as the default, so the two halves of the same function
disagreed with each other.

That default is built as a `str` whatever the pattern is, so a `bytes`
pattern came back with `str` mixed into it:

    >>> re.findall(rb"(a)|(b)", b"ab")
    [(b'a', ''), ('', b'b')]     # CPython: [(b'a', b''), (b'', b'b')]

The empty value is now built once from `isbytes` and both branches use it.
`Match.groups` still reports `None`, which is what CPython does.

Assisted-by: Claude Code:claude-opus-5
---
 crates/vm/src/stdlib/_sre.rs      | 16 ++++++++---
 extra_tests/snippets/stdlib_re.py | 44 +++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+), 3 deletions(-)

diff --git a/crates/vm/src/stdlib/_sre.rs b/crates/vm/src/stdlib/_sre.rs
index a9f98ca7015..4575901d3e8 100644
--- a/crates/vm/src/stdlib/_sre.rs
+++ b/crates/vm/src/stdlib/_sre.rs
@@ -452,15 +452,25 @@ mod _sre {
                 let mut match_list: Vec<PyObjectRef> = Vec::new();
                 let mut iter = SearchIter { req, state };
 
+                // What a group that took no part in the match is reported as.
+                // `findall` hands back the matched text rather than a match
+                // object, so the stand-in has to be an empty value of the type
+                // the pattern works on. `Match.groups` still reports `None` and
+                // is not affected by this.
+                let empty: PyObjectRef = if zelf.isbytes {
+                    vm.ctx.new_bytes(vec![]).into()
+                } else {
+                    vm.ctx.new_str(ascii!("")).into()
+                };
+
                 while iter.next().is_some() {
                     let m = Match::new(&mut iter.state, zelf.clone(), string_args.string.clone());
 
                     let item = if zelf.groups == 0 || zelf.groups == 1 {
                         m.get_slice(zelf.groups, s, vm)
-                            .unwrap_or_else(|| vm.ctx.none())
+                            .unwrap_or_else(|| empty.clone())
                     } else {
-                        m.groups(OptionalArg::Present(vm.ctx.new_str(ascii!("")).into()), vm)?
-                            .into()
+                        m.groups(OptionalArg::Present(empty.clone()), vm)?.into()
                     };
 
                     match_list.push(item);
diff --git a/extra_tests/snippets/stdlib_re.py b/extra_tests/snippets/stdlib_re.py
index 8613ddd30fc..80006aaf3e8 100644
--- a/extra_tests/snippets/stdlib_re.py
+++ b/extra_tests/snippets/stdlib_re.py
@@ -82,3 +82,47 @@
 
 # Combining characters; issue #7518
 assert not re.match(r"\w", "\u0345"), r"\w should not match U+0345 (category Mn)"
+
+
+def test_findall_group_that_did_not_participate():
+    # findall returns the matched text, not a match object, so a group that
+    # took no part in the match stands in as an empty value of the type the
+    # pattern works on. One group used to come back as None, and a bytes
+    # pattern used to mix str into its results.
+    assert re.findall(r"(a)?b", "b ab") == ["", "a"]
+    assert re.findall(r"(x)?", "a") == ["", ""]
+    assert re.findall(r"(a|b)?c", "c ac bc") == ["", "a", "b"]
+    assert re.findall(r"(?P<g>a)?b", "b ab") == ["", "a"]
+    assert re.compile(r"(a)?b").findall("b ab") == ["", "a"]
+
+    assert re.findall(rb"(a)?b", b"b ab") == [b"", b"a"]
+    assert re.findall(rb"(x)?", b"a") == [b"", b""]
+
+    # Two or more groups give a tuple per match, with the same stand-in.
+    assert re.findall(r"(a)|(b)", "ab") == [("a", ""), ("", "b")]
+    assert re.findall(rb"(a)|(b)", b"ab") == [(b"a", b""), (b"", b"b")]
+    assert re.findall(rb"(a)(b)?", b"a ab") == [(b"a", b""), (b"a", b"b")]
+
+    # The type is the pattern's, never the other one.
+    assert [type(x) for x in re.findall(r"(a)?b", "b ab")] == [str, str]
+    assert [type(x) for x in re.findall(rb"(a)?b", b"b ab")] == [bytes, bytes]
+    assert [type(y) for x in re.findall(rb"(a)|(b)", b"ab") for y in x] == [
+        bytes,
+        bytes,
+        bytes,
+        bytes,
+    ]
+
+    # A group that does participate, and no group at all, are unchanged.
+    assert re.findall(r"(a)", "aa") == ["a", "a"]
+    assert re.findall(r"a", "aa") == ["a", "a"]
+    assert re.findall(rb"a", b"aa") == [b"a", b"a"]
+
+    # A match object still reports None, which is where the difference lies.
+    assert re.match(r"(a)?b", "b").groups() == (None,)
+    assert re.match(r"(a)?b", "b").group(1) is None
+    assert [m.groups() for m in re.finditer(r"(a)?b", "b ab")] == [(None,), ("a",)]
+    assert re.split(r"(a)|(b)", "xaybz") == ["x", "a", None, "y", None, "b", "z"]
+
+
+test_findall_group_that_did_not_participate()