Return null for non-participating regex capture groups in strings::extract

cpp/src/strings/regex/regex.inl

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.  All rights reserved.
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 
@@ -468,8 +468,11 @@ __device__ __forceinline__ match_result reprog_device::extract(int32_t const thr
                                                                cudf::size_type end,
                                                                cudf::size_type const group_id) const
 {
-  end = begin.position() + 1;
-  return call_regexec(thread_idx, dstr, begin, end, group_id + 1);
+  end               = begin.position() + 1;
+  auto const result = call_regexec(thread_idx, dstr, begin, end, group_id + 1);
+  // a capture group that did not participate in the overall match
+  // (e.g. an unmatched optional group) has an invalid range
+  return (result && (result->first >= 0)) ? result : cuda::std::nullopt;
 }
 
 template <positional P>

cpp/tests/strings/extract_tests.cpp

@@ -1,5 +1,5 @@
 /*
- * SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION.
+ * SPDX-FileCopyrightText: Copyright (c) 2019-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
  * SPDX-License-Identifier: Apache-2.0
  */
 
@@ -266,6 +266,33 @@ TEST_F(StringsExtractTests, EmptyExtractTest)
   CUDF_TEST_EXPECT_TABLES_EQUAL(*results, table_expected);
 }
 
+TEST_F(StringsExtractTests, NonParticipatingGroup)
+{
+  auto input = cudf::test::strings_column_wrapper({"A1", "B2", "C"});
+  auto sv    = cudf::strings_column_view(input);
+
+  // the optional group does not participate in the match for "C" and must
+  // result in null, not an empty string
+  auto prog    = cudf::strings::regex_program::create("(\\D)(\\d)?");
+  auto results = cudf::strings::extract(sv, *prog);
+
+  std::vector<std::unique_ptr<cudf::column>> columns;
+  columns.push_back(cudf::test::strings_column_wrapper({"A", "B", "C"}).release());
+  columns.push_back(cudf::test::strings_column_wrapper({"1", "2", ""}, {1, 1, 0}).release());
+  auto expected = cudf::table(std::move(columns));
+  CUDF_TEST_EXPECT_TABLES_EQUAL(*results, expected);
+
+  // a group that participates with an empty match remains an empty string
+  auto prog2    = cudf::strings::regex_program::create("(\\D)(\\d*)");
+  auto results2 = cudf::strings::extract(sv, *prog2);
+
+  std::vector<std::unique_ptr<cudf::column>> columns2;
+  columns2.push_back(cudf::test::strings_column_wrapper({"A", "B", "C"}).release());
+  columns2.push_back(cudf::test::strings_column_wrapper({"1", "2", ""}).release());
+  auto expected2 = cudf::table(std::move(columns2));
+  CUDF_TEST_EXPECT_TABLES_EQUAL(*results2, expected2);
+}
+
 TEST_F(StringsExtractTests, ExtractAllTest)
 {
   std::vector<char const*> h_input(

python/cudf/cudf/tests/series/accessors/test_str.py

@@ -2530,6 +2530,19 @@ def test_string_extract(ps_gs, pat, expand, flags, flags_raise):
         assert_eq(expect, got)
 
 
+@pytest.mark.parametrize("pat", [r"(\D)(\d)?", r"(\D)(\d*)"])
+def test_string_extract_nonparticipating_group(pat):
+    # An optional group that does not participate in the match results in
+    # null, while a group that participates with an empty match remains "".
+    s = ["A1", "B2", "C"]
+    gs = cudf.Series(s)
+    ps = pd.Series(s)
+
+    expect = ps.str.extract(pat, expand=True)
+    got = gs.str.extract(pat, expand=True)
+    assert_eq(expect, got)
+
+
 def test_string_extract_named_groups():
     s = ["hello-123", "world-456", "goodbye-789"]
     gs = cudf.Series(s)