[BUG] extract when matching an empty string will some times return a null

[revans2]

Describe the bug
There are cases when strings::extract will produce a null, when it should have matched an empty string and returned that.

Steps/Code to reproduce bug

diff --git a/cpp/tests/strings/extract_tests.cpp b/cpp/tests/strings/extract_tests.cpp
index 5f38c142f..816877ff1 100644
--- a/cpp/tests/strings/extract_tests.cpp
+++ b/cpp/tests/strings/extract_tests.cpp
@@ -28,6 +28,84 @@
 struct StringsExtractTests : public cudf::test::BaseFixture {
 };
 
+TEST_F(StringsExtractTests, EmptyExtractTest)
+{
+  std::vector<const char*> h_strings{
+    "First Last", "Joe Schmoe", "John Smith", "Jane Smith", "Beyonce", "Sting", nullptr, ""};
+
+  cudf::test::strings_column_wrapper strings(
+    h_strings.begin(),
+    h_strings.end(),
+    thrust::make_transform_iterator(h_strings.begin(), [](auto str) { return str != nullptr; }));
+  auto strings_view = cudf::strings_column_view(strings);
+
+  std::vector<const char*> h_expecteds{"First",
+                                       "Joe",
+                                       "John",
+                                       "Jane",
+                                       "Beyonce",
+                                       "Sting",
+                                       nullptr,
+                                       "",
+                                       "Last",
+                                       "Schmoe",
+                                       "Smith",
+                                       "Smith",
+                                       "",
+                                       "",
+                                       nullptr,
+                                       ""};
+
+  std::string pattern = "\\A(\\w*) ?(\\w*)\\Z";
+  auto results        = cudf::strings::extract(strings_view, pattern);
+
+  cudf::test::strings_column_wrapper expected1(
+    h_expecteds.data(),
+    h_expecteds.data() + h_strings.size(),
+    thrust::make_transform_iterator(h_expecteds.begin(), [](auto str) { return str != nullptr; }));
+  cudf::test::strings_column_wrapper expected2(
+    h_expecteds.data() + h_strings.size(),
+    h_expecteds.data() + h_expecteds.size(),
+    thrust::make_transform_iterator(h_expecteds.data() + h_strings.size(),
+                                    [](auto str) { return str != nullptr; }));
+  std::vector<std::unique_ptr<cudf::column>> columns;
+  columns.push_back(expected1.release());
+  columns.push_back(expected2.release());
+  cudf::experimental::table expected(std::move(columns));
+  cudf::test::expect_tables_equal(*results, expected);
+}
+
+TEST_F(StringsExtractTests, EmptyExtractTes2)
+{
+  std::vector<const char*> h_strings{
+    nullptr, "AAA", "AAA_A", "AAA_AAA_", "A__", ""};
+
+  cudf::test::strings_column_wrapper strings(
+    h_strings.begin(),
+    h_strings.end(),
+    thrust::make_transform_iterator(h_strings.begin(), [](auto str) { return str != nullptr; }));
+  auto strings_view = cudf::strings_column_view(strings);
+
+  std::vector<const char*> h_expecteds{nullptr,
+                                       "AAA",
+                                       "A",
+                                       "",
+                                       "",
+                                       ""};
+
+  std::string pattern = "([^_]*)\\Z";
+  auto results        = cudf::strings::extract(strings_view, pattern);
+
+  cudf::test::strings_column_wrapper expected1(
+    h_expecteds.data(),
+    h_expecteds.data() + h_strings.size(),
+    thrust::make_transform_iterator(h_expecteds.begin(), [](auto str) { return str != nullptr; }));
+  std::vector<std::unique_ptr<cudf::column>> columns;
+  columns.push_back(expected1.release());
+  cudf::experimental::table expected(std::move(columns));
+  cudf::test::expect_tables_equal(*results, expected);
+}
+
 TEST_F(StringsExtractTests, ExtractTest)
 {
   std::vector<const char*> h_strings{

Expected behavior
Return an empty string in all cases where it would match, and null for cases where it would not.

[davidwendt]

Looks like Pandas is returning empty strings as well:

>>> import pandas as pd
>>> ps = pd.Series(["AAA","A","","",""])
>>> ps.str.extract("([^_]*)\\Z",expand=True)
     0
0  AAA
1    A
2     
3     
4     
>>> import cudf
>>> ds = cudf.Series(["AAA","A","","",""])
>>> ds.str.extract("([^_]*)\\Z")
      0
0   AAA
1     A
2  None
3  None
4  None