Bug: fix Series.str.split when 'regex=None' for series having 'pd.ArrowDtype(pa.string())' dtype

doc/source/whatsnew/v3.0.0.rst

@@ -526,6 +526,7 @@ Conversion
 
 Strings
 ^^^^^^^
+- Bug in :meth:`Series.str.split` would not treat ``pat`` as regex when ``regex=None`` for series having ``pd.ArrowDtype(pa.string())`` dtype (:issue:`58321`)
 - Bug in :meth:`Series.value_counts` would not respect ``sort=False`` for series having ``string`` dtype (:issue:`55224`)
 -
 

pandas/core/strings/accessor.py

@@ -920,6 +920,9 @@ def split(
             )
         if is_re(pat):
             regex = True
+        elif isinstance(pat, str) and regex is None:
+            # regex is None so link to old behavior #43563
+            regex = len(pat) != 1
         result = self._data.array._str_split(pat, n, expand, regex)
         if self._data.dtype == "category":
             dtype = self._data.dtype.categories.dtype

pandas/tests/extension/test_arrow.py

@@ -2296,6 +2296,27 @@ def test_str_split_pat_none(method):
     tm.assert_series_equal(result, expected)
 
 
+def test_str_split_regex_explicit():
+    # GH 58321
+    # adapted from tests/strings/test_split_partition.py
+    values = pd.Series("xxxjpgzzz.jpg", dtype=ArrowDtype(pa.string()))
+
+    # explicit regex = False split
+    result = values.str.split(r"\.jpg", regex=False)
+    exp = pd.Series(ArrowExtensionArray(pa.array([["xxxjpgzzz.jpg"]])))
+    tm.assert_series_equal(result, exp)
+
+    # non explicit regex split, pattern length == 1
+    result = values.str.split(r".")
+    exp = pd.Series(ArrowExtensionArray(pa.array([["xxxjpgzzz", "jpg"]])))
+    tm.assert_series_equal(result, exp)
+
+    # non explicit regex split, pattern length != 1
+    result = values.str.split(r".jpg")
+    exp = pd.Series(ArrowExtensionArray(pa.array([["xx", "zzz", ""]])))
+    tm.assert_series_equal(result, exp)
+
+
 def test_str_split():
     # GH 52401
     ser = pd.Series(["a1cbcb", "a2cbcb", None], dtype=ArrowDtype(pa.string()))