[ArrowStringArray] PERF: implement ArrowStringArray._str_split

asv_bench/benchmarks/strings.py

@@ -235,13 +235,31 @@ class Split(Dtypes):
 
     def setup(self, dtype, expand):
         super().setup(dtype)
-        self.s = self.s.str.join("--")
+        self.s = self.s.str.join("-")
 
     def time_split(self, dtype, expand):
-        self.s.str.split("--", expand=expand)
+        self.s.str.split("-", expand=expand)
 
     def time_rsplit(self, dtype, expand):
-        self.s.str.rsplit("--", expand=expand)
+        self.s.str.rsplit("-", expand=expand)
+
+
+class SplitPattern(Dtypes):
+
+    params = (Dtypes.params, [None, "--"])
+    param_names = ["dtype", "pat"]
+
+    def setup(self, dtype, pat):
+        super().setup(dtype)
+        if pat is None:
+            pat = "   "
+        self.s = self.s.str.join(pat)
+
+    def time_split(self, dtype, pat):
+        self.s.str.split(pat)
+
+    def time_rsplit(self, dtype, pat):
+        self.s.str.rsplit(pat)
 
 
 class Extract(Dtypes):

pandas/core/arrays/string_arrow.py

@@ -904,6 +904,29 @@ def _str_lower(self):
     def _str_upper(self):
         return type(self)(pc.utf8_upper(self._data))
 
+    def _str_split(self, pat=None, n=-1, expand=False):
+        if pa_version_under3p0 or (pat is not None and len(pat) > 1):
+            return super()._str_split(pat=pat, n=n, expand=expand)
+
+        if n is None or n == 0:
+            n = -1
+
+        if pat is None:
+            result = pc.utf8_split_whitespace(self._data, max_splits=n)
+        else:
+            result = pc.split_pattern(self._data, pattern=pat, max_splits=n)
+
+        mask = np.array(result.is_null())
+        result = np.array(result)
+        result = lib.map_infer_mask(
+            result,
+            lambda x: x.tolist(),
+            mask.view(np.uint8),
+            na_value=self.dtype.na_value,
+            dtype=np.dtype(object),
+        )
+        return result
+
     def _str_strip(self, to_strip=None):
         if pa_version_under4p0:
             return super()._str_strip(to_strip)

pandas/tests/strings/test_split_partition.py

@@ -144,11 +144,14 @@ def test_split_blank_string(any_string_dtype):
 
 def test_split_noargs(any_string_dtype):
     # #1859
+    expected = ["Travis", "Oliphant"]
+
     s = Series(["Wes McKinney", "Travis  Oliphant"], dtype=any_string_dtype)
     result = s.str.split()
-    expected = ["Travis", "Oliphant"]
     assert result[1] == expected
-    result = s.str.rsplit()
+
+    s = Series(["Wes McKinney", "Travis  Oliphant", np.nan], dtype=any_string_dtype)
+    result = s.str.split()
     assert result[1] == expected