Move conditional branch out of infer_dtype_from()

Author: kotamatsuoka

pandas/core/dtypes/cast.py

@@ -660,9 +660,7 @@ def infer_dtype_from_scalar(val, pandas_dtype: bool = False) -> Tuple[DtypeObj,
         # instead of np.empty (but then you still don't want things
         # coming out as np.str_!
 
-        from pandas.core.arrays.string_ import StringDtype
-
-        dtype = StringDtype()
+        dtype = object
 
     elif isinstance(val, (np.datetime64, datetime)):
         val = tslibs.Timestamp(val)

pandas/core/internals/blocks.py

@@ -45,7 +45,6 @@
     is_re,
     is_re_compilable,
     is_sparse,
-    is_string_dtype,
     is_timedelta64_dtype,
     pandas_dtype,
 )
@@ -73,6 +72,7 @@
     PandasDtype,
     TimedeltaArray,
 )
+from pandas.core.arrays.string_ import StringDtype
 from pandas.core.base import PandasObject
 import pandas.core.common as com
 from pandas.core.construction import extract_array
@@ -1008,6 +1008,9 @@ def coerce_to_target_dtype(self, other):
         # if we cannot then coerce to object
         dtype, _ = infer_dtype_from(other, pandas_dtype=True)
 
+        if isinstance(self.dtype, StringDtype):
+            dtype = StringDtype()
+
         if is_dtype_equal(self.dtype, dtype):
             return self
 

pandas/tests/series/methods/test_replace.py

@@ -274,7 +274,7 @@ def test_replace_replacer_equals_replacement(self):
         # GH 20656
         # make sure all replacers are matching against original values
         s = pd.Series(["a", "b"])
-        expected = pd.Series(["b", "a"], dtype="string")
+        expected = pd.Series(["b", "a"])
         result = s.replace({"a": "b", "b": "a"})
         tm.assert_series_equal(expected, result)
 
@@ -346,24 +346,19 @@ def test_replace_with_no_overflowerror(self):
         tm.assert_series_equal(result, expected)
 
     @pytest.mark.parametrize(
-        "ser, to_replace, exp, dtype",
+        "ser, to_replace, exp",
         [
-            ([1, 2, 3], {1: 2, 2: 3, 3: 4}, [2, 3, 4], "int64"),
-            (
-                ["1", "2", "3"],
-                {"1": "2", "2": "3", "3": "4"},
-                ["2", "3", "4"],
-                "string",
-            ),
+            ([1, 2, 3], {1: 2, 2: 3, 3: 4}, [2, 3, 4],),
+            (["1", "2", "3"], {"1": "2", "2": "3", "3": "4"}, ["2", "3", "4"],),
         ],
     )
-    def test_replace_commutative(self, ser, to_replace, exp, dtype):
+    def test_replace_commutative(self, ser, to_replace, exp):
         # GH 16051
         # DataFrame.replace() overwrites when values are non-numeric
 
         series = pd.Series(ser)
 
-        expected = pd.Series(exp, dtype=dtype)
+        expected = pd.Series(exp)
         result = series.replace(to_replace)
 
         tm.assert_series_equal(result, expected)