DEPR: datetimelike inference with strings (#41731)

doc/source/whatsnew/v1.3.0.rst

@@ -698,6 +698,7 @@ Deprecations
 - Deprecated passing arguments (apart from ``value``) as positional in :meth:`DataFrame.fillna` and :meth:`Series.fillna` (:issue:`41485`)
 - Deprecated passing arguments as positional in :meth:`DataFrame.reset_index` (other than ``"level"``) and :meth:`Series.reset_index` (:issue:`41485`)
 - Deprecated construction of :class:`Series` or :class:`DataFrame` with ``DatetimeTZDtype`` data and ``datetime64[ns]`` dtype.  Use ``Series(data).dt.tz_localize(None)`` instead (:issue:`41555`,:issue:`33401`)
+- Deprecated inference of ``timedelta64[ns]``, ``datetime64[ns]``, or ``DatetimeTZDtype`` dtypes in :class:`Series` construction when data containing strings is passed and no ``dtype`` is passed (:issue:`33558`)
 - In a future version, constructing :class:`Series` or :class:`DataFrame` with ``datetime64[ns]`` data and ``DatetimeTZDtype`` will treat the data as wall-times instead of as UTC times (matching DatetimeIndex behavior). To treat the data as UTC times, use ``pd.Series(data).dt.tz_localize("UTC").dt.tz_convert(dtype.tz)`` or ``pd.Series(data.view("int64"), dtype=dtype)`` (:issue:`33401`)
 - Deprecated passing arguments as positional in :meth:`DataFrame.set_axis` and :meth:`Series.set_axis` (other than ``"labels"``) (:issue:`41485`)
 - Deprecated passing arguments as positional in :meth:`DataFrame.where` and :meth:`Series.where` (other than ``"cond"`` and ``"other"``) (:issue:`41485`)

pandas/_libs/lib.pyi

@@ -153,7 +153,7 @@ def ensure_string_array(
 
 def infer_datetimelike_array(
     arr: np.ndarray  # np.ndarray[object]
-) -> str: ...
+) -> tuple[str, bool]: ...
 
 def astype_intsafe(
     arr: np.ndarray,  # np.ndarray[object]

pandas/_libs/lib.pyx

@@ -1558,7 +1558,7 @@ def infer_dtype(value: object, skipna: bool = True) -> str:
     return "mixed"
 
 
-def infer_datetimelike_array(arr: ndarray[object]) -> str:
+def infer_datetimelike_array(arr: ndarray[object]) -> tuple[str, bool]:
     """
     Infer if we have a datetime or timedelta array.
     - date: we have *only* date and maybe strings, nulls
@@ -1576,19 +1576,21 @@ def infer_datetimelike_array(arr: ndarray[object]) -> str:
     Returns
     -------
     str: {datetime, timedelta, date, nat, mixed}
+    bool
     """
     cdef:
         Py_ssize_t i, n = len(arr)
         bint seen_timedelta = False, seen_date = False, seen_datetime = False
         bint seen_tz_aware = False, seen_tz_naive = False
-        bint seen_nat = False
+        bint seen_nat = False, seen_str = False
         list objs = []
         object v
 
     for i in range(n):
         v = arr[i]
         if isinstance(v, str):
             objs.append(v)
+            seen_str = True
 
             if len(objs) == 3:
                 break
@@ -1609,7 +1611,7 @@ def infer_datetimelike_array(arr: ndarray[object]) -> str:
                 seen_tz_aware = True
 
             if seen_tz_naive and seen_tz_aware:
-                return "mixed"
+                return "mixed", seen_str
         elif util.is_datetime64_object(v):
             # np.datetime64
             seen_datetime = True
@@ -1619,16 +1621,16 @@ def infer_datetimelike_array(arr: ndarray[object]) -> str:
             # timedelta, or timedelta64
             seen_timedelta = True
         else:
-            return "mixed"
+            return "mixed", seen_str
 
     if seen_date and not (seen_datetime or seen_timedelta):
-        return "date"
+        return "date", seen_str
     elif seen_datetime and not seen_timedelta:
-        return "datetime"
+        return "datetime", seen_str
     elif seen_timedelta and not seen_datetime:
-        return "timedelta"
+        return "timedelta", seen_str
     elif seen_nat:
-        return "nat"
+        return "nat", seen_str
 
     # short-circuit by trying to
     # actually convert these strings
@@ -1637,14 +1639,14 @@ def infer_datetimelike_array(arr: ndarray[object]) -> str:
     if len(objs):
         try:
             array_to_datetime(objs, errors="raise")
-            return "datetime"
+            return "datetime", seen_str
         except (ValueError, TypeError):
             pass
 
         # we are *not* going to infer from strings
         # for timedelta as too much ambiguity
 
-    return 'mixed'
+    return "mixed", seen_str
 
 
 cdef inline bint is_timedelta(object o):

pandas/core/dtypes/cast.py

@@ -1543,7 +1543,7 @@ def try_timedelta(v: np.ndarray) -> np.ndarray:
         else:
             return td_values.reshape(shape)
 
-    inferred_type = lib.infer_datetimelike_array(ensure_object(v))
+    inferred_type, seen_str = lib.infer_datetimelike_array(ensure_object(v))
 
     if inferred_type == "datetime":
         # error: Incompatible types in assignment (expression has type "ExtensionArray",
@@ -1572,6 +1572,15 @@ def try_timedelta(v: np.ndarray) -> np.ndarray:
                 # "ExtensionArray", variable has type "Union[ndarray, List[Any]]")
                 value = try_datetime(v)  # type: ignore[assignment]
 
+    if value.dtype.kind in ["m", "M"] and seen_str:
+        warnings.warn(
+            f"Inferring {value.dtype} from data containing strings is deprecated "
+            "and will be removed in a future version. To retain the old behavior "
+            "explicitly pass Series(data, dtype={value.dtype})",
+            FutureWarning,
+            stacklevel=find_stack_level(),
+        )
+        # return v.reshape(shape)
     return value
 
 

pandas/tests/apply/test_series_apply.py

@@ -859,7 +859,9 @@ def test_apply_to_timedelta():
     list_of_strings = ["00:00:01", np.nan, pd.NaT, pd.NaT]
 
     a = pd.to_timedelta(list_of_strings)  # noqa
-    b = Series(list_of_strings).apply(pd.to_timedelta)  # noqa
+    with tm.assert_produces_warning(FutureWarning, match="Inferring timedelta64"):
+        ser = Series(list_of_strings)
+    b = ser.apply(pd.to_timedelta)  # noqa
     # Can't compare until apply on a Series gives the correct dtype
     # assert_series_equal(a, b)
 

pandas/tests/arithmetic/test_datetime64.py

@@ -328,7 +328,7 @@ def test_dt64arr_timestamp_equality(self, box_with_array):
             box_with_array if box_with_array not in [pd.Index, pd.array] else np.ndarray
         )
 
-        ser = Series([Timestamp("2000-01-29 01:59:00"), Timestamp("2000-01-30"), "NaT"])
+        ser = Series([Timestamp("2000-01-29 01:59:00"), Timestamp("2000-01-30"), NaT])
         ser = tm.box_expected(ser, box_with_array)
 
         result = ser != ser

pandas/tests/dtypes/test_inference.py

@@ -1169,7 +1169,7 @@ def test_infer_dtype_period_with_na(self, na_value):
         ],
     )
     def test_infer_datetimelike_array_datetime(self, data):
-        assert lib.infer_datetimelike_array(data) == "datetime"
+        assert lib.infer_datetimelike_array(data) == ("datetime", False)
 
     @pytest.mark.parametrize(
         "data",
@@ -1181,11 +1181,11 @@ def test_infer_datetimelike_array_datetime(self, data):
         ],
     )
     def test_infer_datetimelike_array_timedelta(self, data):
-        assert lib.infer_datetimelike_array(data) == "timedelta"
+        assert lib.infer_datetimelike_array(data) == ("timedelta", False)
 
     def test_infer_datetimelike_array_date(self):
         arr = [date(2017, 6, 12), date(2017, 3, 11)]
-        assert lib.infer_datetimelike_array(arr) == "date"
+        assert lib.infer_datetimelike_array(arr) == ("date", False)
 
     @pytest.mark.parametrize(
         "data",
@@ -1200,7 +1200,7 @@ def test_infer_datetimelike_array_date(self):
         ],
     )
     def test_infer_datetimelike_array_mixed(self, data):
-        assert lib.infer_datetimelike_array(data) == "mixed"
+        assert lib.infer_datetimelike_array(data)[0] == "mixed"
 
     @pytest.mark.parametrize(
         "first, expected",
@@ -1218,7 +1218,7 @@ def test_infer_datetimelike_array_mixed(self, data):
     @pytest.mark.parametrize("second", [None, np.nan])
     def test_infer_datetimelike_array_nan_nat_like(self, first, second, expected):
         first.append(second)
-        assert lib.infer_datetimelike_array(first) == expected
+        assert lib.infer_datetimelike_array(first) == (expected, False)
 
     def test_infer_dtype_all_nan_nat_like(self):
         arr = np.array([np.nan, np.nan])

pandas/tests/resample/test_time_grouper.py

@@ -305,27 +305,30 @@ def test_groupby_resample_interpolate():
         .resample("1D")
         .interpolate(method="linear")
     )
-    expected_ind = pd.MultiIndex.from_tuples(
-        [
-            (50, "2018-01-07"),
-            (50, Timestamp("2018-01-08")),
-            (50, Timestamp("2018-01-09")),
-            (50, Timestamp("2018-01-10")),
-            (50, Timestamp("2018-01-11")),
-            (50, Timestamp("2018-01-12")),
-            (50, Timestamp("2018-01-13")),
-            (50, Timestamp("2018-01-14")),
-            (50, Timestamp("2018-01-15")),
-            (50, Timestamp("2018-01-16")),
-            (50, Timestamp("2018-01-17")),
-            (50, Timestamp("2018-01-18")),
-            (50, Timestamp("2018-01-19")),
-            (50, Timestamp("2018-01-20")),
-            (50, Timestamp("2018-01-21")),
-            (60, Timestamp("2018-01-14")),
-        ],
-        names=["volume", "week_starting"],
-    )
+
+    msg = "containing strings is deprecated"
+    with tm.assert_produces_warning(FutureWarning, match=msg):
+        expected_ind = pd.MultiIndex.from_tuples(
+            [
+                (50, "2018-01-07"),
+                (50, Timestamp("2018-01-08")),
+                (50, Timestamp("2018-01-09")),
+                (50, Timestamp("2018-01-10")),
+                (50, Timestamp("2018-01-11")),
+                (50, Timestamp("2018-01-12")),
+                (50, Timestamp("2018-01-13")),
+                (50, Timestamp("2018-01-14")),
+                (50, Timestamp("2018-01-15")),
+                (50, Timestamp("2018-01-16")),
+                (50, Timestamp("2018-01-17")),
+                (50, Timestamp("2018-01-18")),
+                (50, Timestamp("2018-01-19")),
+                (50, Timestamp("2018-01-20")),
+                (50, Timestamp("2018-01-21")),
+                (60, Timestamp("2018-01-14")),
+            ],
+            names=["volume", "week_starting"],
+        )
     expected = DataFrame(
         data={
             "price": [

pandas/tests/series/accessors/test_dt_accessor.py

@@ -679,6 +679,7 @@ def test_dt_timetz_accessor(self, tz_naive_fixture):
             [["2016-01-07", "2016-01-01"], [[2016, 1, 4], [2015, 53, 5]]],
         ],
     )
+    @pytest.mark.filterwarnings("ignore:Inferring datetime64:FutureWarning")
     def test_isocalendar(self, input_series, expected_output):
         result = pd.to_datetime(Series(input_series)).dt.isocalendar()
         expected_frame = DataFrame(

pandas/tests/series/methods/test_combine_first.py

@@ -78,7 +78,11 @@ def test_combine_first_dt64(self):
         s0 = to_datetime(Series(["2010", np.NaN]))
         s1 = Series([np.NaN, "2011"])
         rs = s0.combine_first(s1)
-        xp = Series([datetime(2010, 1, 1), "2011"])
+
+        msg = "containing strings is deprecated"
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            xp = Series([datetime(2010, 1, 1), "2011"])
+
         tm.assert_series_equal(rs, xp)
 
     def test_combine_first_dt_tz_values(self, tz_naive_fixture):

pandas/tests/series/methods/test_fillna.py

@@ -319,8 +319,11 @@ def test_datetime64_fillna(self):
 
         # GH#6587
         # make sure that we are treating as integer when filling
-        # this also tests inference of a datetime-like with NaT's
-        ser = Series([NaT, NaT, "2013-08-05 15:30:00.000001"])
+        msg = "containing strings is deprecated"
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            # this also tests inference of a datetime-like with NaT's
+            ser = Series([NaT, NaT, "2013-08-05 15:30:00.000001"])
+
         expected = Series(
             [
                 "2013-08-05 15:30:00.000001",

pandas/tests/series/test_constructors.py

@@ -900,14 +900,23 @@ def test_constructor_dtype_datetime64_7(self):
 
     def test_constructor_dtype_datetime64_6(self):
         # these will correctly infer a datetime
-        s = Series([None, NaT, "2013-08-05 15:30:00.000001"])
-        assert s.dtype == "datetime64[ns]"
-        s = Series([np.nan, NaT, "2013-08-05 15:30:00.000001"])
-        assert s.dtype == "datetime64[ns]"
-        s = Series([NaT, None, "2013-08-05 15:30:00.000001"])
-        assert s.dtype == "datetime64[ns]"
-        s = Series([NaT, np.nan, "2013-08-05 15:30:00.000001"])
-        assert s.dtype == "datetime64[ns]"
+        msg = "containing strings is deprecated"
+
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            ser = Series([None, NaT, "2013-08-05 15:30:00.000001"])
+        assert ser.dtype == "datetime64[ns]"
+
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            ser = Series([np.nan, NaT, "2013-08-05 15:30:00.000001"])
+        assert ser.dtype == "datetime64[ns]"
+
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            ser = Series([NaT, None, "2013-08-05 15:30:00.000001"])
+        assert ser.dtype == "datetime64[ns]"
+
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            ser = Series([NaT, np.nan, "2013-08-05 15:30:00.000001"])
+        assert ser.dtype == "datetime64[ns]"
 
     def test_constructor_dtype_datetime64_5(self):
         # tz-aware (UTC and other tz's)
@@ -1379,14 +1388,22 @@ def test_constructor_dtype_timedelta64(self):
         assert td.dtype == "object"
 
         # these will correctly infer a timedelta
-        s = Series([None, NaT, "1 Day"])
-        assert s.dtype == "timedelta64[ns]"
-        s = Series([np.nan, NaT, "1 Day"])
-        assert s.dtype == "timedelta64[ns]"
-        s = Series([NaT, None, "1 Day"])
-        assert s.dtype == "timedelta64[ns]"
-        s = Series([NaT, np.nan, "1 Day"])
-        assert s.dtype == "timedelta64[ns]"
+        msg = "containing strings is deprecated"
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            ser = Series([None, NaT, "1 Day"])
+        assert ser.dtype == "timedelta64[ns]"
+
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            ser = Series([np.nan, NaT, "1 Day"])
+        assert ser.dtype == "timedelta64[ns]"
+
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            ser = Series([NaT, None, "1 Day"])
+        assert ser.dtype == "timedelta64[ns]"
+
+        with tm.assert_produces_warning(FutureWarning, match=msg):
+            ser = Series([NaT, np.nan, "1 Day"])
+        assert ser.dtype == "timedelta64[ns]"
 
     # GH 16406
     def test_constructor_mixed_tz(self):

pandas/tests/tools/test_to_timedelta.py

@@ -187,6 +187,16 @@ def test_to_timedelta_via_apply(self):
         result = Series([to_timedelta("00:00:01")])
         tm.assert_series_equal(result, expected)
 
+    def test_to_timedelta_inference_without_warning(self):
+        # GH#41731 inference produces a warning in the Series constructor,
+        #  but _not_ in to_timedelta
+        vals = ["00:00:01", pd.NaT]
+        with tm.assert_produces_warning(None):
+            result = to_timedelta(vals)
+
+        expected = TimedeltaIndex([pd.Timedelta(seconds=1), pd.NaT])
+        tm.assert_index_equal(result, expected)
+
     def test_to_timedelta_on_missing_values(self):
         # GH5438
         timedelta_NaT = np.timedelta64("NaT")
@@ -197,7 +207,8 @@ def test_to_timedelta_on_missing_values(self):
         )
         tm.assert_series_equal(actual, expected)
 
-        actual = to_timedelta(Series(["00:00:01", pd.NaT]))
+        with tm.assert_produces_warning(FutureWarning, match="Inferring timedelta64"):
+            actual = to_timedelta(Series(["00:00:01", pd.NaT]))
         tm.assert_series_equal(actual, expected)
 
         actual = to_timedelta(np.nan)