API/BUG: freq retention in value_counts

doc/source/whatsnew/v3.0.0.rst

@@ -976,6 +976,7 @@ Datetimelike
 - Bug in comparison between objects with pyarrow date dtype and ``timestamp[pyarrow]`` or ``np.datetime64`` dtype failing to consider these as non-comparable (:issue:`62157`)
 - Bug in constructing arrays with :class:`ArrowDtype` with ``timestamp`` type incorrectly allowing ``Decimal("NaN")`` (:issue:`61773`)
 - Bug in constructing arrays with a timezone-aware :class:`ArrowDtype` from timezone-naive datetime objects incorrectly treating those as UTC times instead of wall times like :class:`DatetimeTZDtype` (:issue:`61775`)
+- Bug in retaining frequency in :meth:`value_counts` specifically for :meth:`DatetimeIndex` and :meth:`TimedeltaIndex` (:issue:`33830`)
 - Bug in setting scalar values with mismatched resolution into arrays with non-nanosecond ``datetime64``, ``timedelta64`` or :class:`DatetimeTZDtype` incorrectly truncating those scalars (:issue:`56410`)
 
 

pandas/core/algorithms.py

@@ -868,8 +868,10 @@ def value_counts_internal(
     dropna: bool = True,
 ) -> Series:
     from pandas import (
+        DatetimeIndex,
         Index,
         Series,
+        TimedeltaIndex,
     )
 
     index_name = getattr(values, "name", None)
@@ -934,6 +936,19 @@ def value_counts_internal(
             # Starting in 3.0, we no longer perform dtype inference on the
             #  Index object we construct here, xref GH#56161
             idx = Index(keys, dtype=keys.dtype, name=index_name)
+
+            if (
+                bins is None
+                and not sort
+                and isinstance(values, (DatetimeIndex, TimedeltaIndex))
+                and values.inferred_freq is not None
+                and len(idx) == len(values)
+                and idx.equals(values)
+            ):
+                # freq preservation
+                # Rebuild idx with the correct type and inferred frequency
+                idx.freq = values.inferred_freq
+
             result = Series(counts, index=idx, name=name, copy=False)
 
     if sort:

pandas/tests/base/test_value_counts.py

@@ -339,3 +339,81 @@ def test_value_counts_object_inference_deprecated():
     exp = dti.value_counts()
     exp.index = exp.index.astype(object)
     tm.assert_series_equal(res, exp)
+
+
+@pytest.mark.parametrize(
+    "index",
+    [
+        pd.date_range("2016-01-01", periods=5, freq="D"),
+        pd.timedelta_range(Timedelta(0), periods=5, freq="h"),
+    ],
+    ids=["DatetimeIndex[D]", "TimedeltaIndex[h]"],
+)
+@pytest.mark.parametrize(
+    "build,kwargs,exp_preserve,exp_hasnans,exp_index_fn",
+    [
+        (lambda idx: idx, {"sort": False}, True, False, lambda idx, obj: idx),
+        (
+            lambda idx: idx,
+            {"sort": False, "normalize": True},
+            True,
+            False,
+            lambda idx, obj: idx,
+        ),
+        (lambda idx: idx, {}, False, False, None),
+        (
+            lambda idx: idx.insert(1, idx[1]),
+            {"sort": False},
+            False,
+            False,
+            lambda idx, obj: type(idx)(idx, freq=None),
+        ),
+        (
+            lambda idx: idx.delete(2),
+            {"sort": False},
+            False,
+            False,
+            lambda idx, obj: type(idx)(obj, freq=None),
+        ),
+        (
+            lambda idx: idx.insert(1, pd.NaT),
+            {"sort": False, "dropna": False},
+            False,
+            True,
+            lambda idx, obj: type(idx)(
+                list(idx[:1]) + [pd.NaT] + list(idx[1:]), freq=None
+            ),
+        ),
+        (
+            lambda idx: idx.insert(1, pd.NaT),
+            {"sort": False, "dropna": True},
+            False,
+            False,
+            lambda idx, obj: type(idx)(idx, freq=None),
+        ),
+    ],
+)
+def test_value_counts_freq_datetimelike(
+    index, build, kwargs, exp_preserve, exp_hasnans, exp_index_fn
+):
+    obj = build(index)
+    vc = obj.value_counts(**kwargs)
+
+    # without sort
+    if exp_index_fn is not None:
+        expected_idx = exp_index_fn(index, obj)
+        tm.assert_index_equal(vc.index, expected_idx)
+
+    # freq preservation / drop
+    if exp_preserve:
+        assert vc.index.freq == index.freq
+    else:
+        assert vc.index.freq is None
+
+    # NaT presence
+    assert vc.index.hasnans is exp_hasnans
+
+    # without normalize
+    if kwargs.get("normalize", False):
+        expected_val = 1.0 / len(index)
+        assert np.isclose(vc.to_numpy(), expected_val).all()