Fix explode to preserve datetime unit in Series and DataFrame; update related tests

Author: Narendersingh007

pandas/core/frame.py

@@ -10,7 +10,7 @@
 """
 
 from __future__ import annotations
-
+import pandas as pd
 import collections
 from collections import abc
 from collections.abc import (
@@ -9903,13 +9903,24 @@ def explode(
         df = self.reset_index(drop=True)
         if len(columns) == 1:
             result = df[columns[0]].explode()
+            orig_dtype = df[columns[0]].dtype
+            if pd.api.types.is_datetime64_dtype(orig_dtype):
+                result = result.astype(orig_dtype)
         else:
             mylen = lambda x: len(x) if (is_list_like(x) and len(x) > 0) else 1
             counts0 = self[columns[0]].apply(mylen)
             for c in columns[1:]:
                 if not all(counts0 == self[c].apply(mylen)):
                     raise ValueError("columns must have matching element counts")
-            result = DataFrame({c: df[c].explode() for c in columns})
+            result_data = {}
+            for c in columns:
+                exploded_series = df[c].explode()
+                orig_dtype = df[c].dtype
+                if pd.api.types.is_datetime64_dtype(orig_dtype):
+                    exploded_series = exploded_series.astype(orig_dtype)
+                result_data[c] = exploded_series
+            result = DataFrame(result_data)
+
         result = df.drop(columns, axis=1).join(result)
         if ignore_index:
             result.index = default_index(len(result))

pandas/tests/series/methods/test_explode.py

@@ -175,3 +175,29 @@ def test_explode_pyarrow_non_list_type(ignore_index):
     result = ser.explode(ignore_index=ignore_index)
     expected = pd.Series([1, 2, 3], dtype="int64[pyarrow]", index=[0, 1, 2])
     tm.assert_series_equal(result, expected)
+
+def test_explode_preserves_datetime_unit():
+    # Create datetime64[ms] array manually
+    dt64_ms = np.array(["2020-01-01T00:00:00.000", "2020-01-01T01:00:00.000", "2020-01-01T02:00:00.000"], dtype="datetime64[ms]")
+    s = pd.Series([dt64_ms])
+
+    # Explode the Series
+    result = s.explode()
+
+    # Ensure the dtype (including unit) is preserved
+    assert result.dtype == dt64_ms.dtype, f"Expected dtype {dt64_ms.dtype}, got {result.dtype}"
+
+def test_single_column_explode_preserves_datetime_unit():
+    # Use freq in ms since unit='ms'
+    rng = pd.date_range("2020-01-01T00:00:00Z", periods=3, freq="3600000ms", unit="ms")
+    s = pd.Series([rng])
+    result = s.explode()
+    assert result.dtype == rng.dtype
+
+def test_multi_column_explode_preserves_datetime_unit():
+    rng1 = pd.date_range("2020-01-01", periods=2, freq="3600000ms", unit="ms")
+    rng2 = pd.date_range("2020-01-01", periods=2, freq="3600000ms", unit="ms")
+    df = pd.DataFrame({"A": [rng1], "B": [rng2]})
+    result = df.explode(["A", "B"])
+    assert result["A"].dtype == rng1.dtype
+    assert result["B"].dtype == rng2.dtype