Fix overflow issue in decode_cf_datetime for dtypes <= np.uint32

doc/whats-new.rst

@@ -126,6 +126,9 @@ Bug fixes
 - :py:meth:`isel` with `drop=True` works as intended with scalar :py:class:`DataArray` indexers.
   (:issue:`6554`, :pull:`6579`)
   By `Michael Niklas <https://github.com/headtr1ck>`_.
+- Fixed silent overflow issue when decoding times encoded with 32-bit and below
+  unsigned integer data types (:issue:`6589`, :pull:`6598`). By `Spencer Clark
+  <https://github.com/spencerkclark>`_.
 
 Documentation
 ~~~~~~~~~~~~~

xarray/coding/times.py

@@ -218,16 +218,23 @@ def _decode_datetime_with_pandas(flat_num_dates, units, calendar):
         pd.to_timedelta(flat_num_dates.max(), delta) + ref_date
 
     # To avoid integer overflow when converting to nanosecond units for integer
-    # dtypes smaller than np.int64 cast all integer-dtype arrays to np.int64
-    # (GH 2002).
-    if flat_num_dates.dtype.kind == "i":
+    # dtypes smaller than np.int64 cast all integer and unsigned integer dtype
+    # arrays to np.int64 (GH 2002, GH 6589).  Note this is safe even in the case
+    # of np.uint64 values, because any np.uint64 value that would lead to
+    # overflow when converting to np.int64 would not be representable with a
+    # timedelta64 value, and therefore would raise an error in the lines above.
+    if flat_num_dates.dtype.kind in "iu":
         flat_num_dates = flat_num_dates.astype(np.int64)
 
     # Cast input ordinals to integers of nanoseconds because pd.to_timedelta
     # works much faster when dealing with integers (GH 1399).
     flat_num_dates_ns_int = (flat_num_dates * _NS_PER_TIME_DELTA[delta]).astype(
         np.int64
     )
+    print(_NS_PER_TIME_DELTA[delta])
+    print(flat_num_dates)
+    print(flat_num_dates * _NS_PER_TIME_DELTA[delta])
+    print(flat_num_dates_ns_int)
 
     # Use pd.to_timedelta to safely cast integer values to timedeltas,
     # and add those to a Timestamp to safely produce a DatetimeIndex.  This

xarray/tests/test_coding_times.py

@@ -1121,3 +1121,30 @@ def test_should_cftime_be_used_target_not_npable():
         ValueError, match="Calendar 'noleap' is only valid with cftime."
     ):
         _should_cftime_be_used(src, "noleap", False)
+
+
+@pytest.mark.parametrize("dtype", [np.uint8, np.uint16, np.uint32, np.uint64])
+def test_decode_cf_datetime_uint(dtype):
+    units = "seconds since 2018-08-22T03:23:03Z"
+    num_dates = dtype(50)
+    result = decode_cf_datetime(num_dates, units)
+    expected = np.asarray(np.datetime64("2018-08-22T03:23:53", "ns"))
+    np.testing.assert_equal(result, expected)
+
+
+@requires_cftime
+def test_decode_cf_datetime_uint64_with_cftime():
+    units = "days since 1700-01-01"
+    num_dates = np.uint64(182621)
+    result = decode_cf_datetime(num_dates, units)
+    expected = np.asarray(np.datetime64("2200-01-01", "ns"))
+    np.testing.assert_equal(result, expected)
+
+
+@requires_cftime
+def test_decode_cf_datetime_uint64_with_cftime_overflow_error():
+    units = "microseconds since 1700-01-01"
+    calendar = "360_day"
+    num_dates = np.uint64(1_000_000 * 86_400 * 360 * 500_000)
+    with pytest.raises(OverflowError):
+        decode_cf_datetime(num_dates, units, calendar)