BUG: prevent overflowing diffs raising error in cut (pandas-dev#26045) (pandas-dev#26063)

Author: Batalex

doc/source/whatsnew/v0.25.0.rst

@@ -400,6 +400,7 @@ Reshaping
 - Bug in :func:`concat` where the resulting ``freq`` of two :class:`DatetimeIndex` with the same ``freq`` would be dropped (:issue:`3232`).
 - Bug in :func:`merge` where merging with equivalent Categorical dtypes was raising an error (:issue:`22501`)
 - Bug in :class:`DataFrame` constructor when passing non-empty tuples would cause a segmentation fault (:issue:`25691`)
+- Bug in :func:`pandas.cut` where large bins could incorrectly raise an error due to an integer overflow (:issue:`26045`)
 
 Sparse
 ^^^^^^

pandas/core/reshape/tile.py

@@ -230,7 +230,9 @@ def cut(x, bins, right=True, labels=None, retbins=False, precision=3,
         else:
             bins = np.asarray(bins)
         bins = _convert_bin_to_numeric_type(bins, dtype)
-        if (np.diff(bins) < 0).any():
+
+        # GH 26045: cast to float64 to avoid an overflow
+        if (np.diff(bins.astype('float64')) < 0).any():
             raise ValueError('bins must increase monotonically.')
 
     fac, bins = _bins_to_cuts(x, bins, right=right, labels=labels,

pandas/tests/reshape/test_cut.py

@@ -112,6 +112,35 @@ def test_bins_not_monotonic():
         cut(data, [0.1, 1.5, 1, 10])
 
 
+@pytest.mark.parametrize("x, bins, expected", [
+    (date_range("2017-12-31", periods=3),
+     [Timestamp.min, Timestamp('2018-01-01'), Timestamp.max],
+     IntervalIndex.from_tuples([
+         (Timestamp.min, Timestamp('2018-01-01')),
+         (Timestamp('2018-01-01'), Timestamp.max)])),
+
+    ([-1, 0, 1],
+     np.array([np.iinfo(np.int64).min, 0, np.iinfo(np.int64).max],
+              dtype="int64"),
+     IntervalIndex.from_tuples([
+         (np.iinfo(np.int64).min, 0),
+         (0, np.iinfo(np.int64).max)])),
+
+    ([np.timedelta64(-1), np.timedelta64(0), np.timedelta64(1)],
+     np.array([
+         np.timedelta64(-np.iinfo(np.int64).max),
+         np.timedelta64(0),
+         np.timedelta64(np.iinfo(np.int64).max)]),
+     IntervalIndex.from_tuples([
+         (np.timedelta64(-np.iinfo(np.int64).max), np.timedelta64(0)),
+         (np.timedelta64(0), np.timedelta64(np.iinfo(np.int64).max))])),
+])
+def test_bins_monotonic_not_overflowing(x, bins, expected):
+    # GH 26045
+    result = cut(x, bins)
+    tm.assert_index_equal(result.categories, expected)
+
+
 def test_wrong_num_labels():
     msg = "Bin labels must be one fewer than the number of bin edges"
     data = [.2, 1.4, 2.5, 6.2, 9.7, 2.1]