BUG: IntervalIndex is_monotonic, get_loc, get_indexer_for, contains with np.nan (pandas-dev#41863)

Author: jbrockmendel

doc/source/whatsnew/v1.3.0.rst

@@ -950,6 +950,7 @@ Interval
 - Bug in :meth:`IntervalIndex.intersection` returning duplicates when at least one of the :class:`Index` objects have duplicates which are present in the other (:issue:`38743`)
 - :meth:`IntervalIndex.union`, :meth:`IntervalIndex.intersection`, :meth:`IntervalIndex.difference`, and :meth:`IntervalIndex.symmetric_difference` now cast to the appropriate dtype instead of raising a ``TypeError`` when operating with another :class:`IntervalIndex` with incompatible dtype (:issue:`39267`)
 - :meth:`PeriodIndex.union`, :meth:`PeriodIndex.intersection`, :meth:`PeriodIndex.symmetric_difference`, :meth:`PeriodIndex.difference` now cast to object dtype instead of raising ``IncompatibleFrequency`` when operating with another :class:`PeriodIndex` with incompatible dtype (:issue:`39306`)
+- Bug in :meth:`IntervalIndex.is_monotonic`, :meth:`IntervalIndex.get_loc`, :meth:`IntervalIndex.get_indexer_for`, and :meth:`IntervalIndex.__contains__` when NA values are present (:issue:`41831`)
 
 Indexing
 ^^^^^^^^

pandas/_libs/intervaltree.pxi.in

@@ -36,6 +36,7 @@ cdef class IntervalTree(IntervalMixin):
         object dtype
         str closed
         object _is_overlapping, _left_sorter, _right_sorter
+        Py_ssize_t _na_count
 
     def __init__(self, left, right, closed='right', leaf_size=100):
         """
@@ -67,6 +68,7 @@ cdef class IntervalTree(IntervalMixin):
 
         # GH 23352: ensure no nan in nodes
         mask = ~np.isnan(self.left)
+        self._na_count = len(mask) - mask.sum()
         self.left = self.left[mask]
         self.right = self.right[mask]
         indices = indices[mask]
@@ -116,6 +118,8 @@ cdef class IntervalTree(IntervalMixin):
         Return True if the IntervalTree is monotonic increasing (only equal or
         increasing values), else False
         """
+        if self._na_count > 0:
+            return False
         values = [self.right, self.left]
 
         sort_order = np.lexsort(values)

pandas/core/indexes/interval.py

@@ -61,6 +61,7 @@
     is_scalar,
 )
 from pandas.core.dtypes.dtypes import IntervalDtype
+from pandas.core.dtypes.missing import is_valid_na_for_dtype
 
 from pandas.core.algorithms import take_nd
 from pandas.core.arrays.interval import (
@@ -343,6 +344,8 @@ def __contains__(self, key: Any) -> bool:
         """
         hash(key)
         if not isinstance(key, Interval):
+            if is_valid_na_for_dtype(key, self.dtype):
+                return self.hasnans
             return False
 
         try:
@@ -618,6 +621,8 @@ def get_loc(
             if self.closed != key.closed:
                 raise KeyError(key)
             mask = (self.left == key.left) & (self.right == key.right)
+        elif is_valid_na_for_dtype(key, self.dtype):
+            mask = self.isna()
         else:
             # assume scalar
             op_left = le if self.closed_left else lt
@@ -633,7 +638,12 @@ def get_loc(
             raise KeyError(key)
         elif matches == 1:
             return mask.argmax()
-        return lib.maybe_booleans_to_slice(mask.view("u1"))
+
+        res = lib.maybe_booleans_to_slice(mask.view("u1"))
+        if isinstance(res, slice) and res.stop is None:
+            # TODO: DO this in maybe_booleans_to_slice?
+            res = slice(res.start, len(self), res.step)
+        return res
 
     def _get_indexer(
         self,
@@ -721,9 +731,9 @@ def _get_indexer_pointwise(self, target: Index) -> tuple[np.ndarray, np.ndarray]
         indexer = np.concatenate(indexer)
         return ensure_platform_int(indexer), ensure_platform_int(missing)
 
-    @property
+    @cache_readonly
     def _index_as_unique(self) -> bool:
-        return not self.is_overlapping
+        return not self.is_overlapping and self._engine._na_count < 2
 
     _requires_unique_msg = (
         "cannot handle overlapping indices; use IntervalIndex.get_indexer_non_unique"

pandas/tests/indexes/interval/test_indexing.py

@@ -6,9 +6,11 @@
 from pandas.errors import InvalidIndexError
 
 from pandas import (
+    NA,
     CategoricalIndex,
     Interval,
     IntervalIndex,
+    NaT,
     Timedelta,
     date_range,
     timedelta_range,
@@ -168,6 +170,20 @@ def test_get_loc_non_scalar_errors(self, key):
         with pytest.raises(InvalidIndexError, match=msg):
             idx.get_loc(key)
 
+    def test_get_indexer_with_nans(self):
+        # GH#41831
+        index = IntervalIndex([np.nan, Interval(1, 2), np.nan])
+
+        expected = np.array([True, False, True])
+        for key in [None, np.nan, NA]:
+            assert key in index
+            result = index.get_loc(key)
+            tm.assert_numpy_array_equal(result, expected)
+
+        for key in [NaT, np.timedelta64("NaT", "ns"), np.datetime64("NaT", "ns")]:
+            with pytest.raises(KeyError, match=str(key)):
+                index.get_loc(key)
+
 
 class TestGetIndexer:
     @pytest.mark.parametrize(
@@ -326,6 +342,17 @@ def test_get_indexer_non_monotonic(self):
         expected = np.array([1, 2], dtype=np.intp)
         tm.assert_numpy_array_equal(result, expected)
 
+    def test_get_indexer_with_nans(self):
+        # GH#41831
+        index = IntervalIndex([np.nan, np.nan])
+        other = IntervalIndex([np.nan])
+
+        assert not index._index_as_unique
+
+        result = index.get_indexer_for(other)
+        expected = np.array([0, 1], dtype=np.intp)
+        tm.assert_numpy_array_equal(result, expected)
+
 
 class TestSliceLocs:
     def test_slice_locs_with_interval(self):

pandas/tests/indexes/interval/test_interval.py

@@ -328,6 +328,16 @@ def test_monotonic(self, closed):
         assert idx.is_monotonic_decreasing is True
         assert idx._is_strictly_monotonic_decreasing is True
 
+    def test_is_monotonic_with_nans(self):
+        # GH#41831
+        index = IntervalIndex([np.nan, np.nan])
+
+        assert not index.is_monotonic
+        assert not index._is_strictly_monotonic_increasing
+        assert not index.is_monotonic_increasing
+        assert not index._is_strictly_monotonic_decreasing
+        assert not index.is_monotonic_decreasing
+
     def test_get_item(self, closed):
         i = IntervalIndex.from_arrays((0, 1, np.nan), (1, 2, np.nan), closed=closed)
         assert i[0] == Interval(0.0, 1.0, closed=closed)

pandas/tests/indexing/interval/test_interval.py

@@ -114,6 +114,21 @@ def test_loc_getitem_frame(self):
         with pytest.raises(KeyError, match=r"\[10\] not in index"):
             df.loc[[10, 4]]
 
+    def test_getitem_interval_with_nans(self, frame_or_series, indexer_sl):
+        # GH#41831
+
+        index = IntervalIndex([np.nan, np.nan])
+        key = index[:-1]
+
+        obj = frame_or_series(range(2), index=index)
+        if frame_or_series is DataFrame and indexer_sl is tm.setitem:
+            obj = obj.T
+
+        result = indexer_sl(obj)[key]
+        expected = obj
+
+        tm.assert_equal(result, expected)
+
 
 class TestIntervalIndexInsideMultiIndex:
     def test_mi_intervalindex_slicing_with_scalar(self):