Backport PR #41846 on branch 1.3.x (BUG: DataFrame.at with CategoricalIndex)

doc/source/whatsnew/v1.3.0.rst

@@ -1015,6 +1015,7 @@ Indexing
 - Bug in :meth:`DataFrame.loc.__getitem__` with :class:`MultiIndex` casting to float when at least one index column has float dtype and we retrieve a scalar (:issue:`41369`)
 - Bug in :meth:`DataFrame.loc` incorrectly matching non-Boolean index elements (:issue:`20432`)
 - Bug in :meth:`Series.__delitem__` with ``ExtensionDtype`` incorrectly casting to ``ndarray`` (:issue:`40386`)
+- Bug in :meth:`DataFrame.at` with a :class:`CategoricalIndex` returning incorrect results when passed integer keys (:issue:`41846`)
 - Bug in :meth:`DataFrame.loc` returning a :class:`MultiIndex` in the wrong order if an indexer has duplicates (:issue:`40978`)
 - Bug in :meth:`DataFrame.__setitem__` raising a ``TypeError`` when using a ``str`` subclass as the column name with a :class:`DatetimeIndex` (:issue:`37366`)
 - Bug in :meth:`PeriodIndex.get_loc` failing to raise a ``KeyError`` when given a :class:`Period` with a mismatched ``freq`` (:issue:`41670`)

pandas/core/frame.py

@@ -158,6 +158,7 @@
 from pandas.core.indexers import check_key_length
 from pandas.core.indexes import base as ibase
 from pandas.core.indexes.api import (
+    CategoricalIndex,
     DatetimeIndex,
     Index,
     PeriodIndex,
@@ -3553,6 +3554,11 @@ def _get_value(self, index, col, takeable: bool = False) -> Scalar:
         Returns
         -------
         scalar
+
+        Notes
+        -----
+        Assumes that index and columns both have ax._index_as_unique;
+        caller is responsible for checking.
         """
         if takeable:
             series = self._ixs(col, axis=1)
@@ -3561,20 +3567,21 @@ def _get_value(self, index, col, takeable: bool = False) -> Scalar:
         series = self._get_item_cache(col)
         engine = self.index._engine
 
+        if isinstance(self.index, CategoricalIndex):
+            # Trying to use the engine fastpath may give incorrect results
+            #  if our categories are integers that dont match our codes
+            col = self.columns.get_loc(col)
+            index = self.index.get_loc(index)
+            return self._get_value(index, col, takeable=True)
+
         try:
             loc = engine.get_loc(index)
             return series._values[loc]
-        except KeyError:
-            # GH 20629
-            if self.index.nlevels > 1:
-                # partial indexing forbidden
-                raise
-
-        # we cannot handle direct indexing
-        # use positional
-        col = self.columns.get_loc(col)
-        index = self.index.get_loc(index)
-        return self._get_value(index, col, takeable=True)
+        except AttributeError:
+            # IntervalTree has no get_loc
+            col = self.columns.get_loc(col)
+            index = self.index.get_loc(index)
+            return self._get_value(index, col, takeable=True)
 
     def __setitem__(self, key, value):
         key = com.apply_if_callable(key, self)

pandas/core/indexing.py

@@ -916,8 +916,7 @@ def __getitem__(self, key):
             key = tuple(list(x) if is_iterator(x) else x for x in key)
             key = tuple(com.apply_if_callable(x, self.obj) for x in key)
             if self._is_scalar_access(key):
-                with suppress(KeyError, IndexError, AttributeError):
-                    # AttributeError for IntervalTree get_value
+                with suppress(KeyError, IndexError):
                     return self.obj._get_value(*key, takeable=self._takeable)
             return self._getitem_tuple(key)
         else:
@@ -1004,7 +1003,7 @@ def _is_scalar_access(self, key: tuple) -> bool:
                 # should not be considered scalar
                 return False
 
-            if not ax.is_unique:
+            if not ax._index_as_unique:
                 return False
 
         return True

pandas/tests/indexing/test_at.py

@@ -8,6 +8,7 @@
 
 from pandas import (
     CategoricalDtype,
+    CategoricalIndex,
     DataFrame,
     Series,
     Timestamp,
@@ -141,3 +142,16 @@ def test_at_getitem_mixed_index_no_fallback(self):
             ser.at[0]
         with pytest.raises(KeyError, match="^4$"):
             ser.at[4]
+
+    def test_at_categorical_integers(self):
+        # CategoricalIndex with integer categories that don't happen to match
+        #  the Categorical's codes
+        ci = CategoricalIndex([3, 4])
+
+        arr = np.arange(4).reshape(2, 2)
+        frame = DataFrame(arr, index=ci)
+
+        for df in [frame, frame.T]:
+            for key in [0, 1]:
+                with pytest.raises(KeyError, match=str(key)):
+                    df.at[key, key]