BUG: Implement fillna(..., limit=x) for EAs

doc/source/whatsnew/v3.0.0.rst

@@ -399,7 +399,7 @@ Indexing
 
 Missing
 ^^^^^^^
--
+- Fixed bug in :meth:`DataFrame.fillna` and :meth:`Series.fillna` that would ignore the ``limit`` argument on :class:`.ExtensionArray` dtypes (:issue:`58001`)
 -
 
 MultiIndex
@@ -444,7 +444,6 @@ Sparse
 ExtensionArray
 ^^^^^^^^^^^^^^
 - Fixed bug in :meth:`api.types.is_datetime64_any_dtype` where a custom :class:`ExtensionDtype` would return ``False`` for array-likes (:issue:`57055`)
--
 
 Styler
 ^^^^^^

pandas/core/arrays/_mixins.py

@@ -330,6 +330,12 @@ def _pad_or_backfill(
     @doc(ExtensionArray.fillna)
     def fillna(self, value=None, limit: int | None = None, copy: bool = True) -> Self:
         mask = self.isna()
+        if limit is not None and limit < len(self):
+            modify = mask.cumsum() > limit
+            if modify.any():
+                # Only copy mask if necessary
+                mask = mask.copy()
+                mask[modify] = False
         # error: Argument 2 to "check_value_size" has incompatible type
         # "ExtensionArray"; expected "ndarray"
         value = missing.check_value_size(

pandas/core/arrays/base.py

@@ -1045,19 +1045,12 @@ def fillna(
             Alternatively, an array-like "value" can be given. It's expected
             that the array-like have the same length as 'self'.
         limit : int, default None
-            If method is specified, this is the maximum number of consecutive
-            NaN values to forward/backward fill. In other words, if there is
-            a gap with more than this number of consecutive NaNs, it will only
-            be partially filled. If method is not specified, this is the
-            maximum number of entries along the entire axis where NaNs will be
-            filled.
+            The maximum number of entries where NA values will be filled.
         copy : bool, default True
             Whether to make a copy of the data before filling. If False, then
             the original should be modified and no new memory should be allocated.
             For ExtensionArray subclasses that cannot do this, it is at the
             author's discretion whether to ignore "copy=False" or to raise.
-            The base class implementation ignores the keyword in pad/backfill
-            cases.
 
         Returns
         -------
@@ -1073,6 +1066,12 @@ def fillna(
         Length: 6, dtype: Int64
         """
         mask = self.isna()
+        if limit is not None and limit < len(self):
+            modify = mask.cumsum() > limit
+            if modify.any():
+                # Only copy mask if necessary
+                mask = mask.copy()
+                mask[modify] = False
         # error: Argument 2 to "check_value_size" has incompatible type
         # "ExtensionArray"; expected "ndarray"
         value = missing.check_value_size(

pandas/core/arrays/interval.py

@@ -905,12 +905,7 @@ def fillna(self, value=None, limit: int | None = None, copy: bool = True) -> Sel
             value(s) passed should be either Interval objects or NA/NaN.
         limit : int, default None
             (Not implemented yet for IntervalArray)
-            If method is specified, this is the maximum number of consecutive
-            NaN values to forward/backward fill. In other words, if there is
-            a gap with more than this number of consecutive NaNs, it will only
-            be partially filled. If method is not specified, this is the
-            maximum number of entries along the entire axis where NaNs will be
-            filled.
+            The maximum number of entries where NA values will be filled.
         copy : bool, default True
             Whether to make a copy of the data before filling. If False, then
             the original should be modified and no new memory should be allocated.
@@ -923,6 +918,8 @@ def fillna(self, value=None, limit: int | None = None, copy: bool = True) -> Sel
         """
         if copy is False:
             raise NotImplementedError
+        if limit is not None:
+            raise ValueError("limit must be None")
 
         value_left, value_right = self._validate_scalar(value)
 

pandas/core/arrays/masked.py

@@ -238,6 +238,12 @@ def _pad_or_backfill(
     @doc(ExtensionArray.fillna)
     def fillna(self, value=None, limit: int | None = None, copy: bool = True) -> Self:
         mask = self._mask
+        if limit is not None and limit < len(self):
+            modify = mask.cumsum() > limit
+            if modify.any():
+                # Only copy mask if necessary
+                mask = mask.copy()
+                mask[modify] = False
 
         value = missing.check_value_size(value, mask, len(self))
 

pandas/core/arrays/sparse/array.py

@@ -717,6 +717,7 @@ def fillna(
         ----------
         value : scalar
         limit : int, optional
+            Not supported for SparseArray, must be None.
         copy: bool, default True
             Ignored for SparseArray.
 
@@ -738,6 +739,8 @@ def fillna(
         """
         if value is None:
             raise ValueError("Must specify 'value'.")
+        if limit is not None:
+            raise ValueError("limit must be None")
         new_values = np.where(isna(self.sp_values), value, self.sp_values)
 
         if self._null_fill_value:

pandas/core/internals/blocks.py

@@ -1853,6 +1853,8 @@ def fillna(
     ) -> list[Block]:
         if isinstance(self.dtype, IntervalDtype):
             # Block.fillna handles coercion (test_fillna_interval)
+            if limit is not None:
+                raise ValueError("limit must be None")
             return super().fillna(
                 value=value,
                 limit=limit,

pandas/tests/extension/base/methods.py

@@ -299,6 +299,20 @@ def test_factorize_empty(self, data):
         tm.assert_numpy_array_equal(codes, expected_codes)
         tm.assert_extension_array_equal(uniques, expected_uniques)
 
+    def test_fillna_limit_frame(self, data_missing):
+        # GH#58001
+        df = pd.DataFrame({"A": data_missing.take([0, 1, 0, 1])})
+        expected = pd.DataFrame({"A": data_missing.take([1, 1, 0, 1])})
+        result = df.fillna(value=data_missing[1], limit=1)
+        tm.assert_frame_equal(result, expected)
+
+    def test_fillna_limit_series(self, data_missing):
+        # GH#58001
+        ser = pd.Series(data_missing.take([0, 1, 0, 1]))
+        expected = pd.Series(data_missing.take([1, 1, 0, 1]))
+        result = ser.fillna(value=data_missing[1], limit=1)
+        tm.assert_series_equal(result, expected)
+
     def test_fillna_copy_frame(self, data_missing):
         arr = data_missing.take([1, 1])
         df = pd.DataFrame({"A": arr})

pandas/tests/extension/decimal/test_decimal.py

@@ -144,6 +144,22 @@ def test_fillna_series(self, data_missing):
         ):
             super().test_fillna_series(data_missing)
 
+    def test_fillna_limit_frame(self, data_missing):
+        # GH#58001
+        msg = "ExtensionArray.fillna added a 'copy' keyword"
+        with tm.assert_produces_warning(
+            DeprecationWarning, match=msg, check_stacklevel=False
+        ):
+            super().test_fillna_limit_frame(data_missing)
+
+    def test_fillna_limit_series(self, data_missing):
+        # GH#58001
+        msg = "ExtensionArray.fillna added a 'copy' keyword"
+        with tm.assert_produces_warning(
+            DeprecationWarning, match=msg, check_stacklevel=False
+        ):
+            super().test_fillna_limit_series(data_missing)
+
     @pytest.mark.parametrize("dropna", [True, False])
     def test_value_counts(self, all_data, dropna):
         all_data = all_data[:10]

pandas/tests/extension/json/test_json.py

@@ -149,6 +149,16 @@ def test_fillna_frame(self):
         """We treat dictionaries as a mapping in fillna, not a scalar."""
         super().test_fillna_frame()
 
+    @pytest.mark.xfail(reason="fill value is a dictionary, takes incorrect code path")
+    def test_fillna_limit_frame(self, data_missing):
+        # GH#58001
+        super().test_fillna_limit_frame(data_missing)
+
+    @pytest.mark.xfail(reason="fill value is a dictionary, takes incorrect code path")
+    def test_fillna_limit_series(self, data_missing):
+        # GH#58001
+        super().test_fillna_limit_frame(data_missing)
+
     @pytest.mark.parametrize(
         "limit_area, input_ilocs, expected_ilocs",
         [

pandas/tests/extension/test_interval.py

@@ -84,6 +84,16 @@ class TestIntervalArray(base.ExtensionTests):
     def _supports_reduction(self, ser: pd.Series, op_name: str) -> bool:
         return op_name in ["min", "max"]
 
+    def test_fillna_limit_frame(self, data_missing):
+        # GH#58001
+        with pytest.raises(ValueError, match="limit must be None"):
+            super().test_fillna_limit_frame(data_missing)
+
+    def test_fillna_limit_series(self, data_missing):
+        # GH#58001
+        with pytest.raises(ValueError, match="limit must be None"):
+            super().test_fillna_limit_frame(data_missing)
+
     @pytest.mark.xfail(
         reason="Raises with incorrect message bc it disallows *all* listlikes "
         "instead of just wrong-length listlikes"

pandas/tests/extension/test_numpy.py

@@ -205,6 +205,18 @@ def test_shift_fill_value(self, data):
         # np.array shape inference. Shift implementation fails.
         super().test_shift_fill_value(data)
 
+    @skip_nested
+    def test_fillna_limit_frame(self, data_missing):
+        # GH#58001
+        # The "scalar" for this array isn't a scalar.
+        super().test_fillna_limit_frame(data_missing)
+
+    @skip_nested
+    def test_fillna_limit_series(self, data_missing):
+        # GH#58001
+        # The "scalar" for this array isn't a scalar.
+        super().test_fillna_limit_series(data_missing)
+
     @skip_nested
     def test_fillna_copy_frame(self, data_missing):
         # The "scalar" for this array isn't a scalar.

pandas/tests/extension/test_sparse.py

@@ -264,6 +264,16 @@ def test_fillna_frame(self, data_missing):
 
         tm.assert_frame_equal(result, expected)
 
+    def test_fillna_limit_frame(self, data_missing):
+        # GH#58001
+        with pytest.raises(ValueError, match="limit must be None"):
+            super().test_fillna_limit_frame(data_missing)
+
+    def test_fillna_limit_series(self, data_missing):
+        # GH#58001
+        with pytest.raises(ValueError, match="limit must be None"):
+            super().test_fillna_limit_frame(data_missing)
+
     _combine_le_expected_dtype = "Sparse[bool]"
 
     def test_fillna_copy_frame(self, data_missing):