DEPR: move NumericIndexes into base Index, part 1

Author: Terji Petersen

pandas/__init__.py

@@ -183,38 +183,6 @@
 __git_version__ = v.get("full-revisionid")
 del get_versions, v
 
-# GH 27101
-__deprecated_num_index_names = ["Float64Index", "Int64Index", "UInt64Index"]
-
-
-def __dir__() -> list[str]:
-    # GH43028
-    # Int64Index etc. are deprecated, but we still want them to be available in the dir.
-    # Remove in Pandas 2.0, when we remove Int64Index etc. from the code base.
-    return list(globals().keys()) + __deprecated_num_index_names
-
-
-def __getattr__(name):
-    import warnings
-
-    if name in __deprecated_num_index_names:
-        warnings.warn(
-            f"pandas.{name} is deprecated "
-            "and will be removed from pandas in a future version. "
-            "Use pandas.Index with the appropriate dtype instead.",
-            FutureWarning,
-            stacklevel=2,
-        )
-        from pandas.core.api import Float64Index, Int64Index, UInt64Index
-
-        return {
-            "Float64Index": Float64Index,
-            "Int64Index": Int64Index,
-            "UInt64Index": UInt64Index,
-        }[name]
-
-    raise AttributeError(f"module 'pandas' has no attribute '{name}'")
-
 
 # module level doc-string
 __doc__ = """

pandas/core/frame.py

@@ -11203,7 +11203,7 @@ def quantile(
             )
         if method == "single":
             # error: Argument "qs" to "quantile" of "BlockManager" has incompatible type
-            # "Index"; expected "Float64Index"
+            # "Index"; expected "NumericIndex"
             res = data._mgr.quantile(
                 qs=q, axis=1, interpolation=interpolation  # type: ignore[arg-type]
             )

pandas/core/indexes/base.py

@@ -91,6 +91,7 @@
     ensure_platform_int,
     is_bool_dtype,
     is_categorical_dtype,
+    is_complex_dtype,
     is_dtype_equal,
     is_ea_or_datetimelike_dtype,
     is_extension_array_dtype,
@@ -590,18 +591,14 @@ def _dtype_to_subclass(cls, dtype: DtypeObj):
 
             return TimedeltaIndex
 
-        elif dtype.kind == "f":
-            from pandas.core.api import Float64Index
-
-            return Float64Index
-        elif dtype.kind == "u":
-            from pandas.core.api import UInt64Index
-
-            return UInt64Index
-        elif dtype.kind == "i":
-            from pandas.core.api import Int64Index
+        elif (
+            is_numeric_dtype(dtype)
+            and not is_bool_dtype(dtype)
+            and not is_complex_dtype(dtype)
+        ):
+            from pandas.core.api import NumericIndex
 
-            return Int64Index
+            return NumericIndex
 
         elif dtype.kind == "O":
             # NB: assuming away MultiIndex
@@ -1057,14 +1054,29 @@ def astype(self, dtype, copy: bool = True):
                 ) from err
 
         # pass copy=False because any copying will be done in the astype above
-        if self._is_backward_compat_public_numeric_index:
-            # this block is needed so e.g. NumericIndex[int8].astype("int32") returns
-            # NumericIndex[int32] and not Int64Index with dtype int64.
+        if not self._is_backward_compat_public_numeric_index and not isinstance(
+            self, ABCRangeIndex
+        ):
+            # this block is needed so e.g. Int64Index.astype("int32") returns
+            # Int64Index and not a NumericIndex with dtype int32.
             # When Int64Index etc. are removed from the code base, removed this also.
             if isinstance(dtype, np.dtype) and is_numeric_dtype(dtype):
-                return self._constructor(
-                    new_values, name=self.name, dtype=dtype, copy=False
+                from pandas.core.api import (
+                    Float64Index,
+                    Int64Index,
+                    UInt64Index,
                 )
+
+                if is_signed_integer_dtype(dtype):
+                    klass = Int64Index
+                elif is_unsigned_integer_dtype(dtype):
+                    klass = UInt64Index
+                elif is_float_dtype(dtype):
+                    klass = Float64Index
+                else:
+                    klass = Index
+                return klass(new_values, name=self.name, dtype=dtype, copy=False)
+
         return Index(new_values, name=self.name, dtype=new_values.dtype, copy=False)
 
     _index_shared_docs[
@@ -5265,6 +5277,7 @@ def putmask(self, mask, value) -> Index:
         if self.dtype != object and is_valid_na_for_dtype(value, self.dtype):
             # e.g. None -> np.nan, see also Block._standardize_fill_value
             value = self._na_value
+
         try:
             converted = self._validate_fill_value(value)
         except (LossySetitemError, ValueError, TypeError) as err:
@@ -6133,13 +6146,6 @@ def map(self, mapper, na_action=None):
                 new_values, self.dtype, same_dtype=same_dtype
             )
 
-        if self._is_backward_compat_public_numeric_index and is_numeric_dtype(
-            new_values.dtype
-        ):
-            return self._constructor(
-                new_values, dtype=dtype, copy=False, name=self.name
-            )
-
         return Index._with_infer(new_values, dtype=dtype, copy=False, name=self.name)
 
     # TODO: De-duplicate with map, xref GH#32349
@@ -6616,10 +6622,17 @@ def insert(self, loc: int, item) -> Index:
             loc = loc if loc >= 0 else loc - 1
             new_values[loc] = item
 
-        if self._typ == "numericindex":
-            # Use self._constructor instead of Index to retain NumericIndex GH#43921
-            # TODO(2.0) can use Index instead of self._constructor
-            return self._constructor._with_infer(new_values, name=self.name)
+        if not self._is_backward_compat_public_numeric_index:
+            from pandas.core.indexes.numeric import NumericIndex
+
+            if not isinstance(self, ABCRangeIndex) or not isinstance(
+                self, NumericIndex
+            ):
+                return Index._with_infer(new_values, name=self.name)
+            else:
+                # Use self._constructor instead of Index to retain old-style num. index
+                # TODO(2.0) can use Index instead of self._constructor
+                return self._constructor._with_infer(new_values, name=self.name)
         else:
             return Index._with_infer(new_values, name=self.name)
 

pandas/core/indexes/numeric.py

@@ -13,6 +13,7 @@
 )
 from pandas._typing import (
     Dtype,
+    DtypeObj,
     npt,
 )
 from pandas.util._decorators import (
@@ -347,7 +348,26 @@ def _format_native_types(
 """
 
 
-class IntegerIndex(NumericIndex):
+class TempBaseIndex(NumericIndex):
+    @classmethod
+    def _dtype_to_subclass(cls, dtype: DtypeObj):
+        if is_integer_dtype(dtype):
+            from pandas.core.api import Int64Index
+
+            return Int64Index
+        elif is_unsigned_integer_dtype(dtype):
+            from pandas.core.api import UInt64Index
+
+            return UInt64Index
+        elif is_float_dtype(dtype):
+            from pandas.core.api import Float64Index
+
+            return Float64Index
+        else:
+            return super()._dtype_to_subclass(dtype)
+
+
+class IntegerIndex(TempBaseIndex):
     """
     This is an abstract class for Int64Index, UInt64Index.
     """
@@ -391,7 +411,7 @@ def _engine_type(self) -> type[libindex.UInt64Engine]:
         return libindex.UInt64Engine
 
 
-class Float64Index(NumericIndex):
+class Float64Index(TempBaseIndex):
     _index_descr_args = {
         "klass": "Float64Index",
         "dtype": "float64",

pandas/core/indexes/range.py

@@ -185,9 +185,9 @@ def _simple_new(cls, values: range, name: Hashable = None) -> RangeIndex:
     # error: Return type "Type[Int64Index]" of "_constructor" incompatible with return
     # type "Type[RangeIndex]" in supertype "Index"
     @cache_readonly
-    def _constructor(self) -> type[Int64Index]:  # type: ignore[override]
+    def _constructor(self) -> type[NumericIndex]:  # type: ignore[override]
         """return the class to use for construction"""
-        return Int64Index
+        return NumericIndex
 
     # error: Signature of "_data" incompatible with supertype "Index"
     @cache_readonly

pandas/core/internals/managers.py

@@ -64,8 +64,8 @@
 )
 from pandas.core.indexers import maybe_convert_indices
 from pandas.core.indexes.api import (
-    Float64Index,
     Index,
+    NumericIndex,
     ensure_index,
 )
 from pandas.core.internals.base import (
@@ -1582,7 +1582,7 @@ def _equal_values(self: BlockManager, other: BlockManager) -> bool:
     def quantile(
         self: T,
         *,
-        qs: Float64Index,
+        qs: NumericIndex,
         axis: AxisInt = 0,
         interpolation: QuantileInterpolation = "linear",
     ) -> T:
@@ -1610,7 +1610,7 @@ def quantile(
         assert axis == 1  # only ever called this way
 
         new_axes = list(self.axes)
-        new_axes[1] = Float64Index(qs)
+        new_axes[1] = Index(qs, dtype=np.float64)
 
         blocks = [
             blk.quantile(axis=axis, qs=qs, interpolation=interpolation)

pandas/tests/api/test_api.py

@@ -55,17 +55,14 @@ class TestPDApi(Base):
         "DatetimeIndex",
         "ExcelFile",
         "ExcelWriter",
-        "Float64Index",
         "Flags",
         "Grouper",
         "HDFStore",
         "Index",
-        "Int64Index",
         "MultiIndex",
         "Period",
         "PeriodIndex",
         "RangeIndex",
-        "UInt64Index",
         "Series",
         "SparseDtype",
         "StringDtype",
@@ -93,7 +90,7 @@ class TestPDApi(Base):
     ]
 
     # these are already deprecated; awaiting removal
-    deprecated_classes: list[str] = ["Float64Index", "Int64Index", "UInt64Index"]
+    deprecated_classes: list[str] = []
 
     # external modules exposed in pandas namespace
     modules: list[str] = []

pandas/tests/apply/test_series_apply.py

@@ -186,7 +186,7 @@ def test_apply_datetimetz():
     # change dtype
     # GH 14506 : Returned dtype changed from int32 to int64
     result = s.apply(lambda x: x.hour)
-    exp = Series(list(range(24)) + [0], name="XX", dtype=np.int64)
+    exp = Series(list(range(24)) + [0], name="XX", dtype=np.int32)
     tm.assert_series_equal(result, exp)
 
     # not vectorized
@@ -766,7 +766,7 @@ def test_map_datetimetz():
     # change dtype
     # GH 14506 : Returned dtype changed from int32 to int64
     result = s.map(lambda x: x.hour)
-    exp = Series(list(range(24)) + [0], name="XX", dtype=np.int64)
+    exp = Series(list(range(24)) + [0], name="XX", dtype=np.int32)
     tm.assert_series_equal(result, exp)
 
     # not vectorized

pandas/tests/base/test_unique.py

@@ -5,12 +5,19 @@
 
 import pandas as pd
 import pandas._testing as tm
-from pandas.core.api import NumericIndex
+from pandas.core.api import (
+    Float64Index,
+    Int64Index,
+    NumericIndex,
+    UInt64Index,
+)
 from pandas.tests.base.common import allow_na_ops
 
 
 def test_unique(index_or_series_obj):
     obj = index_or_series_obj
+    if isinstance(obj, (Int64Index, UInt64Index, Float64Index)):
+        pytest.skip("these indexes will be removed in followup")
     obj = np.repeat(obj, range(1, len(obj) + 1))
     result = obj.unique()
 
@@ -36,6 +43,8 @@ def test_unique(index_or_series_obj):
 @pytest.mark.parametrize("null_obj", [np.nan, None])
 def test_unique_null(null_obj, index_or_series_obj):
     obj = index_or_series_obj
+    if isinstance(obj, (Int64Index, UInt64Index, Float64Index)):
+        pytest.skip("these indexes will be removed in followup")
 
     if not allow_na_ops(obj):
         pytest.skip("type doesn't allow for NA operations")

pandas/tests/base/test_value_counts.py

@@ -24,6 +24,10 @@
 def test_value_counts(index_or_series_obj):
     obj = index_or_series_obj
     obj = np.repeat(obj, range(1, len(obj) + 1))
+
+    if obj.dtype == np.float16:
+        pytest.xfail("we don't have Float16Index, so value_counts is problematic")
+
     result = obj.value_counts()
 
     counter = collections.Counter(obj)
@@ -57,6 +61,9 @@ def test_value_counts_null(null_obj, index_or_series_obj):
     orig = index_or_series_obj
     obj = orig.copy()
 
+    if obj.dtype == np.float16:
+        pytest.xfail("we don't have Float16Index, so value_counts is problematic")
+
     if not allow_na_ops(obj):
         pytest.skip("type doesn't allow for NA operations")
     elif len(obj) < 1:

pandas/tests/frame/indexing/test_indexing.py

@@ -736,7 +736,7 @@ def test_getitem_setitem_float_labels(self, using_array_manager):
 
         # positional slicing only via iloc!
         msg = (
-            "cannot do positional indexing on Float64Index with "
+            "cannot do positional indexing on NumericIndex with "
             r"these indexers \[1.0\] of type float"
         )
         with pytest.raises(TypeError, match=msg):

pandas/tests/frame/methods/test_set_index.py

@@ -159,7 +159,7 @@ def test_set_index_cast(self):
         df = DataFrame(
             {"A": [1.1, 2.2, 3.3], "B": [5.0, 6.1, 7.2]}, index=[2010, 2011, 2012]
         )
-        df2 = df.set_index(df.index.astype(np.int32))
+        df2 = df.set_index(df.index.astype(int))
         tm.assert_frame_equal(df, df2)
 
     # A has duplicate values, C does not

pandas/tests/groupby/test_groupby.py

@@ -93,7 +93,7 @@ def test_groupby_nonobject_dtype(mframe, df_mixed_floats):
     result = grouped.sum()
 
     expected = mframe.groupby(key.astype("O")).sum()
-    tm.assert_frame_equal(result, expected)
+    tm.assert_frame_equal(result, expected, check_index_type=False)
 
     # GH 3911, mixed frame non-conversion
     df = df_mixed_floats.copy()
@@ -228,6 +228,7 @@ def test_pass_args_kwargs_duplicate_columns(tsframe, as_index):
         2: tsframe[tsframe.index.month == 2].quantile(0.8),
     }
     expected = DataFrame(ex_data).T
+    expected.index = Index(range(1, 3), dtype=np.int32)
     if not as_index:
         # TODO: try to get this more consistent?
         expected.index = Index(range(2))
@@ -2814,7 +2815,7 @@ def test_groupby_overflow(val, dtype):
     result = df.groupby("a").sum()
     expected = DataFrame(
         {"b": [val * 2]},
-        index=Index([1], name="a", dtype=f"{dtype}64"),
+        index=Index([1], name="a", dtype=f"{dtype}8"),
         dtype=f"{dtype}64",
     )
     tm.assert_frame_equal(result, expected)
@@ -2826,7 +2827,7 @@ def test_groupby_overflow(val, dtype):
     result = df.groupby("a").prod()
     expected = DataFrame(
         {"b": [val * val]},
-        index=Index([1], name="a", dtype=f"{dtype}64"),
+        index=Index([1], name="a", dtype=f"{dtype}8"),
         dtype=f"{dtype}64",
     )
     tm.assert_frame_equal(result, expected)