DEPR: Remove NumericIndex.__new__ + related methods (#51050)

* Remove NumericIndex.__new__ (almost) + various methods

* adds GH-number

* fix errors

* fix error II

Author: topper-123

doc/source/whatsnew/v2.0.0.rst

@@ -619,6 +619,7 @@ Other API changes
   new DataFrame (shallow copy) instead of the original DataFrame, consistent with other
   methods to get a full slice (for example ``df.loc[:]`` or ``df[:]``) (:issue:`49469`)
 - Disallow computing ``cumprod`` for :class:`Timedelta` object; previously this returned incorrect values (:issue:`50246`)
+- Instantiating an :class:`Index` with an numeric numpy dtype with data containing :class:`NA` and/or :class:`NaT` now raises a ``ValueError``. Previously a ``TypeError`` was raised (:issue:`51050`)
 - Loading a JSON file with duplicate columns using ``read_json(orient='split')`` renames columns to avoid duplicates, as :func:`read_csv` and the other readers do (:issue:`50370`)
 - The levels of the index of the :class:`Series` returned from ``Series.sparse.from_coo`` now always have dtype ``int32``. Previously they had dtype ``int64`` (:issue:`50926`)
 - :func:`to_datetime` with ``unit`` of either "Y" or "M" will now raise if a sequence contains a non-round ``float`` value, matching the ``Timestamp`` behavior (:issue:`50301`)

pandas/core/indexes/base.py

@@ -538,6 +538,10 @@ def _ensure_array(cls, data, dtype, copy: bool):
         if data.ndim > 1:
             # GH#13601, GH#20285, GH#27125
             raise ValueError("Index data must be 1-dimensional")
+        elif dtype == np.float16:
+            # float16 not supported (no indexing engine)
+            raise NotImplementedError("float16 indexes are not supported")
+
         if copy:
             # asarray_tuplesafe does not always copy underlying data,
             #  so need to make sure that this happens

pandas/core/indexes/numeric.py

@@ -1,7 +1,5 @@
 from __future__ import annotations
 
-from typing import Callable
-
 import numpy as np
 
 from pandas._typing import Dtype
@@ -10,20 +8,7 @@
     doc,
 )
 
-from pandas.core.dtypes.common import (
-    is_dtype_equal,
-    is_integer_dtype,
-    is_numeric_dtype,
-    is_scalar,
-    pandas_dtype,
-)
-from pandas.core.dtypes.generic import ABCSeries
-
-from pandas.core.construction import sanitize_array
-from pandas.core.indexes.base import (
-    Index,
-    maybe_extract_name,
-)
+from pandas.core.indexes.base import Index
 
 
 class NumericIndex(Index):
@@ -64,102 +49,20 @@ class NumericIndex(Index):
     """
 
     _typ = "numericindex"
-    _values: np.ndarray
     _default_dtype: np.dtype | None = None
-    _dtype_validation_metadata: tuple[Callable[..., bool], str] = (
-        is_numeric_dtype,
-        "numeric type",
-    )
     _can_hold_strings = False
 
     def __new__(
         cls, data=None, dtype: Dtype | None = None, copy: bool = False, name=None
     ) -> NumericIndex:
-        name = maybe_extract_name(name, data, cls)
-
-        subarr = cls._ensure_array(data, dtype, copy)
-        return cls._simple_new(subarr, name=name)
-
-    @classmethod
-    def _ensure_array(cls, data, dtype, copy: bool):
-        """
-        Ensure we have a valid array to pass to _simple_new.
-        """
-        cls._validate_dtype(dtype)
-        if dtype == np.float16:
-
-            # float16 not supported (no indexing engine)
-            raise NotImplementedError("float16 indexes are not supported")
-
-        if not isinstance(data, (np.ndarray, Index)):
-            # Coerce to ndarray if not already ndarray or Index
-            if is_scalar(data):
-                cls._raise_scalar_data_error(data)
-
-            # other iterable of some kind
-            if not isinstance(data, (ABCSeries, list, tuple)):
-                data = list(data)
-
-            if isinstance(data, (list, tuple)):
-                if len(data):
-                    data = sanitize_array(data, index=None)
-                else:
-                    data = np.array([], dtype=np.int64)
-
-        dtype = cls._ensure_dtype(dtype)
-
-        if copy or not is_dtype_equal(data.dtype, dtype):
-            # TODO: the try/except below is because it's difficult to predict the error
-            # and/or error message from different combinations of data and dtype.
-            # Efforts to avoid this try/except welcome.
-            # See https://github.com/pandas-dev/pandas/pull/41153#discussion_r676206222
-            try:
-                subarr = np.array(data, dtype=dtype, copy=copy)
-                cls._validate_dtype(subarr.dtype)
-            except (TypeError, ValueError):
-                raise ValueError(f"data is not compatible with {cls.__name__}")
-            cls._assert_safe_casting(data, subarr)
-        else:
-            subarr = data
-
-        if subarr.ndim > 1:
-            # GH#13601, GH#20285, GH#27125
-            raise ValueError("Index data must be 1-dimensional")
-
-        subarr = np.asarray(subarr)
-        if subarr.dtype == "float16":
-            # float16 not supported (no indexing engine)
-            raise NotImplementedError("float16 indexes are not implemented")
-
-        return subarr
-
-    @classmethod
-    def _validate_dtype(cls, dtype: Dtype | None) -> None:
-        if dtype is None:
-            return
-
-        validation_func, expected = cls._dtype_validation_metadata
-        if not validation_func(dtype):
-            raise ValueError(
-                f"Incorrect `dtype` passed: expected {expected}, received {dtype}"
-            )
-
-    @classmethod
-    def _ensure_dtype(cls, dtype: Dtype | None) -> np.dtype | None:
-        """
-        Assumes dtype has already been validated.
-        """
-        if dtype is None:
-            return cls._default_dtype
-
-        dtype = pandas_dtype(dtype)
-        if not isinstance(dtype, np.dtype):
-            raise TypeError(f"{dtype} not a numpy type")
-        elif dtype == np.float16:
-            # float16 not supported (no indexing engine)
-            raise NotImplementedError("float16 indexes are not supported")
-
-        return dtype
+        # temporary scaffolding, will be removed soon.
+        if isinstance(data, list) and len(data) == 0:
+            data = np.array([], dtype=np.int64)
+        elif isinstance(data, range):
+            data = np.arange(data.start, data.stop, data.step, dtype=np.int64)
+        return super().__new__(
+            cls, data=data, dtype=dtype, copy=copy, name=name
+        )  # type: ignore[return-value]
 
     # ----------------------------------------------------------------
     # Indexing Methods
@@ -168,17 +71,3 @@ def _ensure_dtype(cls, dtype: Dtype | None) -> np.dtype | None:
     @doc(Index._should_fallback_to_positional)
     def _should_fallback_to_positional(self) -> bool:
         return False
-
-    # ----------------------------------------------------------------
-
-    @classmethod
-    def _assert_safe_casting(cls, data: np.ndarray, subarr: np.ndarray) -> None:
-        """
-        Ensure incoming data can be represented with matching signed-ness.
-
-        Needed if the process of casting data from some accepted dtype to the internal
-        dtype(s) bears the risk of truncation (e.g. float to int).
-        """
-        if is_integer_dtype(subarr.dtype):
-            if not np.array_equal(data, subarr):
-                raise TypeError("Unsafe NumPy casting, you must explicitly cast")

pandas/core/indexes/range.py

@@ -98,6 +98,7 @@ class RangeIndex(NumericIndex):
     _typ = "rangeindex"
     _dtype_validation_metadata = (is_signed_integer_dtype, "signed integer")
     _range: range
+    _values: np.ndarray
 
     @property
     def _engine_type(self) -> type[libindex.Int64Engine]:
@@ -178,6 +179,17 @@ def _simple_new(  # type: ignore[override]
         result._reset_identity()
         return result
 
+    @classmethod
+    def _validate_dtype(cls, dtype: Dtype | None) -> None:
+        if dtype is None:
+            return
+
+        validation_func, expected = cls._dtype_validation_metadata
+        if not validation_func(dtype):
+            raise ValueError(
+                f"Incorrect `dtype` passed: expected {expected}, received {dtype}"
+            )
+
     # --------------------------------------------------------------------
 
     # error: Return type "Type[NumericIndex]" of "_constructor" incompatible with return

pandas/tests/indexes/interval/test_constructors.py

@@ -39,8 +39,8 @@ class ConstructorTests:
         params=[
             ([3, 14, 15, 92, 653], np.int64),
             (np.arange(10, dtype="int64"), np.int64),
-            (NumericIndex(range(-10, 11), dtype=np.int64), np.int64),
-            (NumericIndex(range(10, 31), dtype=np.uint64), np.uint64),
+            (NumericIndex(np.arange(-10, 11, dtype=np.int64)), np.int64),
+            (NumericIndex(np.arange(10, 31, dtype=np.uint64)), np.uint64),
             (NumericIndex(np.arange(20, 30, 0.5), dtype=np.float64), np.float64),
             (date_range("20180101", periods=10), "<M8[ns]"),
             (

pandas/tests/indexes/test_base.py

@@ -309,7 +309,6 @@ def test_constructor_dtypes_timedelta(self, attr, klass):
         "klass",
         [
             Index,
-            NumericIndex,
             CategoricalIndex,
             DatetimeIndex,
             TimedeltaIndex,
@@ -873,8 +872,11 @@ def test_isin_nan_common_float64(self, nulls_fixture):
         if nulls_fixture is pd.NaT or nulls_fixture is pd.NA:
             # Check 1) that we cannot construct a float64 Index with this value
             #  and 2) that with an NaN we do not have .isin(nulls_fixture)
-            msg = "data is not compatible with NumericIndex"
-            with pytest.raises(ValueError, match=msg):
+            msg = (
+                r"float\(\) argument must be a string or a (real )?number, "
+                f"not {repr(type(nulls_fixture).__name__)}"
+            )
+            with pytest.raises(TypeError, match=msg):
                 NumericIndex([1.0, nulls_fixture], dtype=np.float64)
 
             idx = NumericIndex([1.0, np.nan], dtype=np.float64)