REF: de-duplicate object-dtype handling (#32168)

Author: jbrockmendel

pandas/core/indexes/base.py

@@ -311,14 +311,7 @@ def __new__(
             # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
             from pandas.core.indexes.interval import IntervalIndex
 
-            closed = kwargs.pop("closed", None)
-            if is_dtype_equal(_o_dtype, dtype):
-                return IntervalIndex(
-                    data, name=name, copy=copy, closed=closed, **kwargs
-                ).astype(object)
-            return IntervalIndex(
-                data, dtype=dtype, name=name, copy=copy, closed=closed, **kwargs
-            )
+            return _maybe_asobject(dtype, IntervalIndex, data, copy, name, **kwargs)
 
         elif (
             is_datetime64_any_dtype(data)
@@ -328,39 +321,19 @@ def __new__(
             # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
             from pandas import DatetimeIndex
 
-            if is_dtype_equal(_o_dtype, dtype):
-                # GH#23524 passing `dtype=object` to DatetimeIndex is invalid,
-                #  will raise in the where `data` is already tz-aware.  So
-                #  we leave it out of this step and cast to object-dtype after
-                #  the DatetimeIndex construction.
-                # Note we can pass copy=False because the .astype below
-                #  will always make a copy
-                return DatetimeIndex(data, copy=False, name=name, **kwargs).astype(
-                    object
-                )
-            else:
-                return DatetimeIndex(data, copy=copy, name=name, dtype=dtype, **kwargs)
+            return _maybe_asobject(dtype, DatetimeIndex, data, copy, name, **kwargs)
 
         elif is_timedelta64_dtype(data) or is_timedelta64_dtype(dtype):
             # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
             from pandas import TimedeltaIndex
 
-            if is_dtype_equal(_o_dtype, dtype):
-                # Note we can pass copy=False because the .astype below
-                #  will always make a copy
-                return TimedeltaIndex(data, copy=False, name=name, **kwargs).astype(
-                    object
-                )
-            else:
-                return TimedeltaIndex(data, copy=copy, name=name, dtype=dtype, **kwargs)
+            return _maybe_asobject(dtype, TimedeltaIndex, data, copy, name, **kwargs)
 
         elif is_period_dtype(data) or is_period_dtype(dtype):
             # Delay import for perf. https://github.com/pandas-dev/pandas/pull/31423
             from pandas import PeriodIndex
 
-            if is_dtype_equal(_o_dtype, dtype):
-                return PeriodIndex(data, copy=False, name=name, **kwargs).astype(object)
-            return PeriodIndex(data, dtype=dtype, copy=copy, name=name, **kwargs)
+            return _maybe_asobject(dtype, PeriodIndex, data, copy, name, **kwargs)
 
         # extension dtype
         elif is_extension_array_dtype(data) or is_extension_array_dtype(dtype):
@@ -5775,3 +5748,40 @@ def _try_convert_to_int_array(
         pass
 
     raise ValueError
+
+
+def _maybe_asobject(dtype, klass, data, copy: bool, name: Label, **kwargs):
+    """
+    If an object dtype was specified, create the non-object Index
+    and then convert it to object.
+
+    Parameters
+    ----------
+    dtype : np.dtype, ExtensionDtype, str
+    klass : Index subclass
+    data : list-like
+    copy : bool
+    name : hashable
+    **kwargs
+
+    Returns
+    -------
+    Index
+
+    Notes
+    -----
+    We assume that calling .astype(object) on this klass will make a copy.
+    """
+
+    # GH#23524 passing `dtype=object` to DatetimeIndex is invalid,
+    #  will raise in the where `data` is already tz-aware.  So
+    #  we leave it out of this step and cast to object-dtype after
+    #  the DatetimeIndex construction.
+
+    if is_dtype_equal(_o_dtype, dtype):
+        # Note we can pass copy=False because the .astype below
+        #  will always make a copy
+        index = klass(data, copy=False, name=name, **kwargs)
+        return index.astype(object)
+
+    return klass(data, dtype=dtype, copy=copy, name=name, **kwargs)