REF: use maybe_convert_objects in pd.array (#56484)

* REF: use maybe_convert_objects in pd.array

* lint fixups

* Update pandas/_libs/lib.pyx

Co-authored-by: Matthew Roeschke <10647082+mroeschke@users.noreply.github.com>

---------

Co-authored-by: Matthew Roeschke <10647082+mroeschke@users.noreply.github.com>

Author: jbrockmendel

pandas/_libs/lib.pyx

@@ -2628,7 +2628,11 @@ def maybe_convert_objects(ndarray[object] objects,
                 seen.object_ = True
                 break
         elif val is C_NA:
-            seen.object_ = True
+            if convert_to_nullable_dtype:
+                seen.null_ = True
+                mask[i] = True
+            else:
+                seen.object_ = True
             continue
         else:
             seen.object_ = True
@@ -2691,6 +2695,12 @@ def maybe_convert_objects(ndarray[object] objects,
             dtype = StringDtype(storage="pyarrow_numpy")
             return dtype.construct_array_type()._from_sequence(objects, dtype=dtype)
 
+        elif convert_to_nullable_dtype and is_string_array(objects, skipna=True):
+            from pandas.core.arrays.string_ import StringDtype
+
+            dtype = StringDtype()
+            return dtype.construct_array_type()._from_sequence(objects, dtype=dtype)
+
         seen.object_ = True
     elif seen.interval_:
         if is_interval_array(objects):
@@ -2734,12 +2744,12 @@ def maybe_convert_objects(ndarray[object] objects,
         return objects
 
     if seen.bool_:
-        if seen.is_bool:
-            # is_bool property rules out everything else
-            return bools.view(np.bool_)
-        elif convert_to_nullable_dtype and seen.is_bool_or_na:
+        if convert_to_nullable_dtype and seen.is_bool_or_na:
             from pandas.core.arrays import BooleanArray
             return BooleanArray(bools.view(np.bool_), mask)
+        elif seen.is_bool:
+            # is_bool property rules out everything else
+            return bools.view(np.bool_)
         seen.object_ = True
 
     if not seen.object_:
@@ -2752,11 +2762,11 @@ def maybe_convert_objects(ndarray[object] objects,
                     result = floats
                 elif seen.int_ or seen.uint_:
                     if convert_to_nullable_dtype:
-                        from pandas.core.arrays import IntegerArray
+                        # Below we will wrap in IntegerArray
                         if seen.uint_:
-                            result = IntegerArray(uints, mask)
+                            result = uints
                         else:
-                            result = IntegerArray(ints, mask)
+                            result = ints
                     else:
                         result = floats
                 elif seen.nan_:
@@ -2771,7 +2781,6 @@ def maybe_convert_objects(ndarray[object] objects,
                         result = uints
                     else:
                         result = ints
-
         else:
             # don't cast int to float, etc.
             if seen.null_:
@@ -2794,6 +2803,22 @@ def maybe_convert_objects(ndarray[object] objects,
                     else:
                         result = ints
 
+        # TODO: do these after the itemsize check?
+        if (result is ints or result is uints) and convert_to_nullable_dtype:
+            from pandas.core.arrays import IntegerArray
+
+            # Set these values to 1 to be deterministic, match
+            #  IntegerArray._internal_fill_value
+            result[mask] = 1
+            result = IntegerArray(result, mask)
+        elif result is floats and convert_to_nullable_dtype:
+            from pandas.core.arrays import FloatingArray
+
+            # Set these values to 1.0 to be deterministic, match
+            #  FloatingArray._internal_fill_value
+            result[mask] = 1.0
+            result = FloatingArray(result, mask)
+
         if result is uints or result is ints or result is floats or result is complexes:
             # cast to the largest itemsize when all values are NumPy scalars
             if itemsize_max > 0 and itemsize_max != result.dtype.itemsize:

pandas/core/construction.py

@@ -7,11 +7,8 @@
 
 from __future__ import annotations
 
-from collections.abc import Sequence
 from typing import (
     TYPE_CHECKING,
-    Optional,
-    Union,
     cast,
     overload,
 )
@@ -23,17 +20,9 @@
 
 from pandas._libs import lib
 from pandas._libs.tslibs import (
-    Period,
     get_supported_dtype,
     is_supported_dtype,
 )
-from pandas._typing import (
-    AnyArrayLike,
-    ArrayLike,
-    Dtype,
-    DtypeObj,
-    T,
-)
 
 from pandas.core.dtypes.base import ExtensionDtype
 from pandas.core.dtypes.cast import (
@@ -46,6 +35,7 @@
     maybe_promote,
 )
 from pandas.core.dtypes.common import (
+    ensure_object,
     is_list_like,
     is_object_dtype,
     is_string_dtype,
@@ -63,11 +53,25 @@
 import pandas.core.common as com
 
 if TYPE_CHECKING:
+    from collections.abc import Sequence
+
+    from pandas._typing import (
+        AnyArrayLike,
+        ArrayLike,
+        Dtype,
+        DtypeObj,
+        T,
+    )
+
     from pandas import (
         Index,
         Series,
     )
-    from pandas.core.arrays.base import ExtensionArray
+    from pandas.core.arrays import (
+        DatetimeArray,
+        ExtensionArray,
+        TimedeltaArray,
+    )
 
 
 def array(
@@ -286,9 +290,7 @@ def array(
         ExtensionArray,
         FloatingArray,
         IntegerArray,
-        IntervalArray,
         NumpyExtensionArray,
-        PeriodArray,
         TimedeltaArray,
     )
     from pandas.core.arrays.string_ import StringDtype
@@ -320,46 +322,58 @@ def array(
         return cls._from_sequence(data, dtype=dtype, copy=copy)
 
     if dtype is None:
-        inferred_dtype = lib.infer_dtype(data, skipna=True)
-        if inferred_dtype == "period":
-            period_data = cast(Union[Sequence[Optional[Period]], AnyArrayLike], data)
-            return PeriodArray._from_sequence(period_data, copy=copy)
-
-        elif inferred_dtype == "interval":
-            return IntervalArray(data, copy=copy)
-
-        elif inferred_dtype.startswith("datetime"):
-            # datetime, datetime64
-            try:
-                return DatetimeArray._from_sequence(data, copy=copy)
-            except ValueError:
-                # Mixture of timezones, fall back to NumpyExtensionArray
-                pass
-
-        elif inferred_dtype.startswith("timedelta"):
-            # timedelta, timedelta64
-            return TimedeltaArray._from_sequence(data, copy=copy)
-
-        elif inferred_dtype == "string":
+        was_ndarray = isinstance(data, np.ndarray)
+        # error: Item "Sequence[object]" of "Sequence[object] | ExtensionArray |
+        # ndarray[Any, Any]" has no attribute "dtype"
+        if not was_ndarray or data.dtype == object:  # type: ignore[union-attr]
+            result = lib.maybe_convert_objects(
+                ensure_object(data),
+                convert_non_numeric=True,
+                convert_to_nullable_dtype=True,
+                dtype_if_all_nat=None,
+            )
+            result = ensure_wrapped_if_datetimelike(result)
+            if isinstance(result, np.ndarray):
+                if len(result) == 0 and not was_ndarray:
+                    # e.g. empty list
+                    return FloatingArray._from_sequence(data, dtype="Float64")
+                return NumpyExtensionArray._from_sequence(
+                    data, dtype=result.dtype, copy=copy
+                )
+            if result is data and copy:
+                return result.copy()
+            return result
+
+        data = cast(np.ndarray, data)
+        result = ensure_wrapped_if_datetimelike(data)
+        if result is not data:
+            result = cast("DatetimeArray | TimedeltaArray", result)
+            if copy and result.dtype == data.dtype:
+                return result.copy()
+            return result
+
+        if data.dtype.kind in "SU":
             # StringArray/ArrowStringArray depending on pd.options.mode.string_storage
             dtype = StringDtype()
             cls = dtype.construct_array_type()
             return cls._from_sequence(data, dtype=dtype, copy=copy)
 
-        elif inferred_dtype == "integer":
+        elif data.dtype.kind in "iu":
             return IntegerArray._from_sequence(data, copy=copy)
-        elif inferred_dtype == "empty" and not hasattr(data, "dtype") and not len(data):
-            return FloatingArray._from_sequence(data, copy=copy)
-        elif (
-            inferred_dtype in ("floating", "mixed-integer-float")
-            and getattr(data, "dtype", None) != np.float16
-        ):
+        elif data.dtype.kind == "f":
             # GH#44715 Exclude np.float16 bc FloatingArray does not support it;
             #  we will fall back to NumpyExtensionArray.
+            if data.dtype == np.float16:
+                return NumpyExtensionArray._from_sequence(
+                    data, dtype=data.dtype, copy=copy
+                )
             return FloatingArray._from_sequence(data, copy=copy)
 
-        elif inferred_dtype == "boolean":
+        elif data.dtype.kind == "b":
             return BooleanArray._from_sequence(data, dtype="boolean", copy=copy)
+        else:
+            # e.g. complex
+            return NumpyExtensionArray._from_sequence(data, dtype=data.dtype, copy=copy)
 
     # Pandas overrides NumPy for
     #   1. datetime64[ns,us,ms,s]

pandas/tests/arrays/test_array.py

@@ -220,6 +220,14 @@ def test_dt64_array(dtype_unit):
             .construct_array_type()
             ._from_sequence(["a", None], dtype=pd.StringDtype()),
         ),
+        (
+            # numpy array with string dtype
+            np.array(["a", "b"], dtype=str),
+            None,
+            pd.StringDtype()
+            .construct_array_type()
+            ._from_sequence(["a", "b"], dtype=pd.StringDtype()),
+        ),
         # Boolean
         (
             [True, None],
@@ -247,6 +255,14 @@ def test_dt64_array(dtype_unit):
             "category",
             pd.Categorical([pd.Period("2000", "D"), pd.Period("2001", "D")]),
         ),
+        # Complex
+        (
+            np.array([complex(1), complex(2)], dtype=np.complex128),
+            None,
+            NumpyExtensionArray(
+                np.array([complex(1), complex(2)], dtype=np.complex128)
+            ),
+        ),
     ],
 )
 def test_array(data, dtype, expected):

pandas/tests/dtypes/test_inference.py

@@ -936,9 +936,9 @@ def test_maybe_convert_objects_bool_nan(self):
     def test_maybe_convert_objects_nullable_boolean(self):
         # GH50047
         arr = np.array([True, False], dtype=object)
-        exp = np.array([True, False])
+        exp = BooleanArray._from_sequence([True, False], dtype="boolean")
         out = lib.maybe_convert_objects(arr, convert_to_nullable_dtype=True)
-        tm.assert_numpy_array_equal(out, exp)
+        tm.assert_extension_array_equal(out, exp)
 
         arr = np.array([True, False, pd.NaT], dtype=object)
         exp = np.array([True, False, pd.NaT], dtype=object)