BUG: 2D ndarray of dtype 'object' is always copied upon construction (pandas-dev#39263)

Author: irgolic

doc/source/whatsnew/v1.3.0.rst

@@ -280,6 +280,7 @@ Datetimelike
 - Bug in comparisons between :class:`Timestamp` object and ``datetime64`` objects just outside the implementation bounds for nanosecond ``datetime64`` (:issue:`39221`)
 - Bug in :meth:`Timestamp.round`, :meth:`Timestamp.floor`, :meth:`Timestamp.ceil` for values near the implementation bounds of :class:`Timestamp` (:issue:`39244`)
 - Bug in :func:`date_range` incorrectly creating :class:`DatetimeIndex` containing ``NaT`` instead of raising ``OutOfBoundsDatetime`` in corner cases (:issue:`24124`)
+- Bug in :class:`DataFrame` constructor always copying 2D object arrays (:issue:`39272`)
 
 Timedelta
 ^^^^^^^^^

pandas/core/internals/__init__.py

@@ -18,7 +18,6 @@
     BlockManager,
     SingleBlockManager,
     create_block_manager_from_arrays,
-    create_block_manager_from_blocks,
 )
 
 __all__ = [
@@ -40,5 +39,4 @@
     "concatenate_block_managers",
     # those two are preserved here for downstream compatibility (GH-33892)
     "create_block_manager_from_arrays",
-    "create_block_manager_from_blocks",
 ]

pandas/core/internals/construction.py

@@ -60,8 +60,8 @@
     union_indexes,
 )
 from pandas.core.internals.managers import (
+    create_block_manager_from_array,
     create_block_manager_from_arrays,
-    create_block_manager_from_blocks,
 )
 
 if TYPE_CHECKING:
@@ -230,36 +230,29 @@ def init_ndarray(values, index, columns, dtype: Optional[DtypeObj], copy: bool):
     index, columns = _get_axes(
         values.shape[0], values.shape[1], index=index, columns=columns
     )
-    values = values.T
+
+    array = values.T
 
     # if we don't have a dtype specified, then try to convert objects
     # on the entire block; this is to convert if we have datetimelike's
     # embedded in an object type
-    if dtype is None and is_object_dtype(values.dtype):
-
-        if values.ndim == 2 and values.shape[0] != 1:
-            # transpose and separate blocks
-
-            dvals_list = [maybe_infer_to_datetimelike(row) for row in values]
-            for n in range(len(dvals_list)):
-                if isinstance(dvals_list[n], np.ndarray):
-                    dvals_list[n] = dvals_list[n].reshape(1, -1)
-
-            from pandas.core.internals.blocks import make_block
-
-            # TODO: What about re-joining object columns?
-            block_values = [
-                make_block(dvals_list[n], placement=[n], ndim=2)
-                for n in range(len(dvals_list))
+    if dtype is None and is_object_dtype(array.dtype):
+        if array.ndim == 2 and array.shape[0] != 1:
+            maybe_datetime = [
+                maybe_infer_to_datetimelike(instance) for instance in array
             ]
-
+            # don't convert (and copy) the objects if no type inference occurs
+            if any(
+                not is_dtype_equal(instance.dtype, array.dtype)
+                for instance in maybe_datetime
+            ):
+                return create_block_manager_from_arrays(
+                    maybe_datetime, columns, [columns, index]
+                )
         else:
-            datelike_vals = maybe_infer_to_datetimelike(values)
-            block_values = [datelike_vals]
-    else:
-        block_values = [values]
+            array = maybe_infer_to_datetimelike(array)
 
-    return create_block_manager_from_blocks(block_values, [columns, index])
+    return create_block_manager_from_array(array, [columns, index])
 
 
 def init_dict(data: Dict, index, columns, dtype: Optional[DtypeObj] = None):

pandas/core/internals/managers.py

@@ -1633,32 +1633,6 @@ def fast_xs(self, loc):
 # Constructor Helpers
 
 
-def create_block_manager_from_blocks(blocks, axes: List[Index]) -> BlockManager:
-    try:
-        if len(blocks) == 1 and not isinstance(blocks[0], Block):
-            # if blocks[0] is of length 0, return empty blocks
-            if not len(blocks[0]):
-                blocks = []
-            else:
-                # It's OK if a single block is passed as values, its placement
-                # is basically "all items", but if there're many, don't bother
-                # converting, it's an error anyway.
-                blocks = [
-                    make_block(
-                        values=blocks[0], placement=slice(0, len(axes[0])), ndim=2
-                    )
-                ]
-
-        mgr = BlockManager(blocks, axes)
-        mgr._consolidate_inplace()
-        return mgr
-
-    except ValueError as e:
-        blocks = [getattr(b, "values", b) for b in blocks]
-        tot_items = sum(b.shape[0] for b in blocks)
-        raise construction_error(tot_items, blocks[0].shape[1:], axes, e)
-
-
 def create_block_manager_from_arrays(
     arrays, names: Index, axes: List[Index]
 ) -> BlockManager:
@@ -1678,6 +1652,19 @@ def create_block_manager_from_arrays(
         raise construction_error(len(arrays), arrays[0].shape, axes, e)
 
 
+def create_block_manager_from_array(
+    array,
+    axes: List[Index],
+) -> BlockManager:
+    try:
+        block = make_block(values=array, placement=slice(0, len(axes[0])), ndim=2)
+        mgr = BlockManager([block], axes)
+        mgr._consolidate_inplace()
+    except ValueError as e:
+        raise construction_error(array.shape[0], array.shape[1:], axes, e)
+    return mgr
+
+
 def construction_error(tot_items, block_shape, axes, e=None):
     """ raise a helpful message about our construction """
     passed = tuple(map(int, [tot_items] + list(block_shape)))

pandas/tests/frame/test_constructors.py

@@ -1909,6 +1909,14 @@ def test_constructor_series_copy(self, float_frame):
 
         assert not (series["A"] == 5).all()
 
+    def test_object_array_does_not_copy(self):
+        a = np.array(["a", "b"], dtype="object")
+        b = np.array([["a", "b"], ["c", "d"]], dtype="object")
+        df = DataFrame(a)
+        assert np.shares_memory(df.values, a)
+        df2 = DataFrame(b)
+        assert np.shares_memory(df2.values, b)
+
     def test_constructor_with_nas(self):
         # GH 5016
         # na's in indices