REF: tighter typing in constructor functions (pandas-dev#49591)

Author: jbrockmendel

pandas/core/construction.py

@@ -597,6 +597,15 @@ def sanitize_array(
                 #  e.g. test_constructor_floating_data_int_dtype
                 # TODO: where is the discussion that documents the reason for this?
                 subarr = np.array(data, copy=copy)
+
+        elif dtype is None:
+            subarr = data
+            if data.dtype == object:
+                subarr = maybe_infer_to_datetimelike(data)
+
+            if subarr is data and copy:
+                subarr = subarr.copy()
+
         else:
             # we will try to copy by-definition here
             subarr = _try_cast(data, dtype, copy)
@@ -754,7 +763,7 @@ def _maybe_repeat(arr: ArrayLike, index: Index | None) -> ArrayLike:
 
 def _try_cast(
     arr: list | np.ndarray,
-    dtype: np.dtype | None,
+    dtype: np.dtype,
     copy: bool,
 ) -> ArrayLike:
     """
@@ -764,7 +773,7 @@ def _try_cast(
     ----------
     arr : ndarray or list
         Excludes: ExtensionArray, Series, Index.
-    dtype : np.dtype or None
+    dtype : np.dtype
     copy : bool
         If False, don't copy the data if not needed.
 
@@ -774,30 +783,7 @@ def _try_cast(
     """
     is_ndarray = isinstance(arr, np.ndarray)
 
-    if dtype is None:
-        # perf shortcut as this is the most common case
-        if is_ndarray:
-            arr = cast(np.ndarray, arr)
-            if arr.dtype != object:
-                if copy:
-                    return arr.copy()
-                return arr
-
-            out = maybe_infer_to_datetimelike(arr)
-            if out is arr and copy:
-                out = out.copy()
-            return out
-
-        else:
-            # i.e. list
-            varr = np.array(arr, copy=False)
-            # filter out cases that we _dont_ want to go through
-            #  maybe_infer_to_datetimelike
-            if varr.dtype != object or varr.size == 0:
-                return varr
-            return maybe_infer_to_datetimelike(varr)
-
-    elif is_object_dtype(dtype):
+    if is_object_dtype(dtype):
         if not is_ndarray:
             subarr = construct_1d_object_array_from_listlike(arr)
             return subarr

pandas/core/frame.py

@@ -738,7 +738,7 @@ def __init__(
 
         # For data is list-like, or Iterable (will consume into list)
         elif is_list_like(data):
-            if not isinstance(data, (abc.Sequence, ExtensionArray)):
+            if not isinstance(data, abc.Sequence):
                 if hasattr(data, "__array__"):
                     # GH#44616 big perf improvement for e.g. pytorch tensor
                     data = np.asarray(data)

pandas/core/internals/construction.py

@@ -53,9 +53,7 @@
 )
 from pandas.core.arrays import (
     Categorical,
-    DatetimeArray,
     ExtensionArray,
-    TimedeltaArray,
 )
 from pandas.core.construction import (
     ensure_wrapped_if_datetimelike,
@@ -277,14 +275,20 @@ def ndarray_to_mgr(
 
         return arrays_to_mgr(values, columns, index, dtype=dtype, typ=typ)
 
-    elif is_extension_array_dtype(vdtype) and not is_1d_only_ea_dtype(vdtype):
-        # i.e. Datetime64TZ, PeriodDtype
+    elif is_extension_array_dtype(vdtype):
+        # i.e. Datetime64TZ, PeriodDtype; cases with is_1d_only_ea_dtype(vdtype)
+        #  are already caught above
         values = extract_array(values, extract_numpy=True)
         if copy:
             values = values.copy()
         if values.ndim == 1:
             values = values.reshape(-1, 1)
 
+    elif isinstance(values, (np.ndarray, ExtensionArray, ABCSeries, Index)):
+        # drop subclass info
+        values = np.array(values, copy=copy_on_sanitize)
+        values = _ensure_2d(values)
+
     else:
         # by definition an array here
         # the dtypes will be coerced to a single dtype
@@ -496,51 +500,50 @@ def treat_as_nested(data) -> bool:
 # ---------------------------------------------------------------------
 
 
-def _prep_ndarraylike(
-    values, copy: bool = True
-) -> np.ndarray | DatetimeArray | TimedeltaArray:
-    if isinstance(values, TimedeltaArray) or (
-        isinstance(values, DatetimeArray) and values.tz is None
-    ):
-        # By retaining DTA/TDA instead of unpacking, we end up retaining non-nano
-        pass
-
-    elif not isinstance(values, (np.ndarray, ABCSeries, Index)):
-        if len(values) == 0:
-            return np.empty((0, 0), dtype=object)
-        elif isinstance(values, range):
-            arr = range_to_ndarray(values)
-            return arr[..., np.newaxis]
-
-        def convert(v):
-            if not is_list_like(v) or isinstance(v, ABCDataFrame):
-                return v
-
-            v = extract_array(v, extract_numpy=True)
-            res = maybe_convert_platform(v)
-            return res
-
-        # we could have a 1-dim or 2-dim list here
-        # this is equiv of np.asarray, but does object conversion
-        # and platform dtype preservation
-        if is_list_like(values[0]):
-            values = np.array([convert(v) for v in values])
-        elif isinstance(values[0], np.ndarray) and values[0].ndim == 0:
-            # GH#21861 see test_constructor_list_of_lists
-            values = np.array([convert(v) for v in values])
-        else:
-            values = convert(values)
-
+def _prep_ndarraylike(values, copy: bool = True) -> np.ndarray:
+    # values is specifically _not_ ndarray, EA, Index, or Series
+    # We only get here with `not treat_as_nested(values)`
+
+    if len(values) == 0:
+        return np.empty((0, 0), dtype=object)
+    elif isinstance(values, range):
+        arr = range_to_ndarray(values)
+        return arr[..., np.newaxis]
+
+    def convert(v):
+        if not is_list_like(v) or isinstance(v, ABCDataFrame):
+            return v
+
+        v = extract_array(v, extract_numpy=True)
+        res = maybe_convert_platform(v)
+        # We don't do maybe_infer_to_datetimelike here bc we will end up doing
+        #  it column-by-column in ndarray_to_mgr
+        return res
+
+    # we could have a 1-dim or 2-dim list here
+    # this is equiv of np.asarray, but does object conversion
+    # and platform dtype preservation
+    # does not convert e.g. [1, "a", True] to ["1", "a", "True"] like
+    #  np.asarray would
+    if is_list_like(values[0]):
+        values = np.array([convert(v) for v in values])
+    elif isinstance(values[0], np.ndarray) and values[0].ndim == 0:
+        # GH#21861 see test_constructor_list_of_lists
+        values = np.array([convert(v) for v in values])
     else:
+        values = convert(values)
 
-        # drop subclass info
-        values = np.array(values, copy=copy)
+    return _ensure_2d(values)
 
+
+def _ensure_2d(values: np.ndarray) -> np.ndarray:
+    """
+    Reshape 1D values, raise on anything else other than 2D.
+    """
     if values.ndim == 1:
         values = values.reshape((values.shape[0], 1))
     elif values.ndim != 2:
         raise ValueError(f"Must pass 2-d input. shape={values.shape}")
-
     return values