COMPAT: np 1.18 wants explicit dtype=object)

pandas/core/arrays/numpy_.py

@@ -1,5 +1,6 @@
 import numbers
 from typing import Union
+import warnings
 
 import numpy as np
 from numpy.lib.mixins import NDArrayOperatorsMixin
@@ -158,7 +159,11 @@ def _from_sequence(cls, scalars, dtype=None, copy=False):
         if isinstance(dtype, PandasDtype):
             dtype = dtype._dtype
 
-        result = np.asarray(scalars, dtype=dtype)
+        with warnings.catch_warnings():
+            # See https://github.com/numpy/numpy/issues/15041
+            warnings.filterwarnings("ignore", ".*with automatic object dtype.*")
+            result = np.asarray(scalars, dtype=dtype)
+
         if copy and result is scalars:
             result = result.copy()
         return cls(result)

pandas/core/common.py

@@ -10,6 +10,7 @@
 from functools import partial
 import inspect
 from typing import Any, Iterable, Union
+import warnings
 
 import numpy as np
 
@@ -224,7 +225,13 @@ def asarray_tuplesafe(values, dtype=None):
     if isinstance(values, list) and dtype in [np.object_, object]:
         return construct_1d_object_array_from_listlike(values)
 
-    result = np.asarray(values, dtype=dtype)
+    if dtype is None:
+        with warnings.catch_warnings():
+            # See https://github.com/numpy/numpy/issues/15041
+            warnings.filterwarnings("ignore", ".*with automatic object dtype.*")
+            result = np.asarray(values)
+    else:
+        result = np.asarray(values, dtype=dtype)
 
     if issubclass(result.dtype.type, str):
         result = np.asarray(values, dtype=object)

pandas/core/dtypes/cast.py

@@ -1,6 +1,7 @@
 """ routings for casting """
 
 from datetime import datetime, timedelta
+import warnings
 
 import numpy as np
 
@@ -1046,7 +1047,11 @@ def maybe_infer_to_datetimelike(value, convert_dates: bool = False):
 
     if not is_list_like(v):
         v = [v]
-    v = np.array(v, copy=False)
+
+    with warnings.catch_warnings():
+        # See https://github.com/numpy/numpy/issues/15041
+        warnings.filterwarnings("ignore", ".*with automatic object dtype.*")
+        v = np.array(v, copy=False)
 
     # we only care about object dtypes
     if not is_object_dtype(v):

pandas/core/dtypes/concat.py

@@ -1,6 +1,7 @@
 """
 Utility functions related to concat
 """
+import warnings
 
 import numpy as np
 
@@ -134,6 +135,7 @@ def is_nonempty(x) -> bool:
                 # coerce to object
                 to_concat = [x.astype("object") for x in to_concat]
 
+    to_concat = [_safe_array(x) for x in to_concat]
     return np.concatenate(to_concat, axis=axis)
 
 
@@ -172,7 +174,7 @@ def concat_categorical(to_concat, axis: int = 0):
     to_concat = [
         x._internal_get_values()
         if is_categorical_dtype(x.dtype)
-        else np.asarray(x).ravel()
+        else _safe_array(x).ravel()
         if not is_datetime64tz_dtype(x)
         else np.asarray(x.astype(object))
         for x in to_concat
@@ -183,6 +185,15 @@ def concat_categorical(to_concat, axis: int = 0):
     return result
 
 
+def _safe_array(x):
+    # FIXME: kludge
+    with warnings.catch_warnings():
+        # See https://github.com/numpy/numpy/issues/15041
+        warnings.filterwarnings("ignore", ".*with automatic object dtype.*")
+        arr = np.asarray(x)
+    return arr
+
+
 def union_categoricals(
     to_union, sort_categories: bool = False, ignore_order: bool = False
 ):

pandas/core/internals/blocks.py

@@ -1764,6 +1764,9 @@ def get_values(self, dtype=None):
         return values
 
     def to_dense(self):
+        if self.dtype.kind == "O":
+            # See https://github.com/numpy/numpy/issues/15041
+            return np.asarray(self.values, dtype=object)
         return np.asarray(self.values)
 
     def to_native_types(self, slicer=None, na_rep="nan", quoting=None, **kwargs):

pandas/io/formats/format.py

@@ -1485,6 +1485,11 @@ def _format_strings(self) -> List[str]:
         if is_categorical_dtype(values.dtype):
             # Categorical is special for now, so that we can preserve tzinfo
             array = values._internal_get_values()
+        elif values.dtype.kind == "O":
+            # numpy>=1.18 wants object dtype passed explicitly
+            # Note: dtype.kind check works for json extension tests, while
+            #  dtype == object check does not.
+            array = np.asarray(values, dtype=object)
         else:
             array = np.asarray(values)
 

pandas/tests/extension/base/interface.py

@@ -32,7 +32,11 @@ def test_memory_usage(self, data):
         assert result == s.nbytes
 
     def test_array_interface(self, data):
-        result = np.array(data)
+        if hasattr(data, "dtype") and data.dtype.kind == "O":
+            # e.g. JSONArray
+            result = np.array(data, dtype=object)
+        else:
+            result = np.array(data)
         assert result[0] == data[0]
 
         result = np.array(data, dtype=object)

pandas/tests/extension/test_numpy.py

@@ -51,7 +51,7 @@ def data_missing(allow_in_pandas, dtype):
     if dtype.numpy_dtype == "object":
         if _np_version_under1p16:
             raise pytest.skip("Skipping for NumPy <1.16")
-        return PandasArray(np.array([np.nan, (1,)]))
+        return PandasArray(np.array([np.nan, (1,)], dtype=object))
     return PandasArray(np.array([np.nan, 1.0]))
 
 
@@ -78,7 +78,7 @@ def data_for_sorting(allow_in_pandas, dtype):
     if dtype.numpy_dtype == "object":
         # Use an empty tuple for first element, then remove,
         # to disable np.array's shape inference.
-        return PandasArray(np.array([(), (2,), (3,), (1,)])[1:])
+        return PandasArray(np.array([(), (2,), (3,), (1,)], dtype=object)[1:])
     return PandasArray(np.array([1, 2, 0]))
 
 
@@ -90,7 +90,7 @@ def data_missing_for_sorting(allow_in_pandas, dtype):
     A < B and NA missing.
     """
     if dtype.numpy_dtype == "object":
-        return PandasArray(np.array([(1,), np.nan, (0,)]))
+        return PandasArray(np.array([(1,), np.nan, (0,)], dtype=object))
     return PandasArray(np.array([1, np.nan, 0]))
 
 
@@ -106,7 +106,9 @@ def data_for_grouping(allow_in_pandas, dtype):
         a, b, c = (1,), (2,), (3,)
     else:
         a, b, c = np.arange(3)
-    return PandasArray(np.array([b, b, np.nan, np.nan, a, a, b, c]))
+    return PandasArray(
+        np.array([b, b, np.nan, np.nan, a, a, b, c], dtype=dtype.numpy_dtype)
+    )
 
 
 @pytest.fixture