Backport PR #57265 on branch 2.2.x (COMPAT: Numpy 2.0 casting compat) (#57271)

Backport PR #57265: COMPAT: Numpy 2.0 casting compat

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

Author: meeseeksmachine

pandas/core/dtypes/cast.py

@@ -1682,6 +1682,7 @@ def maybe_cast_to_integer_array(arr: list | np.ndarray, dtype: np.dtype) -> np.n
     arr = np.asarray(arr)
 
     if np.issubdtype(arr.dtype, str):
+        # TODO(numpy-2.0 min): This case will raise an OverflowError above
         if (casted.astype(str) == arr).all():
             return casted
         raise ValueError(f"string values cannot be losslessly cast to {dtype}")

pandas/core/internals/blocks.py

@@ -1425,7 +1425,14 @@ def setitem(self, indexer, value, using_cow: bool = False) -> Block:
             if isinstance(casted, np.ndarray) and casted.ndim == 1 and len(casted) == 1:
                 # NumPy 1.25 deprecation: https://github.com/numpy/numpy/pull/10615
                 casted = casted[0, ...]
-            values[indexer] = casted
+            try:
+                values[indexer] = casted
+            except (TypeError, ValueError) as err:
+                if is_list_like(casted):
+                    raise ValueError(
+                        "setting an array element with a sequence."
+                    ) from err
+                raise
         return self
 
     def putmask(

pandas/tests/indexing/test_loc.py

@@ -1235,13 +1235,7 @@ def test_loc_setitem_empty_append_raises(self):
         with pytest.raises(KeyError, match=msg):
             df.loc[[0, 1], "x"] = data
 
-        msg = "|".join(
-            [
-                "cannot copy sequence with size 2 to array axis with dimension 0",
-                r"could not broadcast input array from shape \(2,\) into shape \(0,\)",
-                "Must have equal len keys and value when setting with an iterable",
-            ]
-        )
+        msg = "setting an array element with a sequence."
         with pytest.raises(ValueError, match=msg):
             df.loc[0:2, "x"] = data
 
@@ -1575,16 +1569,10 @@ def test_loc_setitem_2d_to_1d_raises(self):
         # float64 dtype to avoid upcast when trying to set float data
         ser = Series(range(2), dtype="float64")
 
-        msg = "|".join(
-            [
-                r"shape mismatch: value array of shape \(2,2\)",
-                r"cannot reshape array of size 4 into shape \(2,\)",
-            ]
-        )
+        msg = "setting an array element with a sequence."
         with pytest.raises(ValueError, match=msg):
             ser.loc[range(2)] = data
 
-        msg = r"could not broadcast input array from shape \(2,2\) into shape \(2,?\)"
         with pytest.raises(ValueError, match=msg):
             ser.loc[:] = data
 

pandas/tests/series/test_constructors.py

@@ -1958,9 +1958,15 @@ def test_constructor_int64_dtype(self, any_int_dtype):
 
     def test_constructor_raise_on_lossy_conversion_of_strings(self):
         # GH#44923
-        with pytest.raises(
-            ValueError, match="string values cannot be losslessly cast to int8"
-        ):
+        if not np_version_gt2:
+            raises = pytest.raises(
+                ValueError, match="string values cannot be losslessly cast to int8"
+            )
+        else:
+            raises = pytest.raises(
+                OverflowError, match="The elements provided in the data"
+            )
+        with raises:
             Series(["128"], dtype="int8")
 
     def test_constructor_dtype_timedelta_alternative_construct(self):