Merge pull request #1964 from IntelPython/usm_ndrray-array-protocol-must-raise

Usm ndrray array protocol must raise

Author: oleksandr-pavlyk

CHANGELOG.md

@@ -21,6 +21,7 @@ and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0
 * Improved performance of `tensor.cumulative_sum`, `tensor.cumulative_prod`, `tensor.cumulative_logsumexp` as well as performance of boolean indexing [gh-1923](https://github.com/IntelPython/dpctl/pull/1923)
 * Improved performance of `tensor.min`, `tensor.max`, `tensor.logsumexp`, `tensor.reduce_hypot` for floating point type arrays by at least 2x [gh-1932](https://github.com/IntelPython/dpctl/pull/1932)
 * Extended `tensor.asarray` to support objects that implement `__usm_ndarray__` property to be interpreted as `usm_ndarray` objects [gh-1959](https://github.com/IntelPython/dpctl/pull/1959)
+* `dpctl.tensor.usm_ndarray` object disallows implicit conversions to NumPy array [gh-1964](https://github.com/IntelPython/dpctl/pull/1964)
 
 ### Fixed
 

dpctl/tensor/_usmarray.pyx

@@ -1552,6 +1552,21 @@ cdef class usm_ndarray:
     def __repr__(self):
         return usm_ndarray_repr(self)
 
+    def __array__(self, dtype=None, /, *, copy=None):
+        """NumPy's array protocol method to disallow implicit conversion.
+
+	Without this definition, `numpy.asarray(usm_ar)` converts
+	usm_ndarray instance into NumPy array with data type `object`
+	and every element being 0d usm_ndarray.
+
+        https://github.com/IntelPython/dpctl/pull/1384#issuecomment-1707212972
+	"""
+        raise TypeError(
+            "Implicit conversion to a NumPy array is not allowed. "
+	    "Use `dpctl.tensor.asnumpy` to copy data from this "
+	    "`dpctl.tensor.usm_ndarray` instance to NumPy array"
+	)
+
 
 cdef usm_ndarray _real_view(usm_ndarray ary):
     """

dpctl/tests/test_usm_ndarray_ctor.py

@@ -2659,3 +2659,12 @@ def test_setitem_copy_as_contig_alignment(dt):
     x[1:, ...] = vals
     assert dpt.all(x[0] == 0)
     assert dpt.all(x[1:, :] == vals)
+
+
+def test_asarray_property():
+    get_queue_or_skip()
+
+    x = dpt.ones(11, dtype="i4")
+
+    with pytest.raises(TypeError):
+        np.asarray(x)