Merge pull request #62 from Quansight-Labs/f_contiguous

add ndarrays.flags.f_contiguous

Author: ev-br

torch_np/_ndarray.py

@@ -112,7 +112,22 @@ def base(self):
     @property
     def flags(self):
         # Note contiguous in torch is assumed C-style
-        return Flags({"C_CONTIGUOUS": self._tensor.is_contiguous()})
+
+        # check if F contiguous
+        from itertools import accumulate
+
+        f_strides = tuple(accumulate(list(self._tensor.shape), func=lambda x, y: x * y))
+        f_strides = (1,) + f_strides[:-1]
+        is_f_contiguous = f_strides == self._tensor.stride()
+
+        return Flags(
+            {
+                "C_CONTIGUOUS": self._tensor.is_contiguous(),
+                "F_CONTIGUOUS": is_f_contiguous,
+                "OWNDATA": self._tensor._base is None,
+                "WRITEABLE": True,  # pytorch does not have readonly tensors
+            }
+        )
 
     @property
     def T(self):

torch_np/tests/numpy_tests/core/test_indexing.py

@@ -377,7 +377,6 @@ def test_subclass_writeable(self):
         assert_(d[...].flags.writeable)
         assert_(d[0].flags.writeable)
 
-    @pytest.mark.xfail(reason="can't determine f-style contiguous in torch")
     def test_memory_order(self):
         # This is not necessary to preserve. Memory layouts for
         # more complex indices are not as simple.

torch_np/tests/numpy_tests/core/test_numeric.py

@@ -2194,7 +2194,7 @@ def check_function(self, func, fill_value=None):
                        **fill_kwarg)
 
             assert_equal(arr.dtype, dtype)
-            # assert_(getattr(arr.flags, self.orders[order]))   # XXX: no ndarray.flags
+            assert_(getattr(arr.flags, self.orders[order]))
 
             if fill_value is not None:
                 val = fill_value

torch_np/tests/numpy_tests/lib/test_function_base.py

@@ -272,7 +272,7 @@ def test_basic(self):
         assert_equal(a[0, 0], 1)
         assert_equal(a_copy[0, 0], 10)
 
-    @pytest.mark.xfail(reason="ndarray.flags not implemented")
+    @pytest.mark.xfail(reason="order='F' not implemented")
     def test_order(self):
         # It turns out that people rely on np.copy() preserving order by
         # default; changing this broke scikit-learn: