BUG: torch/meshgrid: stop ignoring the "indexing" argument

Author: ev-br

array_api_compat/torch/_aliases.py

@@ -826,7 +826,7 @@ def sign(x: Array, /) -> Array:
 def meshgrid(*arrays: Array, indexing: Literal['xy', 'ij'] = 'xy') -> list[Array]:
     # enforce the default of 'xy'
     # TODO: is the return type a list or a tuple
-    return list(torch.meshgrid(*arrays, indexing='xy'))
+    return list(torch.meshgrid(*arrays, indexing=indexing))
 
 
 __all__ = ['asarray', 'result_type', 'can_cast',

tests/test_torch.py

@@ -117,3 +117,16 @@ def test_meshgrid():
 
     assert Y.shape == Y_xy.shape
     assert xp.all(Y == Y_xy)
+
+    # repeat with an explicit indexing
+    X, Y = xp.meshgrid(x, y, indexing='ij')
+
+    # output of torch.meshgrid(x, y, indexing='ij')
+    X_ij, Y_ij = xp.asarray([[1], [2]]), xp.asarray([[4], [4]])
+
+    assert X.shape == X_ij.shape
+    assert xp.all(X == X_ij)
+
+    assert Y.shape == Y_ij.shape
+    assert xp.all(Y == Y_ij)
+