BUG: Calling ufuncs with a positional output argument causes the result to inherit the output's mask

This brings `np.add(a, b, out)` in line with  `np.add(a, b, out=out)`.
These previously differed because gh-10459 causes them to call __array_wrap__ in different ways (with and without the output argument in the context tuple, respectively).

Since the data in the `out` argument is never used by ufuncs, it seems consistent that the mask should not be either.

1 files changed, 7 insertions, 5 deletions

diff --git a/numpy/ma/core.py b/numpy/ma/core.py
index 719961d1b..56c761da1 100644
--- a/numpy/ma/core.py
+++ b/numpy/ma/core.py
@@ -3029,18 +3029,20 @@ class MaskedArray(ndarray):
 
         if context is not None:
             result._mask = result._mask.copy()
-            (func, args, _) = context
-            m = reduce(mask_or, [getmaskarray(arg) for arg in args])
+            func, args, out_i = context
+            # args sometimes contains outputs (gh-10459), which we don't want
+            input_args = args[:func.nin]
+            m = reduce(mask_or, [getmaskarray(arg) for arg in input_args])
             # Get the domain mask
             domain = ufunc_domain.get(func, None)
             if domain is not None:
                 # Take the domain, and make sure it's a ndarray
-                if len(args) > 2:
+                if len(input_args) > 2:
                     with np.errstate(divide='ignore', invalid='ignore'):
-                        d = filled(reduce(domain, args), True)
+                        d = filled(reduce(domain, input_args), True)
                 else:
                     with np.errstate(divide='ignore', invalid='ignore'):
-                        d = filled(domain(*args), True)
+                        d = filled(domain(*input_args), True)
 
                 if d.any():
                     # Fill the result where the domain is wrong