BUG: Make ndarray inplace operators forward calls when needed. #9021
Conversation
|
Needs documentation changes too - this broken behaviour was explicitly documented |
|
So I think the conclusion of that discussion was specifically that if the RHS object sets If I'm reading right, I think that the |
|
@eric-wieser Yeah, documentation still needs fixes. No new tests are needed after the small fix in this PR. Probably should add a note there. |
|
@njsmith Sounds like we need an |
|
Thinking about it, we should probably raise a deprecation warning. Sparse (and others), really should not expect it to work with ndarrays because Python does the wrong thing there. |
2357509 to
c612792
Compare
|
OK, fixed up so that in-place ops forward on I'm going to wait on documentation until #9014 goes in so as to avoid conficts. |
numpy/core/src/multiarray/number.c
Outdated
| return PyArray_GenericInplaceBinaryFunction(m1, m2, n_ops.add); | ||
| } | ||
|
|
||
| static PyObject * | ||
| array_inplace_subtract(PyArrayObject *m1, PyObject *m2) | ||
| { | ||
| INPLACE_GIVE_UP_IF_NEEDED( | ||
| m1, m2, nb_inplace_subtract,array_inplace_subtract); |
For backward compatibility, we need to let `__array_priority__` determine the override for inplace ops. If `__array_ufunc__ = None` on the right hand side it is still the case that TypeError will be raised. This fixes SciPy test failures of the type ndarray += sparse_array Closes numpy#9019.
c612792 to
105557a
Compare
|
@mhvk Thanks for the quick review. |
For backward compatibility, we need to let
__array_priority__determine the override for inplace ops. If
__array_ufunc__ = Noneon the right hand side it is still the case that TypeError will be
raised.
This fixes SciPy test failures of the type
Closes #9019.