Consider adding Put symbol or similar.

[arcondello]

Something like numpy.put().

Users have expressed a desire to do operations like

from dwave.optimization import Model

model = Model()

array = model.constant(np.zeros(5, 5))
x = model.integer(5)

array[0, :] += x  # this raises an error

Unfortunately, due to the way that __setitem__ works in Python, we cannot give that exact syntax. But we can get pretty close. So the vision would be something like

from dwave.optimization import put

indices = model.constant([0, 1, 2, 3, 4])
array = put(array, indices, x)

[wbernoudy]

The reason we can't use __setitem__ is that we need to create a new symbol/object, but there's no machinery for returning that with __setitem__, right?

[arcondello]

Yes, __setitem__() always acts in-place on the object. Unlike say __iadd__() which can return a new object. It's possible we could do some fiddling with globals() or whatever, but that way lies madness and bugs.

[wbernoudy]

It would also obscure the fact that you are making a new symbol every assignment which can clearly get expensive.

I'm slightly worried that a Put symbol would encourage use-cases that are better served by concatenation (so fewer intermediate arrays of "full-size"). But at least needing to call put and getting a new symbol back makes it clear that it's a full copy of the array.