Should the generator of a VirtualArray always yield unique copies of an array?

[pfackeldey]

          Do we assume that the result of each generator invocation is unique (i.e. doesn't need to be copied)? That would be nice, and means that this is correct.

Originally posted by @agoose77 in #3364 (comment)

[ikrommyd]

I had replied to @agoose77's original comment. Each generator creates a new numpy.ndarray that doesn't share memory with anything else no?

def generator():
    return np.array([0.0, 1.1, 2.2, 3.3, 4.4, 5.5, 6.6, 7.7, 8.8, 9.9], dtype=np.float64)

This will return a unique numpy array upon each invocation right?

[pfackeldey]

This does, yes, but not necessarily uproot's reading that uses internal caches.