Retrieving results of non-trivial storages

[beojan]

It doesn't seem to be possible to retrieve the results as a NumPy array when a non-trivial storage is used (e.g. storage.weight). It's known that view isn't supported, but np.asarray also crashes the Jupyter kernel.

[henryiii]

It is not currently possible - a decision has/had to be made about that the storages produce as a numpy array. I think we do have a plan now, but it has not been implemented yet. There is an incoming feature to Boost.Histogram C++ (I think today) that will help us here.

h.view() and np.asarray(h) use the same underlying machinery.

As a workaround, you can still iterate over the histogram cells and access them one-by-one as Python objects.

[beojan]

a decision has/had to be made about that the storages produce as a numpy array

I would expect to be able to access each stored quantity as an array (i.e. .values() and .variances() as for individual cells in the C++ API).

[henryiii]

But then the returned item is no longer a numpy array (killing asarray). We could possibly introduce a difference between asarray and .view() here, though.