[BUG/FEAT] hist division should convert to floats

[andrzejnovak]

a = hist.Hist.new.Reg(10,-2,2).Int64().fill(np.random.uniform(-2,2,size=1000))
b = hist.Hist.new.Reg(10,-2,2).Int64().fill(np.random.normal(0,1,size=1000))

print(a.values()/b.values())
print((a/b).values())
>>> [2.51351351 1.46153846 0.92380952 0.82857143 0.59119497 0.70666667
 0.85714286 0.89534884 1.625      3.8125    ]
>>> [2 1 0 0 0 0 0 0 1 3]

@henryiii

[henryiii]

hists currently do not change storages under operations. @HDembinski should they? I could probably do that if they should.

[HDembinski]

In Python, I think the conversion should happen, as this is the least surpring thing. The unlimited storage also converts in this case.

[nsmith-]

In addition it would be good if scaling a histogram by a scalar value automatically converts scalar storage (Int64, Double etc.) to Weight storage so that the variance is correctly propagated. In the case here, where two histograms are divided, it is a bit ambiguous what the appropriate variance would be so I am not sure I can recommend conversion to Weight storage.