[Numpy-discussion] histogramdd shapes

[Ben Granett]

Hi,
There seems to be a problem with histogramdd (numpy 1.0.1).  Depending on the
number of bins in each axis, it may not produce an array with the dimensions
oriented correctly.  The bug is in the axis swapping code at the end of the
routine.

To reproduce, you can run,
>>> x = random.randn(100,3)
>>> H, edg = histogramdd(x, bins=(7,5,6))
>>> print H.shape
(5, 7, 6)



A fix is to replace this:
    # Shape into a proper matrix
    hist = hist.reshape(sort(nbin))
    for i,j in enumerate(ni):
        hist = hist.swapaxes(i,j)
        if (hist.shape == nbin).all():
            break


with this:
    for i in arange(nbin.size):
	j = ni[i]
	hist = hist.swapaxes(i,j)
	ni[i],ni[j] = ni[j],ni[i]


That is, elements of ni need to be swapped too.

I hope this wasn't a known bug, but I couldn't find any mention of it.  Thanks,
Ben