Does numpy.where always output the indices in an ascending order?

Question

I have a 1D boolean array a, for example

a = numpy.array([0, 1, 0, 0, 1, 1, 1, 0, 1], dtype='bool')

I want to use numpy.where to find the indices of the True elements

idx = numpy.where(a)[0]

Can I assume that the output idx is always sorted in an ascending order? Does numpy.where guarantee that? I am asking this question because the document does not say anything about it and the numpy.where is written in C, which is not obvious to tell how it works.

Answer 1

If absolute maximum speed is not of utmost importance you can simply code out your assumption explicitly:

idx = np.arange(len(a))[a]

That said, np.nonzero(a) is guaranteed to give the indices in order as per the documentation. It's just slightly less explicit than the above, but will be faster.