Using nditer, is it possible to manually handle dimensions  with different lengths? <div><br><div>For example, lets say I had an array A[5, 100] and I wanted to sample every 10 along the second axis so I would end up with an array B[5,10]. Is it possible to do this with nditer, handling the iteration over the second axis manually of course (probably in cython)?</div>

</div><div><br></div><div>I want something like this (modified from <a href="http://docs.scipy.org/doc/numpy/reference/arrays.nditer.html#putting-the-inner-loop-in-cython" target="_blank">http://docs.scipy.org/doc/numpy/reference/arrays.nditer.html#putting-the-inner-loop-in-cython</a>)</div>
<div><br></div><div><div>@cython.boundscheck(False)</div><div>def sum_squares_cy(arr):</div><div>    cdef np.ndarray[double] x</div><div>    cdef np.ndarray[double] y</div><div>    cdef int size</div>
<div>    cdef double value</div><div>    cdef int j</div><div><br></div><div>    axeslist = list(arr.shape)</div><div>    axeslist[1] = -1</div><div><br></div><div>    out = zeros((arr.shape[0], 10))</div><div>    it = np.nditer([arr, out], flags=[&#39;reduce_ok&#39;, &#39;external_loop&#39;,</div>
<div>                                      &#39;buffered&#39;, &#39;delay_bufalloc&#39;],</div>
<div>                op_flags=[[&#39;readonly&#39;], [&#39;readwrite&#39;, &#39;no_broadcast&#39;]],</div><div>                op_axes=[None, axeslist],</div><div>                op_dtypes=[&#39;float64&#39;, &#39;float64&#39;])</div>

<div>    it.operands[1][...] = 0</div><div>    it.reset()</div><div>    for xarr, yarr in it:</div><div>        x = xarr</div><div>        y = yarr</div><div>        size = x.shape[0]</div><div>        j = 0</div><div>        for i in range(size):</div>

<div>           #some magic here involving indexing into x[i] and y[j]</div><div>    return it.operands[1]</div></div><div><br></div><div>Does this make sense? Is it possible to do?</div>