[NumPy-Tickets] [NumPy] #1580: shape lost inside record arrays

[NumPy Trac]

#1580: shape lost inside record arrays
------------------------+---------------------------------------------------
 Reporter:  batripler   |       Owner:  somebody
     Type:  defect      |      Status:  new     
 Priority:  high        |   Milestone:          
Component:  numpy.core  |     Version:  1.4.0   
 Keywords:              |  
------------------------+---------------------------------------------------
 The following code replicates the problem:

 {{{
 import numpy
 def foo(dim):
   sdt=numpy.dtype([('v',(float,(1,)))])
   print numpy.empty(dim,sdt)['v'].shape
   print
 numpy.empty(dim,numpy.dtype([('d1',sdt),('d2',sdt)]))['d1']['v'].shape
 def test():
   foo((2,3))
   foo((0,3))
 }}}

 Output:
 {{{
 (2, 3, 1)
 (2, 3, 1)
 (0, 3, 1)
 (1, 0, 3)
 }}}

 The 1st line matches the 2nd line of output, since the array is of non-
 zero size. The same should be true for the 3rd and 4th lines, when the
 array happens to be zero-sized (i.e. zero along at least one dimension).
 Indexing zero-sized record arrays with subarray fields doesn't return
 arrays of the correct shape.

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1580>
NumPy <http://projects.scipy.org/numpy>
My example project