[NumPy-Tickets] [NumPy] #1420: irfft2 handles axes argument improperly

[NumPy Trac]

#1420: irfft2 handles axes argument improperly
----------------------------+-----------------------------------------------
 Reporter:  sivaldimarsson  |       Owner:  somebody
     Type:  defect          |      Status:  new     
 Priority:  normal          |   Milestone:          
Component:  numpy.fft       |     Version:          
 Keywords:  irfft2          |  
----------------------------+-----------------------------------------------
 The following output of an interactive session shows how numpy.fft.irfft
 handles axes argument badly. The workaround is to use irfftn with a shape
 argument as shown.

 siv@siv-desktop:~/Documents$ ipython

 Python 2.6.4 (r264:75706, Dec  7 2009, 18:43:55)
 Type "copyright", "credits" or "license" for more information.

 IPython 0.10 -- An enhanced Interactive Python.
 ?         -> Introduction and overview of IPython's features.
 %quickref -> Quick reference.
 help      -> Python's own help system.
 object?   -> Details about 'object'. ?object also works, ?? prints more.

 In [1]: import numpy as np

 In [3]: import numpy.fft as npfft

 In [4]: a = np.zeros((256,256))

 In [5]: a.shape

 Out[5]: (256, 256)

 In [9]: a[0:10,0:100] = 1

 In [11]: ahat = npfft.rfft2(a,axes=(1,0))

 In [12]: ahat.shape

 Out[12]: (129, 256)

 In [14]: ahatinvhat = npfft.irfft2(ahat,axes=(1,0))

 In [15]: ahatinvhat.shape

 Out[15]: (129, 510)

 In [16]: ahatinvhat - a

 ValueError                   Traceback (most recent call last)

 /home/siv/Documents/<ipython console> in <module>()

 ValueError: shape mismatch: objects cannot be broadcast to a single shape

 In [22]: chat = npfft.rfftn(a,s=a.shape,axes=(1,0))

 In [23]: chat.shape

 Out[23]: (129, 256)

 In [24]: np.max(np.abs(ahat-chat))

 Out[24]: 0.0

 In [25]: chatinvhat = npfft.irfftn(chat,s=a.shape,axes=(1,0))

 In [26]: chatinvhat.shape

 Out[26]: (256, 256)

 In [27]: np.max(np.abs(a-chatinvhat))

 Out[27]: 6.6613381477509392e-16

-- 
Ticket URL: <http://projects.scipy.org/numpy/ticket/1420>
NumPy <http://projects.scipy.org/numpy>
My example project