scipy.weave does not work well with big-endian numpy arrays (Trac #1153)

[scipy-gitbot]

Original ticket http://projects.scipy.org/scipy/ticket/1153 on 2010-04-13 by trac user trobitaille, assigned to unknown.

The following code:

import numpy as np
from scipy.weave import converters
import scipy.weave as weave

nx = 10
x = np.logspace(-5, 4, nx).astype('>f4')
y = np.logspace(-5, 4, nx).astype('<f4')

code = """
       for (int i=0; i<nx-1; ++i) {
          printf("%11.3e %11.3e \\n",x(i), y(i));
       }
       return_val = 0;
       """

weave.inline(code, ['x', 'y', 'nx'], type_converters=converters.blitz)

produces the following output:

 -5.603e-12   1.000e-05 
  1.188e-24   1.000e-04 
  4.534e+28   1.000e-03 
  2.072e-32   1.000e-02 
 -4.295e+08   1.000e-01 
  4.601e-41   1.000e+00 
  1.157e-41   1.000e+01 
  7.184e-41   1.000e+02 
  4.386e-41   1.000e+03 

This suggests that big-endian numpy arrays are not correctly converted when passed to the C code in weave.

[scipy-gitbot]

Milestone changed to Unscheduled by @rgommers on 2011-06-13

[scipy-gitbot]

@sturlamolden wrote on 2012-07-30

Endianness is not converted at all, which is correct or incorrect depending on your point of view. I think the ticket is invalid.

[andreas-h]

if valid, this ticket should be transferred to the weave project. is this possible, or do we have to open a new ticket in weave project and reference this one from there?

[rgommers]

We should transfer all of the weave tickets via the Github API. This can probably re-use code that was written for the Trac --> Github move. It's not very high on my todo list, so if you want to give it a shot please do.

[person142]

Closing since weave is no longer a part of SciPy. This issue is now tracked at scipy/weave#7.