The small example that I posted used pyximport. The extension module that made me originally see the issue was built with a setup.py distutils script. I have the same issue with both of them. They work on a normal python distribution, but fail if there is not a system- wide install of numpy. The issue happens if the setup.py script uses either the original .pyx files or the generated .c files.
On Feb 2, 9:33 pm, Robert Bradshaw <[email protected]> wrote: > On Feb 2, 2010, at 4:11 PM, M. Yurko wrote: > > > > > On Feb 2, 6:44 pm, "Georg S. Weber" <[email protected]> > > wrote: > >> On 2 Feb., 21:26, Michael Yurko <[email protected]> wrote: > > >>> Hello All, > > >>> I've recently been building an extension module using both cython > >>> and numpy. > >>> It works fine with my system-wide install of python with numpy and > >>> cython as > >>> well as my install of sage. However, when I try to run it from > >>> sage on a > >>> system that doesn't have a system wide install of cython and > >>> numpy, it fails > >>> with: > >>> /home/myurko/.pyxbld/temp.linux-x86_64-2.6/pyrex/numpy_test.c: > >>> 144:31: error: > >>> numpy/arrayobject.h: No such file or > >>> directory > >>> /home/myurko/.pyxbld/temp.linux-x86_64-2.6/pyrex/numpy_test.c: > >>> 145:31: error: > >>> numpy/ufuncobject.h: No such file or directory > > >>> and then pages of issues relating to the failure to include > >>> arrayobject.h > >>> and ufuncobect.h . This issue is simply illustrated by the two > >>> short > >>> files:http://modular.math.washington.edu/home/myurko/misc/npt_wrapper.pyand > >>> .... Before > >>> I open a ticket, I just wanted to make sure that there already > >>> isn't one > >>> open, or there is not some other issue at play. > > >>> Thanks, > >>> Mike Yurko > > >> Try adding the line (yes, it's a *comment*) > > >> #cinclude $SAGE_ROOT/local/lib/python/site-packages/numpy/core/ > >> include > > >> to your *.pyx file. I don't remember right now where this stuff > >> (#cinclude, #clib, ...) for .pyx modules in conjunction with Sage is > >> documented. But it's not a bug, it's a feature ... > > >> Cheers, > >> Georg > > > Hmm. That doesn't seem to work either in my trivial example, or my > > original extension module. I did find mention on #cinclude in the > > documentation (http://www.sagemath.org/doc/developer/ > > coding_in_other.html). > > That's only for using %cython in the notebook. > > > However, I probably should have been more > > specific. For now, I'm just trying to use sage as a python install > > that has both cython and numpy. I've been running the script with > > "sage -python ..." It seems odd to me that the behavior of the > > included python would be modified. > > How did you build your .pyx modules? Did you make a setup.py and run > "sage -python setup.py"? > > - Robert -- To post to this group, send an email to [email protected] To unsubscribe from this group, send an email to [email protected] For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
