Mailman 3 December 2006 - SciPy-User

feed-forward neural network for python
by Marek Wojciechowski 07 Dec '06

07 Dec '06

Hi! I released feed-forward neural network for python project at sourceforge (ffnet). I'm announcing it here because it depends on numpy/scipy tandem, so you, folks, are potential users/testers :) If anyone is interested please visit ffnet.sourceforge.net (and then post comments if any...) Greetings -- Marek Wojciechowski

4 3

Re: [SciPy-user] blas/lapack issue on a freebsd box...
by fred 07 Dec '06

07 Dec '06

> > This is part of your problem. Do you have a site.cfg file? If so, please > post it. > > Ok. I have this one, in /usr/local/lib/python2.4/site-packages/numpy/distutils: [atlas] library_dirs = /usr/lib/atlas/3dnow/ atlas_libs = lapack, blas Looks wrong, no ? > You ought to have sections that look like these: > > > [DEFAULT] > library_dirs=/usr/local/lib > include_dirs=/usr/local/include > > [blas_opt] > libraries=f77blas, cblas, atlas > > [lapack_opt] > libraries=lapack, f77blas, cblas, atlas > > I have deleted the lines above, and put yours. But nothing changes :-( > Possibly, you might also need to add g2c to the end of the libraries= lines. > > Hmm, I have no g2c on my freebsd box. > Now, those libraries lists are from what I normally see installed from > ATLAS. > Given that I've never seen libalapack_r.so.1 before, it's possible that > ATLAS > gets build strangely on your machine. Please also give us a list of the > libraries that ATLAS installs. Also tell us where you got > /usr/local/lib/liblapack.so and /usr/local/lib/libblas.so . > Information for atlas-devel-3.7.11: Files: /usr/local/include/cblas.h /usr/local/include/clapack.h /usr/local/include/blas.h /usr/local/include/lapack.h /usr/local/lib/libalapack.a /usr/local/lib/libalapack.so.1 /usr/local/lib/libalapack.so /usr/local/lib/libalapack_r.a /usr/local/lib/libalapack_r.so.1 /usr/local/lib/libalapack_r.so /usr/local/lib/libatlas.a /usr/local/lib/libatlas.so.1 /usr/local/lib/libatlas.so /usr/local/lib/libatlas_r.a /usr/local/lib/libatlas_r.so.1 /usr/local/lib/libatlas_r.so /usr/local/lib/libcblas.a /usr/local/lib/libcblas.so.1 /usr/local/lib/libcblas.so /usr/local/lib/libcblas_r.a /usr/local/lib/libcblas_r.so.1 /usr/local/lib/libcblas_r.so /usr/local/lib/libf77blas.a /usr/local/lib/libf77blas.so.1 /usr/local/lib/libf77blas.so /usr/local/lib/libf77blas_r.a /usr/local/lib/libf77blas_r.so.1 /usr/local/lib/libf77blas_r.so /usr/local/lib/libptcblas.a /usr/local/lib/libptcblas.so.1 /usr/local/lib/libptcblas.so /usr/local/lib/libptf77blas.a /usr/local/lib/libptf77blas.so.1 /usr/local/lib/libptf77blas.so /usr/local/lib/libtstatlas.a /usr/local/lib/libtstatlas.so.1 /usr/local/lib/libtstatlas.so /usr/local/lib/libtstatlas_r.a /usr/local/lib/libtstatlas_r.so.1 /usr/local/lib/libtstatlas_r.so /usr/local/share/doc/atlas/AtlasCredits.txt /usr/local/share/doc/atlas/ChangeLog /usr/local/share/doc/atlas/DirStruct.txt /usr/local/share/doc/atlas/INDEX.txt /usr/local/share/doc/atlas/LibReadme.txt /usr/local/share/doc/atlas/TestTime.txt /usr/local/share/doc/atlas/TroubleShoot.txt /usr/local/share/doc/atlas/Windows.txt /usr/local/share/doc/atlas/atlas_contrib.ps /usr/local/share/doc/atlas/atlas_devel.ps /usr/local/share/doc/atlas/atlas_over.ps /usr/local/share/doc/atlas/cblas.ps /usr/local/share/doc/atlas/cblasqref.ps /usr/local/share/doc/atlas/f77blasqref.ps /usr/local/share/doc/atlas/lapackqref.ps I got liblapack.so from lapack-3.0_1 and libblas.so from blas-1.0 binary packages. Thanks in advance. PS : building scipy-0.5.1 requires that atlas be built with static libs. PS2 : thanks to added me to the EditorsList ;-) -- http://scipy.org/FredericPetit

2 3

blas/lapack issue on a freebsd box...
by fred 06 Dec '06

06 Dec '06

Hi all, I'm trying to build scipy and all its stuff on my freebsd box (r6.2). (binaries pkg are not yet provided on freebsd ftp servers). I have successfully built scipy 0.5.1, but in fact, it fails when I want to import some module: IPython 0.7.2 -- An enhanced Interactive Python. ? -> Introduction to IPython's features. %magic -> Information about IPython's 'magic' % functions. help -> Python's own help system. object? -> Details about 'object'. ?object also works, ?? prints more. In [1]: from scipy.special import jv --------------------------------------------------------------------------- exceptions.ImportError Traceback (most recent call last) /usr/home/fred/<ipython console> /usr/local/lib/python2.4/site-packages/scipy/special/__init__.py 8 from basic import * 9 import specfun ---> 10 import orthogonal 11 from orthogonal import legendre, chebyt, chebyu, chebyc, chebys, \ 12 jacobi, laguerre, genlaguerre, hermite, hermitenorm, gegenbauer, \ /usr/local/lib/python2.4/site-packages/scipy/special/orthogonal.py 65 import _cephes as cephes 66 _gam = cephes.gamma ---> 67 from scipy.linalg import eig 68 69 def poch(z,m): /usr/local/lib/python2.4/site-packages/scipy/linalg/__init__.py 6 from linalg_version import linalg_version as __version__ 7 ----> 8 from basic import * 9 from decomp import * 10 from matfuncs import * /usr/local/lib/python2.4/site-packages/scipy/linalg/basic.py 15 #from blas import get_blas_funcs 16 from flinalg import get_flinalg_funcs ---> 17 from lapack import get_lapack_funcs 18 from numpy import asarray,zeros,sum,newaxis,greater_equal,subtract,arange,\ 19 conjugate,ravel,r_,mgrid,take,ones,dot,transpose,sqrt,add,real /usr/local/lib/python2.4/site-packages/scipy/linalg/lapack.py 15 import numpy 16 ---> 17 from scipy.linalg import flapack 18 from scipy.linalg import clapack 19 _use_force_clapack = 1 ImportError: /usr/local/lib/libalapack_r.so.1: Undefined symbol "cblas_dswap" As google gives no result about this issue, any suggestion are welcome. Cheers, PS : python 2.4.3, numpy 1.0r1, scipy 0.5.1, atlas 3.6.0.1, blas 1.0 -- http://scipy.org/FredericPetit

5 10

Re: [SciPy-user] Scipy Superpack (was Re: pajer@iname.com has sent you a, file.)
by R. Padraic Springuel 06 Dec '06

06 Dec '06

Well, I got scipy and matplotlib working now. For scipy, installing gfortran worked just as advertised. For maplotlib, I decided to download it directly from their sourceforge site and install it along with wxPython for back end purposes (it's the same backend I use on Windows so I like the consistency). It was easier for me than trying to figure out if more than one matplotlib file was missing. Is there anyway that the Superpack could include a Readme.txt file with installation instructions? I know that I'll sometimes download stuff to install later when I'm offline and I'm sure that others do too. Having the installation instructions in a txt file in the download would help in those situations. Also, the installation instructions on the Scipy website aren't clear about when they apply to the Superpack installation and when they apply to a build from scratch installation. Some clarification here would be useful. I'm willing to help write the installation instructions for the Superpack (to the best of my ability), but don't know where they might be submitted (direction?). Also, since I use wxPython for my matplotlib backend, if there are any specific instructions for PyMC (which I think is another backend) that would make the installation go smoother, I wouldn't be able to provide those. -- R. Padraic Springuel Teaching Assistant Department of Physics and Astronomy University of Maine Bennett 309 Office Hours: Wednesday 2-3pm

1 0

Re: [SciPy-user] How to: Pass NumPy arrays to C & use SWIG to generate modules
by vallis.35530172＠bloglines.com 05 Dec '06

05 Dec '06

If all you want to do is pass an array of doubles (with its length) to C, you can use the attached typemap and swig file. I think they represent the maximum possible distillation of the process. Read the short comments in example.i. (Uh, I'm afraid the posting might break the newlines... I will also send you the files in attachment). Let me know if you need help with compiling. Cheers, Michele ==== the following is my example.i (the SWIG file) ==== %module example %{ #include "example.h" %} // suppose your example.h contains the function declaration // void examplefunction(double *array,long arraylength); // to which you want to pass a numpy 1-D array; // what we do here is to apply the numpy-to-array typemap // defined in numpy_typemaps.i to your function, which is then // included by the %include line below %include numpy_typemaps.i %typemap(in) (double *array,long arraylength) = (double *numpyarray,long numpyarraysize); %include "example.h" ==== the following is file numpy_typemaps.i ==== #include "numpy/arrayobject.h" #define ISCONTIGUOUS(m) (PyArray_ISCONTIGUOUS(m)) #define PyArray_CONTIGUOUS(m) (ISCONTIGUOUS(m) ? Py_INCREF(m), m : \ (PyArrayObject *)(PyArray_ContiguousFromObject((PyObject *)(m), (m)->descr->type_num, 0,0))) %} %init %{ import_array(); %} /* %typemap(in) (double* numarray, long length) { */ %typemap(in) (double *numpyarray,long numpyarraysize) { PyArrayObject *arr; if (!PyArray_Check($input)) { PyErr_SetString(PyExc_TypeError,"First argument is not an array"); return NULL; } if (PyArray_ObjectType($input,0) != PyArray_DOUBLE) { PyErr_SetString(PyExc_TypeError, "Incorrect array type: we need an array of DOUBLE"); return NULL; } arr = PyArray_CONTIGUOUS((PyArrayObject *)$input); if (arr->nd == 1) { $1 = (double *)arr->data; $2 = (long)arr->dimensions[0]; } else if (arr->nd == 2) { $1 = (double *)arr->data; $2 = (long)arr->dimensions[0] * (long)arr->dimensions[1]; } else { PyErr_SetString(PyExc_TypeError, "Incorrect number of dims: we want a 1D or 2D array"); return NULL; } Py_DECREF(arr); /* Release our local copy of the PyArray */ } ==== --- SciPy Users List <scipy-user(a)scipy.org wrote: Thanks. I saw that site in a previous google. But it > seems to be aimed at Windows. I am on a Mac. Rather > than fight that battle now, I will try today to just > generate my own simple C extension test routine. If I > get that far, then that will probably be the way for > me to go. I haven't checked out pyrex, though. I > might look at that when I take a break. > > If I get the C extension working, I will post an > announcement about it. It might help non-gurus like > me get started. > > -- Lou Pecora > Naval Research Lab, Washington, DC > > --- Gael Varoquaux <gael.varoquaux(a)normalesup.org> > wrote: > > > Lou, > > > > I am not a very advanced python user. I do not > > master writing C > > extentions. I have tried simple things, and had > > success with them because > > they were simple enough for me. In order of > > difficulty this is what I > > did: > > > > - scipy.weave (great, but won't probably suit your > > purposes, have a look > > at the wiki page) > > > > - pyrex (same reamrk than above, though the concept > > is fabulous) > > > > - passing array to C code with ctypes. I think this > > can suit you. It > > might not be as versatile or powerful as SWIG, but > > I found it was > > simple enough for my limited knowledge. I started > > out with simple hello > > word C functions, and was able to pass numpy > > arrays back and forth > > between numpy and C thanks to Albert Strasheim's > > wiki page > > http://scipy.org/Cookbook/Ctypes2 > > > > I cannot help you out with SWIG, as I have decided > > it was to complicated > > for me so far. > > > > Ga�l > > > > > ____________________________________________________________________________________ > Want to start your own business? > Learn how on Yahoo! Small Business. > http://smallbusiness.yahoo.com/r-index > _______________________________________________ > SciPy-user mailing list > SciPy-user(a)scipy.org > http://projects.scipy.org/mailman/listinfo/scipy-user >

1 0

How to: Pass NumPy arrays to C & use SWIG to generate modules
by Lou Pecora 05 Dec '06

05 Dec '06

I have some number crunching routines I want to code in C or C++ and then use SWIG to create a module I can import. I can get the simple example.i stuff in the SWIG package to work, but I need to pass NumPy arrays back and forth. I have looked up typemaps and googled examples, but whatever I find is pretty general and high level for me and my needs -- at least until I get more experience. Does anyone know of simple examples of passing NumPy arrays and using SWIG? I really don't want to branch out to other objects, yet. No C++ templates or fancy stuff. Just NumPy arrays to C arrays (or something reasonable in C) will get me a long way in my coding for now. Thanks for any suggestions or pointers. -- Lou Pecora US Naval Research Lab ____________________________________________________________________________________ Do you Yahoo!? Everyone is raving about the all-new Yahoo! Mail beta. http://new.mail.yahoo.com

5 10

geostatistics
by Joshua Petterson 05 Dec '06

05 Dec '06

Hi all, I can't find any package for geostatistics in scipy (kriging, variograms, ...), do I have to use R or is there another place for that? Best regards

4 3

optimize.leastsq() question
by palazzol＠comcast.net 05 Dec '06

05 Dec '06

Hi, First, I want to thank you all for making SciPy so great. I am prototyping an application using VPython and SciPy together, and I can't believe how well everything is working after <2 weeks of coding. Now, on to my question. I've boiled it down to a small test program which shows the situation, and attached it here. I am passing an array into my minimization function, intended to hold the return values. Apparently, this causes some kind of issue, when the array is updated elementwise. If you run it as-is, you will see the iteration stop after a few tries, with an error. If you uncomment either of the #err=xxx lines, it works fine. I am running under WinXP, Python2.4, a slightly older version of SciPy 0.4.9, with numpy-0.9.8. I'm hoping to upgrade soon, but haven't yet. Thanks, -Frank

3 2

General parallel processing question
by Anand Patil 05 Dec '06

05 Dec '06

Hi again everyone, Say I have two objects A and B. A's member functions _occasionally_ want to call the member functions of B, and vice versa. Both A and B have to do a lot of work between calls to each other's member functions. I'd like to push B off to a new process, but be able to program as if it were still in the same process as A. That is, I'd like to be able to call B's member functions from A without having to teach A how to do interprocess communication. The solution I've been thinking of is to write 'avatar' objects that represent objects living in different processes and know how to pass member function calls along to their 'true selves' (probably using mpi4py and/or IPython1, but I haven't worked out the details yet). If A lives in process 0 and B gets pushed to process 1, I would create an avatar of B in process 0 and one of A in process 1. That scheme would get kind of klunky, though. Has anyone thought about/ dealt with this situation before? Thanks, Anand

4 5

Editor of the wiki
by Gael Varoquaux 04 Dec '06

04 Dec '06

Hi, Can an editor add me please to the editor page, I would like to rename a page. Cheers, Gaël

4 3