Mailman 3 January 2006 - SciPy-User

numpy.asarray and Numeric.asarray
by Steve Schmerler Jan. 24, 2006

Jan. 24, 2006

Hi Using optimize.fmin_powell from a recent svn build with the initial guess x0 being a scalar I get this error: Traceback (most recent call last): File "trans.py", line 151, in ? tau = fmin_powell(func, tau_start) File "/usr/lib/python2.3/site-packages/scipy/optimize/optimize.py", line 1453, in fmin_powell N = len(x) TypeError: len() of unsized object where x = asarray(x0). Comparing numpy's and Numeric's asarray I found =================================================================================== In [8]: import numpy as nu In [9]: import Numeric as NU In [10]: x = nu.asarray([1,2]); len(x) Out[10]: 2 In [11]: y = NU.asarray([1,2]); len(y) Out[11]: 2 In [12]: x = nu.asarray(3); len(x) --------------------------------------------------------------------------- exceptions.TypeError Traceback (most recent call last) /home/elcorto/<console> TypeError: len() of unsized object In [13]: y = NU.asarray(3); len(y) Out[13]: 1 =================================================================================== cheers, steve -- "People like Blood Sausage too. People are Morons!" -- Phil Connors, Groundhog Day

1 1

scipy.special.erf randomly returns NANs on OSX
by Evan Monroig Jan. 24, 2006

Jan. 24, 2006

Hi, I have this strange problem that the special erf function returns NANs when the input is an array: >>> import scipy,numpy >>> x = numpy.arange(0., 2., 0.1) >>> x array([ 0. , 0.1, 0.2, 0.3, 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, 1. , 1.1, 1.2, 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9]) >>> scipy.special.erf(x) array([ 0. , 0.11246292, 0.22270259, 0.32862676, 0.42839236, 0.52049988, 0.60385609, 0.67780119, 0.74210096, 0.79690821, nan, 0.88020507, 0.91031398, 0.93400794, 0.95228512, 0.96610515, 0.97634838, 0.98379046, 0.9890905 , 0.99279043]) >>> x = numpy.arange(1., 2., 0.1) >>> scipy.special.erf(x) array([ 0.84270079, 0.88020507, 0.91031398, 0.93400794, 0.95228512, 0.96610515, 0.97634838, 0.98379046, 0.9890905 , 0.99279043]) scipy is 0.4.4.1526, numpy is 0.9.3.1836, I am running on Mac OS X with MacPython 2.4.1 Can anyone reproduce the problem? Evan

5 9

IPython 0.7.1 is out.
by Fernando Perez Jan. 24, 2006

Jan. 24, 2006

[ As usual, sorry for the cross-post. Given that many of you use ipython and aren't on the lists, I hope you'll find this useful. But let me know if this is considered spam, and I'll stop ] Hi all, I have released IPython 0.7.1, which is mainly a bugfix release over 0.7.0. As expected in that release, given the large changes made, some problems inevitably appeared. I believe all regressions and known bugs have been fixed, along with some useful new features. This release marks the end of my tenure on the stable branch of ipython. Ville Vainio will continue as maintainer of the ipython trunk/, while I will continue working on the development branch of IPython (chainsaw). I will remain available on the list as usual, both to assist Ville as needed, and to discuss withe anyone interested. But I'll try to limit my time and effort spent on trunk to a minimum, so we can really advance the (fairly ambitious) chainsaw project. Where to get it --------------- IPython's homepage is at: http://ipython.scipy.org and downloads are at: http://ipython.scipy.org/dist I've provided: - source downloads (.tar.gz) - RPMs (for Python 2.3 and 2.4, built under Fedora Core 3). - Python Eggs (http://peak.telecommunity.com/DevCenter/PythonEggs). - a native win32 installer for both Python 2.3 and 2.4. Fedora users should note that IPython is now officially part of the Extras repository, so they can get the update from there as well (though it may lag by a few days). Debian, Fink and BSD packages for this version should be coming soon, as the respective maintainers have the time to follow their packaging procedures. Many thanks to Jack Moffit, Norbert Tretkowski, Andrea Riciputi and Dryice Liu). The eggs included now are lighter, as they don't include documentation and other ancillary data. If you want a full ipython installation, use the source tarball or your distribution's favorite system. Many thanks to Enthought for their continued hosting support for IPython. Release notes ------------- As always, the full ChangeLog is at http://ipython.scipy.org/ChangeLog. The highlights of this release follow. - *FIXED: this wasn't really working right in 0.7.0* Support for input with empty lines. If you have auto-indent on, this means that you need to either hit enter _twice_, or add/remove a space to your last blank line, to indicate you're done entering input. These changes also allow us to provide copy/paste of code with blank lines. - *FIXED: this wasn't really working right in 0.7.0* Support for pasting multiline input even with autoindent on. The code will look wrong on screen, but it will be stored and executed correctly internally. Note that if you have blank lines in your code, these need to be indented like their surroundings (pure empty lines will make ipython assume you are finished with your input). This is a limitation also of the plain python interpreter, and can't really be fixed in a line-oriented system like ipython. - Fixed bug where macros were not working correctly in threaded shells. - Catch exceptions which can be triggered asynchronously by signal handlers. This fixes a rare and obscure problem, but which could crash ipython; reported by Colin Kingsley <tercel-AT-gentoo.org>. - Added new '-r' option to %hist, to see the raw input history (without conversions like %ls -> ipmagic("ls")). - Fix problems with GTK under win32 (excessive keyboard lag and cpu usage). - Added ipython.py script to root directory of download. This allows you to unpack ipython and execute it in-place, without needing to install it at all. - Improved handling of backslashes (\) in magics. This improves the usability of ipython especially under win32, which uses \ as a path separator (though it also benefits Unix for escaping characters in shell usage). - More autocall fixes (rare, but critical). - New IPython.ipapi module to begin exposing 'officially' IPython's public API. This should ease the task of those building systems on top of ipython. - New IPython.platutils module, exposing various platform-dependent utilities (such as terminal title control). - Implemented exception-based 'chain of command' for IPython's hooks. - Added Jason Orendorff's "path" module to IPython tree, http://www.jorendorff.com/articles/python/path/. You can get path objects conveniently through %sc, and !!, e.g.: sc files=ls for p in files.paths: # or files.p print p,p.mtime - Fix '%run -d' with Python 2.4 (pdb changed in Python 2.4). - Various other small fixes and enhancements. Enjoy, and as usual please report any problems. Regards, and thanks for all this time using IPython! Fernando.

1 0

use of blas routines in scipy 0.4.4
by Darren Dale Jan. 23, 2006

Jan. 23, 2006

I'm having a bit of trouble using a blas function in scipy-0.4.4. I've tested the following script with version 0.3.2 and Numeric, which works, and 0.4.4 and numpy which gives me an error. # For scipy-0.4.4 with numpy-0.9.4.1971 from scipy.lib.blas import get_blas_funcs from numpy import array # For scipy-0.3.2, with Numeric-24.2: ##from Numeric import array ##from scipy.linalg.blas import get_blas_funcs a = array([[1,1,1]]) b=array([[1],[1],[1]]) dot, = get_blas_funcs(['gemm',], (a,b), debug=1) res = dot(1, a, b) print res Here's the error: dgemm:n=1 Traceback (most recent call last): File "blasdot.py", line 12, in ? res = dot(1, a, b) fblas.error: (trans_b?kb==k:ldb==k) failed for hidden n Is there a bug, or am I not aware of some important change? Thanks, Darren

2 3

problems numpy 0.9.4 and matplotlib
by Aurimas Anskaitis Jan. 23, 2006

Jan. 23, 2006

I have some problems with numpy-0.9.4 and matplotlib 0.86.1. When I try to launch very simple script like this: import pylab from scipy import * y = array([1, 2, 3]) pylab.plot(y) pylab.show() I get a long error message ending with line "ValueError: arrays must have same number of dimensions". May somebody explain me what's happening? Everything works just fine using numpy 0.9.2. --------------------------------- Yahoo! Photos Ring in the New Year with Photo Calendars. Add photos, events, holidays, whatever.

4 4

Compiling djbfft
by Nils Wagner Jan. 23, 2006

Jan. 23, 2006

Hi all, I saw the comments wrt djbfft. http://pong.tamu.edu/tiki/tiki-view_blog_post.php?blogId=6&postId=97 The compilation problem can be fixed by adding a line #include <errno.h> to error.h. The default installation directory is /usr/local/djbfft. However djbfft_info: NOT AVAILABLE Pearu, How should I modify system_info.py in ~/numpy/numpy/distutils such that the library djbfft.a in /usr/local/djbfft/lib/ will be found. Nils

1 0

sparse matrices and umfpack
by Barry Drake Jan. 23, 2006

Jan. 23, 2006

I'm trying to use UMFSparse, but there is no module umfpack which should be imported at the beginning of UMFSparse.py: """UMFSparse: A minimal sparse-matrix class for Python, useful for interfacing with the UMFPACK library for solving large linear systems, This file defines a class: UMFmatrix that defines a sparse-matrix class. It is designed so that the attribute data points to the non-zero values in the matrix and index is a list of arrays which detail how to place the non-zero values in the array. Currently only a compressed sparse row storage is implemented. """ import scipy_base import scipy.io as io import types import struct import umfpack ... Where is umfpack? I'm using the latest version of Enthought's Python distro on a Win XP Pro system. I've searched my hard drive, the message archives, googled, and gmaned. So far nothing that helps. Anyone know where the umfpack module is? Thanks. Barry Drake

2 1

FFT Problem
by Frederick Ross Jan. 23, 2006

Jan. 23, 2006

I'm running the SVN builds from http://homepage.mac.com/fonnesbeck/mac/ on MacOS 10.4.4 on a 2005 PowerBook G4, and have the following peculiar problem: >>> import numpy >>> a = numpy.array([1, 3, 5, 4]) >>> print numpy.ifft(numpy.fft(a)) [ 1.+0.j 3.+0.j 5.+0.j 4.+0.j] >>> import scipy Overwriting fft=<function fft at 0x783c70> from scipy.fftpack.basic (was <function fft at 0x6ae6b0> from numpy.dft.fftpack) Overwriting ifft=<function ifft at 0x783cb0> from scipy.fftpack.basic (was <function inverse_fft at 0x6ae6f0> from numpy.dft.fftpack) >>> print scipy.ifft(scipy.fft(a)) [ 1. +0.j 3.5-0.j 5. +0.j 3.5+0.j] Any suggestions on what might be causing this would be much appreciated. -- Frederick Ross Graduate Fellow The Rockefeller University

3 3

vector subtraction error
by Ryan Krauss Jan. 22, 2006

Jan. 22, 2006

I am still using old scipy, so maybe this is no longer an issue in the new NumPy, but I seem to do this to myself a fair ammount. I think I have to 1-d vectors and I need to subtract them, but some how there shapes are (n,) and (n,1) and when I subtract them I get something that is shape (n,n): (Pdb) shape(cb.dBmag()) Out[3]: (4250,) (Pdb) shape(curb.dBmag()) Out[3]: (4250, 1) (Pdb) temp=cb.dBmag()-curb.dBmag() (Pdb) shape(temp) Out[3]: (4250, 4250) Would there be a terrible performance cost to check for this when array subtraction is called? Would this be different in the new NumPy? Thanks, Ryan

6 8

Installing SciPy on Linux (with enhancements)
by Robert Hetland Jan. 22, 2006

Jan. 22, 2006

In case anyone is interested, here is a link to unofficial directions for installing SciPy on Linux (slightly more detailed than on the SciPy Install page, but more specific for Linux), written by Steve Baum (http://stommel.tamu.edu/~baum): http://pong.tamu.edu/tiki/tiki-view_blog_post.php?blogId=6&postId=97 -Rob. ----- Rob Hetland, Assistant Professor Dept of Oceanography, Texas A&M University p: 979-458-0096, f: 979-845-6331 e: hetland(a)tamu.edu, w: http://pong.tamu.edu

2 1