Mailman 3 May 2009 - NumPy-Discussion

Home for pyhdf5io?
by albert.thuswaldner＠gmail.com 22 May '09

22 May '09

Dear list, I'm writing this because i have developed a small python module that might be of interest to you the readers of this list: http://code.google.com/p/pyhdf5io/ It basically implements load/save functions that mimic the behaviour of those found in Matlab, ie with them you can store your variables from within the interactive shell (IPython, python) or from within a function, and then load them back in again. One important difference is that the hdf5 format is used to store the variables, which comes with aa number of benefits: - a open standard file format which is supported by many applications. - completely portable file format across different platforms. Read more here: http://www.hdfgroup.org/HDF5/whatishdf5.html And now to the question: I think that this module is to small to be developed and maintained on its on, I think It would be better if it could be part of some larger project. So where would pyhdf5io fit in? Any tips and ideas are highly appreciated. Thanks. /Albert

5 5

wiki page correction
by Grant Kelly 22 May '09

22 May '09

I believe there is an error on this wiki page: http://www.scipy.org/NumPy_for_Matlab_Users MATLAB y=x(2,:) PYTHON y = x[2,:].copy() shouldn't the Python version be: y = x[1,:].copy() If not, please advise. Thanks, Grant

2 3

memoryerror with numpy.fromfile
by Michael Hearne 22 May '09

22 May '09

I am getting a MemoryError from a numpy.fromfile() call in an application I am trying to deploy. Normally I would assume that this would mean that I don't have enough memory available on the system. However, if I run vmstat (Linux) at the same time as my process, I see that I have 3+ Gigabytes of memory free, and no swap space being used. I can't think of a way to track down this problem, so I'm punting to the list. The only thing I can imagine is that someone Python has been allocated X amount of space (very small relative to the memory actually available), and is asking for more than X. I don't know if this is true, or if there is even a way to check it. Sigh. Version info: Linux RedHat 5 kernel 2.6.18-128.1.10.el5PAE Python 2.5.4 -- EPD_Py25 4.3.0 numpy 1.3.0 Can anyone suggest more tests that I can do? Thanks, Mike

3 4

ANN: Cython 0.11.2 released
by Dag Sverre Seljebotn 21 May '09

21 May '09

I'm happy to announce the release of Cython 0.11.2. http://sage.math.washington.edu/home/dagss/Cython-0.11.2.tar.gz http://sage.math.washington.edu/home/dagss/Cython-0.11.2.zip (Will be present on the Cython front page in a few days.) New features: * There's now native complex floating point support! C99 complex will be used if complex.h is included, otherwise explicit complex arithmetic working on all C compilers is used. [Robert Bradshaw] cdef double complex a = 1 + 0.3j cdef np.ndarray[np.complex128_t, ndim=2] arr = \ np.zeros(10, np.complex128) * Cython can now generate a main()-method for embedding of the Python interpreter into an executable (see #289) [Robert Bradshaw] * @wraparound directive (another way to disable arr[idx] for negative idx) [Dag Sverre Seljebotn] * Correct support for NumPy record dtypes with different alignments, and "cdef packed struct" support [Dag Sverre Seljebotn] * @callspec directive, allowing custom calling convention macros [Lisandro Dalcin] * Bug fixes and smaller improvements. For the full list, see [1]. Contributors to this release: - Stefan Behnel - Robert Bradshaw - Lisandro Dalcin - Dag Sverre Seljebotn Thanks also to everybody who's helping us out in our discussions on the mailing list. [1] <URL:http://trac.cython.org/cython_trac/query?group=component&milestone=0.11…> -- Dag Sverre

2 2

how to use ldexp?
by dmitrey 21 May '09

21 May '09

hi all, I have tried the example from numpy/add_newdocs.py np.ldexp(5., 2) but instead of the 20 declared there it yields TypeError: function not supported for these types, and can't coerce safely to supported types I have tried arrays but it yields same error >>> np.ldexp(np.array([5., 2.]), np.array([2, 1])) Traceback (innermost last): File "<stdin>", line 1, in <module> TypeError: function not supported for these types, and can't coerce safely to supported types So, how can I use ldexp? np.__version__ = '1.4.0.dev6972' Thank you in advance, D.

4 7

binary builds against older numpys
by John Hunter 21 May '09

21 May '09

We are trying to build and test mpl installers for python2.4, 2.5 and 2.6. What we are finding is that if we build mpl against a more recent numpy than the installed numpy on a test machine, the import of mpl extension modules which depend on numpy trigger a segfault. Eg, on python2.5 and python2.6, we build the mpl installers against the numpy-1.3.0-win32.superpack installation, and if I test the installer on a python2.5 machine with numpy-1.2.1-win32.superpack installed, I get the segfault. If I install numpy-1.3.0-win32.superpack on the test machine, then the mpl binaries work fine. Is there an known binary incompatibly between 1.2.1 and 1.3.0? One solution we may consider is building our 2.5 binaries against 1.2.1 and seeing if they work with both 1.2.1 and 1.3.0 installations, but wanted to check in here to see if there were known issues or solutions we should be considering. Our test installers are at http://drop.io/rlck8ph if you are interested. Thanks, JDH

4 4

numpy failure under Windows Vista 64 bit
by "Klaus Nökel" 21 May '09

21 May '09

David, > > Klaus Noekel wrote: > > I doubt that the DLL was not physically present and rather suspect a > > dependency on some other DLL that was missing. The INSTALL.TXT > > unfortunately was not helpful. Can anybody please explain what other > > dependencies exist? Anything else I need to install? > > > > This exact problem is specific to IDLE - I don't know what triggers it. > Today, the best solution for a 64 bits numpy on windows is to built it > yourself with MS compilers - the distributed one is built with mingw > compilers, and there still seems to be some stability problems with > those. Unfortunately, as the mingw debugger does not work either on 64 > bits archs, finding the problem is quite hard. > I don't believe that the problem is specific to IDLE. Python also crashes when I put nothing but "import numpy" in a file and execute it with python.exe. Regarding the note on building numpy myself: the discussion in this forum scared me a little, because of the challenge to build LAPACK with a compatible Fortran compiler etc. That and the fact that I do not have MSVC 2008 (only 2005) keeps me from trying it. Any chance that a MS-based installer will materialize soon? Or are there any mingw-specific runtime libraries that I need to install so that the mingw-based numpy works? Thanks for your help! Klaus Noekel -- Neu: GMX FreeDSL Komplettanschluss mit DSL 6.000 Flatrate + Telefonanschluss für nur 17,95 Euro/mtl.!* http://dslspecial.gmx.de/freedsl-aktionspreis/?ac=OM.AD.PD003K11308T4569a

3 2

view takes no keyword arguments exception
by Jochen Schroeder 21 May '09

21 May '09

Hi all, I'm trying to help someone out with some problems with pyfftw (pyfftw.berlios.de). He is seeing an exception, TypeError: view() takes no keyword arguments This doesn't only happen when he uses pyfftw but also when he does the following: >>> import numpy as np >>> a=np.arange(10) >>> print a.view(dtype='float') Traceback (most recent call last): File "<stdin>", line 1, in <module> TypeError: view() takes no keyword arguments I he's on Windows and sees this error both with numpy 1.1.1 and 1.3. I'm a bit lost anybody have an idea what could be the problem? Cheers Jochen

4 4

Implementing ufuncs in Cython
by Stéfan van der Walt 21 May '09

21 May '09

Hi all, Mark Lodato outlines how to write ufuncs in Cython at http://wiki.cython.org/MarkLodato/CreatingUfuncs This is also a great way of adding generalised ufuncs: http://projects.scipy.org/numpy/wiki/GeneralLoopingFunctions Super useful! Regards Stéfan

2 1

binary shift for ndarray
by dmitrey 21 May '09

21 May '09

hi all, suppose I have A that is numpy ndarray of floats, with shape n x n. I want to obtain dot(A, b), b is vector of length n and norm(b)=1, but instead of exact multiplication I want to approximate b as a vector [+/- 2^m0, ± 2^m1, ± 2^m2 ,,, ± 2^m_n], m_i are integers, and then invoke left_shift(vector_m) for rows of A. So, what is the simplest way to do it, without cycles of course? Or it cannot be implemented w/o cycles with current numpy version? Thank you in advance, D.

2 3