matrix algebra

Bas wegwerp at gmail.com
Wed Sep 24 15:58:05 CEST 2008


On Sep 22, 10:02 am, Al Kabaila <akaba... at pcug.org.au> wrote:
> There are several packages for matrix algebra. I tried Numeric, numpy and
> numarray. All three are very good, but each uses different syntax.
That argument might have been valid 3 years ago, but as already said
by others, Numeric and Numarray are deprecated. Numpy should be the
only thing needed for new users. I suggest you investigate a little
bit more the next time you make such efforts, since this fact should
be widely known among the users of the mentioned packages, see e.g.
the huge warning at the numarray page:
http://www.stsci.edu/resources/software_hardware/numarray/numarray.html

> 1. Is there any interest in matrix algebra "for the masses" (I mean interest
> in a wrapper for a subset of functions of the packages with a unified
> simple syntax)?
In my opinion, no. I might be biased, since with my matlab background
I find numpy simple enough as is. But I don't see how A = B*C+D or
E=dot(F,G) is complicated for a beginner of linear algebra.

> My OS is Linux (openSUSE 10.3) and my interest in retirement is Python
> applications to Structural Analysis of  Civil Engineering structures,
> currently in 2 dimensions only (under GPL). Modern Structural Analysis is
> highly matrix oriented, but requires only a few basic matrix operations,
> namely matrix creation, transposition, multiplication, invertion and
> linear equation solution. For stability analysis one would require
> Eigenvalues and Eigenvectors. In 3 dimensions, additionally highly
> desirable would be vector algebra. The packages do have all these
> functions, but currently only the basic functions are in the wrapper.
If you care about contributing something useful to the community, I
think your time and skills are better spent writing some cool
mechanical analysis tool for inclusion in Scipy.

Bas



More information about the Python-list mailing list