A faster Python?

[Siegfried Gonzi]

Emile van Sebille wrote:
> 
> Nick Arnett
> > To be specific, I do a lot of time series, correlation, regression and
> > clustering analysis.  Those can really bog down.  While I'm fine with
> > exporting to Excel, SPSS, etc., for heavy-duty analysis,
> 
> This surprised me.  I like python with Numeric exactly because it's
> faster and processes more data than Excel.  I'm now regularly
> calculating and delivering pivot-style results from a million row source
> with 240 columns.

Where it comes --otherwise to your experience-- that people believe that
the Numeric-package is slow? I did not compare Numeric to lets say
Matlab or IDL. But why is it that people believe that Numeric is slower
than for example Matlab (for me, when one mentions Numric in combination
with speed, this implies that Numeric is way slower than Matlab or IDL)?

My accompanied Yorick manual says that Yorick can gain 10% of the speed
of compiled C code. I, personaly, made the experience that Yorick is
quite as fast as Matlab or IDL. 


Please, people should realize that Matlab or IDL too is slow compared to
C!

The above writing will not help improving the speed of Python (which
itself, without compiled libraries, would be really  a pain in the
neck).

I think it is important to mention Numeric in combination with Matlab or
IDL or Yorick or...otherwise one could easily fall for it and believe
that Numeric is inherently slow; but I believe my notion is not far away
from reality that there are not any tools available  like
Matlab,IDL,Yorick,Numric  which have matured to compiled C speed in a
side by side comparison.

Or are there any hints that Numric is really way slower than Matlab?
Please no matrix-matrix multiplication; this test is useful for a
compiled language and when on does implement it himself. But there are
often quite a few differences how a specific task is implemented (e.g.
the IDL FFT does care on arrays which dimensionality is not exactly
2^n).


S. Gonzi