bug in python/numarray

[Terry Reedy]

<meng at are.berkeley.edu> wrote in message 
news:5.1.1.5.2.20050802103138.04c79d50 at are.berkeley.edu...
> Hi, there,
>
> I got different results by running the same lines of code on windows and
> debian. Here is the code:
>
> a = kroneckerproduct(ones((4195,1)), identity(12))

I don't know what this does, but wonder if it uses a lot of memory.

> print a.mean()
>
> This works perfectly well in windows but it gave the following error 
> while
> running it in debian:

Were you running on one dual boot machine (with the same memory available 
to each OS) or on two different machines, especially with different amounts 
of memory.

> Traceback (most recent call last):
>   File "<stdin>", line 1, in ?
>   File "/usr/lib/python2.4/site-packages/numarray/numarraycore.py", line
> 1137, in mean
>     return self.sum()/(self.nelements()*1.0)
>   File "/usr/lib/python2.4/site-packages/numarray/numarraycore.py", line
> 1133, in sum
>     return ufunc.add.reduce(ufunc.add.areduce(self, type=type).flat,
> type=type)
> IndexError: too many indices.
>
> But if I reduce the number 4195 to 419, it works.

Try using binary search (ie, try 2200 next) to pin down the exact breakdown 
point (as in 2123 works, 2124 does not) if there is one, or whether the 
breakdown is run dependent.  Unload other apps and non-essential background 
process to see if that changes anything

> Is it a bug in Python or  Numarray?

If it is not a resource issue, assume the latter since it crashed in 
Numarray.  I think its maintainers would want to know about it.

Terry J. Reedy