Re: [Pysparse-users] Compiling pysparse on a 64 bit machine

[Artur P. <art...@gm...>]

Hello Daniel

Once again I manage to solve the problem.

When I was trying to make a simple program that mimics the problem I
have, to send you, as you asked, I found the problem. Maybe there is a
bug with pysparse or the problem is mine.

The simplest program is:

import numpy
import pysparse

GN = numpy.array([0,1,2,3])

B = pysparse.spmatrix.ll_mat(10, 10, 10)

for i in range(0,2):
    B[GN[i], GN[i]] += 10

print B


If you do this it works. But if you use:

GN = numpy.array([0,1,2,3], dtype="int32")

It gives the error I stated.

If one prints the dtype of GN of the first code, it states that it is
int64. Hence the problem is that one cannot give in32 as an index to a
ll_mat on a 64bit machine, or at least on mine.

Once again thank you for your help!

-artur palha

On Wed, Jul 9, 2008 at 3:16 PM, Artur Palha <art...@gm...> wrote:
> Thank you for the help, again.
>
> The error it gives is:
>
> IndexError: first index is invalid
>
> And the error comes up when I try to assign a value to an entry of the
> B matrix in this way:
>
> B[ GN[e,n,m], GN[e,i,j] ] += something
>
> B is the sparse matrix and GN is a 3d array.
>
> I have checked that the values of GN are correct. The exact same code
> works on another machine. I even tried a simpler code and it gives the
> same error. The strange thing is that if I try to do it in ipython,
> line by line, it works. Even if I assign a the value of GN to another
> variable, as:
>
> a = GN[e,n,m]
> b = GN[e,i,j]
>
> and then try:
>
> B[a,b] += something
>
> it does not work also, or if I assing to a and b constant values as:
>
> a = 1
> b = 2
>
> B[a,b] += something
>
> It only works if I do:
>
> B[1,2] += something
>
> Any help?
>
> Thank you
>
> -artur palha
>
> On Thu, Jul 3, 2008 at 6:06 PM, Daniel Wheeler
> <dan...@gm...> wrote:
>> On Thu, Jul 3, 2008 at 8:50 AM, Artur Palha <art...@gm...> wrote:
>>> I hope someone can help me
>>>
>>> I am trying to compile pysparse on a 64 bit machine with openSuse 10.3.
>>>
>>> It compiles but gives an error acessing the sparse arrays.
>>
>> What is the error? I use pysparse on a 64 bit machine without problems.
>>
>>> I read on
>>> the README file that there is an incompatibility issue with pysparse
>>> and numpy on 64bit machines and that a patch is needed.
>>
>> According to the CVS log, the message was there from the initial
>> import of pysparse. I'll remove it.
>>
>>> I have
>>> installed version 1.1.0 of Numpy do I need the patch or the problem is
>>> somewhere else?
>>
>> Can you give some more details?
>>
>> --
>> Daniel Wheeler
>>
>