Oops. I missed that I was on the hook for code review. I've merged and pushed 
to develop. Thanks for pointing this out.

On Jan 16, 2014, at 2:44 PM, Charles Reid <[email protected]>
 wrote:

> Hi all,
> 
> FYI, I just noticed this was also reported as an issue on Trac 
> (http://matforge.org/fipy/ticket/658) and solved at a better/more upstream 
> location 
> (http://matforge.org/fipy/changeset/2f015d7d9acd1953462a93a52e03c18303c8e2a6/fipy).
>  But it doesn't look like that's made it to the main git branch yet, it's in 
> its own branch.
> 
> 
> 
> 
> Charles
> 
> 
> On Thu, Jan 16, 2014 at 11:24 AM, Charles Reid <[email protected]> wrote:
> Hi,
> 
> I noticed that if I am solving a vector equation, running sweep() instead of 
> solve() raises an error from the pysparseMatrix class's multiplication 
> method. I looked deeper into the error and found that the problem was that a 
> square matrix and a vector, both with the same number of total elements, were 
> being multiplied together, but failing because their sizes were mismatched. 
> The patch contains 5 lines that add a simple conditional, to check if the 
> square matrix and the vector have the same number of elements, in which case 
> I use ravel() on the square matrix (which expands the matrix into a vector).
> 
> Using the ravel() method requires assuming that matrix-vector conversions 
> using C (row-major) order are correct. I verified this was the case using a 
> very simple problem. Any thoughts on whether this is a safe assumption 
> generally?
> 
> I've attached a patch file, which can be applied from in the fipy/ source 
> directory:
> 
> cd fipy/
> patch -p1 < pysparseMatrix.patch
> 
> 
> Charles
> 
> 
> 
> 
> 
> _______________________________________________
> fipy mailing list
> [email protected]
> http://www.ctcms.nist.gov/fipy
>  [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]


_______________________________________________
fipy mailing list
[email protected]
http://www.ctcms.nist.gov/fipy
  [ NIST internal ONLY: https://email.nist.gov/mailman/listinfo/fipy ]

Reply via email to