blanking out rows and columns

Hi!

I'm using some large sparse matrices, at points in my code i have to "blank 
out" rows and columns. With that i mean i have my large sparse matrix, and i 
want a number, around 10% of the rows and columns maximum, of those blanked 
out to resemble the identity matrix. Rows and columns are always tied 
together, so if i want row i blanked, i will also want column i blanked. 
Example:

1 2 3        1 0 3
4 5 6 =>   0 1 0
7 8 9        7 0 9

In this example i have blanked row and column 1, 1 on the diagonal and 0 on 
all other entries.

The question is, how do i do this the fastest using MTL? It is pure horror 
if i just iterate through it using a for(i = x; i < y; i++).

I hope you can help me out :)

/Michael 


_______________________________________________
This list is archived at http://www.osl.iu.edu/MailArchives/mtl-devel/