Re: [OctDev] Re: sparse tests

[Paul K. <pki...@us...>]

On Dec 13, 2003, at 1:48 PM, Andy Adler wrote:

> On Sat, 13 Dec 2003, Paul Kienzle wrote:
>
>> It would be a little more convincing if other operators didn't
>> convert to full as necessary, like + and /.  I haven't checked
>> functions like cos/sin though.
>
> I didn't think + converted to full. "/" will, though. I didn't
> think it was used enough to bother implementing.
>
> All the advanced functions (sin/cos etc.) will convert to full.
>
> For cos it doesn't make sense to have a sparse implementation
> since cos(0)=1. ie. the matrix will become filled with ones.
>
> For sin, sinh, tan, erf, abs, conj, floor, round, ceil, fix,
>     tanh, sign, asin, asinh, atan, sqrt (any more?)
> in which func(0)= 0, I could write a template that does
> the function on each non-sparse element. Then we could define
> sp_func for each, which could dispatch for sparse operands.
>
> Is this worth the work?

These are easy enough to add as scripts:

function x = spfn(fn,y)
    [i,j,v,m,n]=spfind(y);
    x = sparse(i,j,feval(fn,v),m,n);
end

Or if they want ease of use:

## PKG_ADD: dispatch('sin','spsin','sparse');
## PKG_ADD: dispatch('sin','spsin','complex_sparse');
function x=spsin(y), x=spfn('sin',y); end