[Coin-discuss] CommonUnitTest: matrix equivalency

[Matthew Saltzman]

On Fri, 25 Apr 2003, Laszlo Ladanyi wrote:

> I won't. Especially that I plan to rewrite the objective value testing :-).
> Your ideas were really good ones, I think. Although you have still assumed
> that there is a primal and a dual *feasible* solution... The volume alg has
> only a dual feasible sol, the primal value might be above or below that...

Does it make sense to test a primal solution against a primal objective
value and a dual solution against a dual objective value, and to give the
solver realization's unit test the option of which to test (and against
what value)?

>
> --Laci
>
> On Fri, 25 Apr 2003, Lou Hafer wrote:
>
> >
> > 	Would anyone object if I rewrite the exmip1 matrix equivalency tests
> > in OsiSolverInterfaceCommonUnitTest to use the isEquivalent method of the
> > CoinPackedMatrix class? The current test is sensitive to the order of entries
> > in the elements and indices vectors of the packed matrix. Using isEquivalent
> > removes the order sensitivity.
> >
> > 	I've written the necessary code, and it seems to work just fine. As
> > best I can see from looking through the code, isEquivalent seems to do all
> > the tests that are currently performed.

This seems like the right thing to do.  In general, it seems like we want
to use the class methods to do these kinds of things.  If they don't
already, they should be extended.

-- 
		Matthew Saltzman

Clemson University Math Sciences
mjs at clemson.edu
http://www.math.clemson.edu/~mjs