[PPL-devel] Zero-dimensional constraint systems.

[Roberto Bagnara]

Enea Zaffanella wrote:
> 
> In various places in the source files, we have assertions requiring that
> matrices (of constraints and generators) have a number of columns <> 1.
> These assertions can be violated by the following code, 

[...]

> There are three ways to solve such a situation:
> 

[Two kludges removed ;-)]

> 3. we remove all assertions of this kind. Non-empty 0-dim constraint
> systems will be legal and will contain constraints that are either
> always false (like the one above) or always true (like the positivity
> constraint 1 >= 0). Note that the representation of the empty polyhedron
> needs not to be changed, i.e., we just look at the status flag and do
> not need to explicitly insert have a matrix for the always false
> constraint.
> 
> The easiest-to-implement solution is 1.
> The cleanest one, in my opinion, is solution 3. Here are some other
> reasons:

[Good reasons removed]

> Is anyone foreseeing other problems for such a change ?

I am very much in favor of solution 3.  In fact, while developing
the initial draft of the Prolog interface I have come across the
same problems Enea pointed out and I made also some choices that
are in the direction of solution 3 above.

Elisa, can you give this problem (i.e., seeing whether solution 3
is feasible) the highest priority?
Ciao,

    Roberto

-- 
Roberto Bagnara
Computer Science Group
Department of Mathematics, University of Parma, Italy
http://www.cs.unipr.it/~bagnara/
mailto:bagnara at cs.unipr.it