Dear all,

Added to my last email, I would like to know the possibilities from PPL to symbolically manipulate interval domains? The attachment is a slide of Yang Hongseok on the subject widening/narrowing. He gave an example to show that finite abstract domains are in general less powerful than widening/narrowing approaches.  

Could you please tell me how to use uniquely PPL's interval abstract domain to infer the relational properties indicated in the attached slide?  

Thanks a lot.

Zell.

On Sun, Aug 19, 2012 at 2:56 PM, Zell <zell08v@orange.fr> wrote:
Hello,

I still have a superficial question.

Consider the domain  D:= Var-> Interval, where "Interval"  contains two special  intervals, " Empty " and  " Universe " . I am now using Int_64 from PPL to model this domain. However I feel confused when I need to represent elements that " mixes the 'Empty' and 'Universe' " . One such element is,  { x->[3,5], y->[4,7], z->empty, w->universe}.

I find that, whenever a constraint like "z->empty" is added to a constraints_system, the constraints_system becomes "false". This seems reasonable because the constraint_system in PPL models a conjunction of constraints and thus one constraint is false make the whole constraint system unsatisfiable. However, I do not see how to represent elements  like the one given above that mixes 'z->Empty' and 'w->Universe' . ( I would like to use 'z->empty' to mean 'z is not yet initialized', and 'w->universe' to mean that 'w is unknown' .)

I probably misunderstand something basic here. What do you think?

Thanks.
Zell.

p.s Due to the help of Enea and Roberto, I am not able to use PPL in my analysis. I really appreciate your help!!! By the way, I have been using the pseudo name "Zell" because I used to write my name as "Zell".  For information, my passport name is "Zhoulai", which can be more or less pronounced as "zell".