[Soot-list] hash function for the instancekey

[Zhoulai]

Than you for your replies.

Do you think it is *implementable*  to  hash instance keys s.t.
*          whenever two keys have different hashes  then the intersection
of their points-to sets is empty?*

This seems  motivating because such hash function would make instance keys
still easier to use, and thus make it more popular.

I formalized this  naive question at the forum
http://cs.stackexchange.com/questions/3480/algorithm-that-hashes-a-collection-of-sets-following-their-disjointness-relation

but no constructive answers yet... Please share with me if you have any
idea.

Thanks.

Zell.

Do you have some ideas?




On Mon, Sep 17, 2012 at 8:22 PM, Eric Bodden <eric.bodden at ec-spride.de>wrote:

> >> But when I read the " Objective representatives"
> >> paper, I think  the author was insisting that instancekey can almost be
> used
> >> as run-time objects in some extent.
> >
> > What I wrote is nevertheless correct in the sense that also at runtime
> > two different objects may have the same hash code, so also there the
> > conclusion "different hash codes => different objects" does not hold.
>
> Oups, sorry, my mistake. Of course "different hash codes => different
> objects" holds at runtime. But with InstanceKey objects that's not the
> case, they may or may not alias (to tell the difference, one needs an
> additional must-not-alias check). That's just the result of the
> three-valued logic that instance keys implement.
>
> For details I recommend to also look at our recent TOPLAS paper (section
> 7.6.1):
> http://www.bodden.de/pubs/blh12partially.pdf
>
> Eric
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.cs.mcgill.ca/pipermail/soot-list/attachments/20120917/ed48cd92/attachment.html