Lattice Parameterization Inconsistency

[DIvanov503]

Lattice Parameterization Inconsistency

[algobardo]

I think the right change would rather be on the ReversePowersetLattice.
Note that the approach of using the characteristic function still allows you to pass a Set to construct the lattice as Set[A] extends (A) => Boolean: http://www.scala-lang.org/api/2.12.3/scala/collection/Set.html

[DIvanov503]

I take your point. On the other hand, Idea doesn't seem to recognize the type A if I just replace s: Set[A] by ch: A => Boolean in the ReversePowersetLattice, e.g., it complains about the AvailableExpAnalysis and doesn't compile. Adding asInstanceOf operators fix the problem but it doesn't seem to be a good style. However, I'm a Scala novice and might have missed an elegant way of solving it.

By the way, I'm curious whether there exists a use case where defining a lattice over a subset of a type, that itself does not form a type, is needed.

[DIvanov503]

Quickfix: replacing s: Set[A] by ch: A => Boolean works well for my reaching definitions analysis implementation. However, there are still typing warnings from Idea that can be eliminated by asInstanceOf. And it requires explicit construction of the bottom element in the ReversePowersetLattice.

[algobardo]

• As for the use case, having something to pass when constructing the object save us from annotating the generic type. Moreover if you want to be pedantic/defensive the powerset lattice implementation coud check that the lub operation is performed on sets that are subset of the one denoted by ch

• As for the Idea warnings, Idea scala plugin is not as clever as the scala compiler. You have to live with them. I think there is some work going on for making a Language Server for giving the editors compiler info, but I guess we have to wait for that.

[DIvanov503]

• Thank you! The lub return was one of my concerns as well.
• Thanks for clarification!