Lattice constrains not enforced on input in ascent_run

[B-Lorentz]

The following fails in 0.5.0:

 assert_eq!(ascent_run! {
        lattice a(i64) = vec![(0,),(1,)];
    }.a, vec![(1,)]);

Presumably because the lattice joins are not executed on the incoming vector, as it is seen as 'input'
I think it would make sense to generate an intermediate input relation, so the lattice's rules are upheld.

[s-arash]

I see! The issue (or one issue) with trying to "fix" this is that Ascent never removes elements from input vectors, and I think modifying this behavior is probably not worth it to get the above program to behave as you would expect.

Note that you could easily achieve what you are looking for by doing something like this:

ascent! {
  relation a_proto(i64) = vec![(0,), (1,)];
  lattice a(i64);
  a(x) <-- a_proto(x);
}

Or this:

ascent! {
  lattice a(i64);
  a(x) <-- for x in vec![0, 1];
}

[B-Lorentz]

@s-arash
Thanks. I'm aware of the workarounds. What if ascent just refused programs where lattices are direct input relations, in order to prevent the user being surprised (as either her input facts disappearing or the lattice rules not being obeyed can be surprising)?
This wouldn't remove any expressive power, as you have shown easy workarounds exist.