Add add_cons_local, add_cons_node

[Joao-Dionisio]

Still not done, want to take my time to make things properly, as in adding tests and actually understanding what chatGPT combined with Rust's compiler have created. Ah, and probably want to create the constraint directly on the function, and also want to add add_cons_node.

@mmghannam a question about the russcip philosophy. Is it like PySCIPOpt where you want to mimic SCIP as much as possible?

[mmghannam]

@Joao-Dionisio happy to see you here :)

Unlike PySCIPOpt one could here call the SCIP api directly from the ffi module. So russcip is more free in what it can offer. The philosophy is mainly two things:

1. Make the default use case easy.
2. Try to avoid as many user errors as possible at compile time.

[mmghannam]

I think in this case an add_cons method on the node struct (that takes a ConsBuilder object) would be more ergonomic.

[Joao-Dionisio]

I think in this case an add_cons method on the node struct (that takes a ConsBuilder object) would be more ergonomic.

Same, that's what made me question russcip's philosophy :) Just need to grab the local model from the node, but it should be okay

EDIT: on the other hand, I'm not seeing how to access the local scip

EDIT2: Should be mostly done. Just need to add a test, which will be a bit annoying because of the branching requirement... unless it's enough to test if adding at the root node is enough 🤔

[mmghannam]

I think these need to be only on Model<Solving>. I am not sure if it makes sense at any other stage. Then, you could modify the test in the separator.rs to addConsLocal instead, and you can modify the custom branching rule test to additionally use addConsNode.

[Joao-Dionisio]

Wait @mmghannam , did you manage to fix the underlying problem?

Why wasn't node_get_n_added_cons not working? I'm assuming it was due to SCIP somehow treating the current node differently, but I couldn't figure out what was wrong.

[mmghannam]

Wait @mmghannam , did you manage to fix the underlying problem?

Why wasn't node_get_n_added_cons not working? I'm assuming it was due to SCIP somehow treating the current node differently, but I couldn't figure out what was wrong.

The function was working with the use of add_cons_node() in the branching rule tests. I think add_cons_local doesn't have the same effect on the current node, it might be upgraded to a global constraint since it was running in the root node but I'm not sure.