Change addConsNode to (also) accept ExprCons

[CharJon]

Based on discussion in #416
To keep the prototype simple a new addExprConsNode method is added.
Do you prefer having addConsNode accept ExprCons only or both (Constraint and ExprCons)?

[fserra]

thanks. I think it would make sense to factor out the code that creates the linear constraint since it also appears in _addLinCons

Regarding the name and if it should also be part of addConsNode I am unsure. I am pinging @mattmilten here.
So we have addCons and addPyCons for when the argument is an ExprCons or a Constraint, resp.
So for consistency, we could have addConsNode and addPyConsNode.
But then we also have addConsLocal and we would need to do the same there.

An alternative would be to remove addPyCons and let addCons handle both cases.

[CharJon]

Personally I think the addPyCons naming is a bit confusing, as the Constraint objects are the ones actually having a pointer to a SCIPConstraint, while the ExprCons addCons expects as input don't. If you prefer having a single method for both cases I will restructure the code.

[mattmilten]

I prefer having a nice and easy-to-use method addCons that just accepts whatever you throw at it - be it ExprCons or Constraint.
And the same also holds for addConsLocal, etc., of course.

[CharJon]

I created a new method _createCons and made changes to addCons and addConsNode to avoid duplicate code. The current version of addConsNode works for Constraint and ExprCons. There a two TODOs in the code as I'm still unsure about how to handle the reference counting of SCIP. If you have any sources I could study to understand why and when PY_SCIP_CALL(SCIPreleaseCons(self._scip, &scip_cons)) and Py_INCREF(cons) are necessary please tell me.

[mattmilten]

Looks good to me. Just the CHANGELOG entry is missing.

[SanielDous]

Did this find its way into master branch already? Would be really nice

[zzy-opt]

I am implementing the branch and price where I need to add a linear constraint to the newly created node in branching rule. I am facing the exact same problem stated here.
The suggested solution definitely can solve my problem. Howerever I don't see this updated version of addConsNode function exists. What goes wrong here? Isn't it successfully merged?

[SanielDous]

@zzy-opt In #416 there is some more recent information about this topic.

[zzy-opt]

@SanielDous Thank you. It seems this cannot be fixed in short term. Maybe I had better figure out how to use cython and change source code myself.