Track all row variables in Constraint.vars

[jackthepunished]

Description

• Constraint.vars on a quadratic row held only the linear-term variables; variables that appear only in quadratic terms were missing.
• updateConstraint did not add newly referenced variables to Constraint.vars, so compute_slack raised KeyError for them.
• Both pre-existing; split out of Add algebraic __str__ and detailed __repr__ to Python LP API classes #1400 at review request. Stores references to existing Variables only, no coefficient copies.

Checklist

• I am familiar with the Contributing Guidelines.
• Testing
  • New or existing tests cover these changes
  • Added tests
  • Created an issue to follow-up
  • NA
• Documentation
  • The documentation is up to date with these changes
  • Added new documentation
  • NA

[copy-pr-bot]

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

[Iroy30]

@jackthepunished Thanks for putting up this PR! Just for context, we're updating the slack computation in #1615, and currently self.vars in the Constraint class is only used for that purpose—it isn't part of the public API intended for users. As part of that work, we'll also be re-evaluating whether we need to store variables in the Constraint class at all.

That said, it looks like you need this for #1400, so we can merge these fixes for now. I've suggested a few changes—let us know if they address both the bug you reported and the issue in the related PR.

[Iroy30]

do we need to use id() here? We could perhaps update this part to hash index

            self.vars = {}
            # Variables may appear only in quadratic terms, so the linear
            # terms of the expression are not enough to cover the row.
            for var in (*expr.vars, *expr.qvars1, *expr.qvars2, *expr.qvars):
                self.vars[var.index] = var

where self.vars is now a dict and can be directly used in compute_slack() rather than reformulating it.

You would need to make the list to dict changes for self.vars in the following lines when expression is Linear, in updateConstraint which would then just add to the dict, plus update the tests.