Struct ConstraintGraph 

pub struct ConstraintGraph { /* private fields */ }

Constraint dependency graph.

Models which constraints share variables. This enables:

• Graph coloring to find independent constraint sets
• Parallel propagation within independent sets
• Efficient incremental solving after small changes

Implementations

impl ConstraintGraph

pub fn new(num_vars: usize) -> Self

Create a new empty constraint graph over num_vars variables.

pub fn num_constraints(&self) -> usize

Number of constraints in the graph.

pub fn num_vars(&self) -> usize

Number of variables in the graph.

pub fn add_constraint( &mut self, constraint: Box<dyn FastConstraint>, var_indices: Vec<usize>, )

Add a constraint that touches the given variable indices.

The adjacency list is updated so that this constraint becomes adjacent to all existing constraints that share at least one variable index.

pub fn independent_sets(&self) -> Vec<Vec<usize>>

Greedy graph coloring — returns independent sets (constraints with same color).

Uses Welsh-Powell ordering (decreasing degree) for better coloring quality. Constraints in the same independent set can be checked/projected in parallel.

Returns

A Vec<Vec<usize>> where each inner Vec is a set of constraint indices that are mutually non-adjacent (i.e., share no variables).

pub fn propagate_parallel(&self, x: &mut Array1<f32>) -> PropagationResult

Parallel constraint propagation.

Projects x using constraints in independent-set order. Within each independent set, constraints are applied in parallel via rayon (each one operates on disjoint variable subsets, so there are no races).

Falls back to sequential application when variable sets overlap within a color (the graph coloring guarantees no overlap for correctly registered constraints).