Struct TensorDependencyGraph 

pub struct TensorDependencyGraph {
    pub nodes: HashSet<u64>,
    pub edges: Vec<DependencyEdge>,
    pub dirty: DirtySet,
}

Tracks data dependencies between tensors and rules, enabling incremental recomputation when tensors are updated.

Fields

nodes: HashSet<u64>

All registered node IDs.

edges: Vec<DependencyEdge>

All dependency edges.

dirty: DirtySet

Set of dirty (stale) node IDs.

Implementations

impl TensorDependencyGraph

pub fn new() -> Self

Create a new, empty graph.

pub fn add_node(&mut self, id: u64)

Register a node — idempotent.

pub fn add_edge(&mut self, edge: DependencyEdge)

Add a dependency edge. Also registers the from and to nodes.

pub fn remove_node(&mut self, id: u64)

Remove a node and all edges that reference it. Clears any dirty flag.

pub fn mark_dirty(&mut self, id: u64)

Mark id dirty and transitively propagate to all reachable dependents via TensorOutput, RuleImplication, and SharedFact edges where from_id == id.

pub fn recompute_order(&self) -> Vec<u64>

Topological sort (Kahn’s algorithm) of dirty nodes only. Returns nodes in processing order (dependencies before dependents). On cycle detection, appends remaining nodes in arbitrary order.

pub fn dependents_of(&self, id: u64) -> Vec<u64>

Direct successors of id (edges where from_id == id), sorted ascending.

pub fn dependencies_of(&self, id: u64) -> Vec<u64>

Direct predecessors of id (edges where to_id == id), sorted ascending.

pub fn stats(&self) -> GraphStats

Compute and return a statistics snapshot.

pub fn clear_all_dirty(&mut self)

Clear the entire dirty set.

Trait Implementations

impl Default for TensorDependencyGraph

fn default() -> Self

Returns the “default value” for a type.