Struct cranelift_egraph::EGraph

pub struct EGraph<L: Language, A: Analysis<L = L>> {
    pub nodes: Vec<L::Node>,
    pub classes: PrimaryMap<Id, EClass>,
    pub unionfind: UnionFind,
    pub analysis: Option<(A, SecondaryMap<Id, A::Value>)>,
    /* private fields */
}

An egraph.

Fields

nodes: Vec<L::Node>

Node-allocation arena.

classes: PrimaryMap<Id, EClass>

Eclass definitions. Each eclass consists of an enode, and child pointer to the rest of the eclass.

unionfind: UnionFind

Union-find for canonical ID generation. This lets us name an eclass with a canonical ID that is the same for all generations of the class.

analysis: Option<(A, SecondaryMap<Id, A::Value>)>

Analysis and per-node state.

Implementations

impl<L: Language, A: Analysis<L = L>> EGraph<L, A>where L::Node: 'static,

pub fn new(analysis: Option<A>) -> Self

Create a new aegraph.

pub fn with_capacity(nodes: usize, analysis: Option<A>) -> Self

Create a new aegraph with the given capacity.

pub fn add(&mut self, node: L::Node, node_ctx: &L) -> NewOrExisting<Id>

Add a new node.

pub fn union(&mut self, ctx: &L, a: Id, b: Id) -> Id

Merge one eclass into another, maintaining the acyclic property (args must have lower eclass Ids than the eclass containing the node with those args). Returns the Id of the merged eclass.

pub fn canonical_id_mut(&mut self, eclass: Id) -> Id

Get the canonical ID for an eclass. This may be an older generation, so will not be able to see all enodes in the eclass; but it will allow us to unambiguously refer to an eclass, even across merging.

pub fn canonical_id(&self, eclass: Id) -> Id

Get the canonical ID for an eclass. This may be an older generation, so will not be able to see all enodes in the eclass; but it will allow us to unambiguously refer to an eclass, even across merging.

pub fn enodes(&self, eclass: Id) -> NodeIter<L, A>

Get the enodes for a given eclass.

pub fn analysis_value(&self, eclass: Id) -> &A::Value

Get the analysis value for a given eclass. Panics if no analysis is present.