Struct rustc_trait_selection::solve::search_graph::SearchGraph

pub(super) struct SearchGraph<'tcx> {
    mode: SolverMode,
    stack: IndexVec<StackDepth, StackEntry<'tcx>>,
    provisional_cache: FxHashMap<CanonicalInput<'tcx>, ProvisionalCacheEntry<'tcx>>,
    cycle_participants: FxHashSet<CanonicalInput<'tcx>>,
}

Fields

mode: SolverMode

stack: IndexVec<StackDepth, StackEntry<'tcx>>

The stack of goals currently being computed.

An element is deeper in the stack if its index is lower.

provisional_cache: FxHashMap<CanonicalInput<'tcx>, ProvisionalCacheEntry<'tcx>>

cycle_participants: FxHashSet<CanonicalInput<'tcx>>

We put only the root goal of a coinductive cycle into the global cache.

If we were to use that result when later trying to prove another cycle participant, we can end up with unstable query results.

See tests/ui/next-solver/coinduction/incompleteness-unstable-result.rs for an example of where this is needed.

Implementations

impl<'tcx> SearchGraph<'tcx>

pub(super) fn new(mode: SolverMode) -> SearchGraph<'tcx>

pub(super) fn solver_mode(&self) -> SolverMode

fn on_cache_hit(&mut self, additional_depth: usize, encountered_overflow: bool)

Update the stack and reached depths on cache hits.

fn pop_stack(&mut self) -> StackEntry<'tcx>

Pops the highest goal from the stack, lazily updating the the next goal in the stack.

Directly popping from the stack instead of using this method would cause us to not track overflow and recursion depth correctly.

pub(super) fn global_cache( &self, tcx: TyCtxt<'tcx> ) -> &'tcx EvaluationCache<'tcx>

The trait solver behavior is different for coherence so we use a separate cache. Alternatively we could use a single cache and share it between coherence and ordinary trait solving.

pub(super) fn is_empty(&self) -> bool

pub(super) fn current_goal_is_normalizes_to(&self) -> bool

fn allowed_depth_for_nested( tcx: TyCtxt<'tcx>, stack: &IndexVec<StackDepth, StackEntry<'tcx>> ) -> Option<Limit>

Returns the remaining depth allowed for nested goals.

This is generally simply one less than the current depth. However, if we encountered overflow, we significantly reduce the remaining depth of all nested goals to prevent hangs in case there is exponential blowup.

fn stack_coinductive_from( tcx: TyCtxt<'tcx>, stack: &IndexVec<StackDepth, StackEntry<'tcx>>, head: StackDepth ) -> bool

fn tag_cycle_participants( stack: &mut IndexVec<StackDepth, StackEntry<'tcx>>, cycle_participants: &mut FxHashSet<CanonicalInput<'tcx>>, usage_kind: HasBeenUsed, head: StackDepth )

fn clear_dependent_provisional_results( provisional_cache: &mut FxHashMap<CanonicalInput<'tcx>, ProvisionalCacheEntry<'tcx>>, head: StackDepth )

pub(super) fn with_new_goal( &mut self, tcx: TyCtxt<'tcx>, input: CanonicalInput<'tcx>, inspect: &mut ProofTreeBuilder<'tcx>, prove_goal: impl FnMut(&mut Self, &mut ProofTreeBuilder<'tcx>) -> QueryResult<'tcx> ) -> QueryResult<'tcx>

Probably the most involved method of the whole solver.

Given some goal which is proven via the prove_goal closure, this handles caching, overflow, and coinductive cycles.

fn response_no_constraints( tcx: TyCtxt<'tcx>, goal: CanonicalInput<'tcx>, certainty: Certainty ) -> QueryResult<'tcx>

Layout

Note: Most layout information is completely unstable and may even differ between compilations. The only exception is types with certain repr(...) attributes. Please see the Rust Reference's “Type Layout” chapter for details on type layout guarantees.

Size: 96 bytes