pub struct ScopeTree {
    pub root_body: Option<HirId>,
    pub parent_map: FxIndexMap<Scope, (Scope, ScopeDepth)>,
    var_map: FxIndexMap<ItemLocalId, Scope>,
    destruction_scopes: FxIndexMap<ItemLocalId, Scope>,
    pub rvalue_candidates: FxHashMap<HirId, RvalueCandidateType>,
    pub yield_in_scope: FxHashMap<Scope, Vec<YieldData>>,
    pub body_expr_count: FxHashMap<BodyId, usize>,
Expand description

The region scope tree encodes information about region relationships.


root_body: Option<HirId>

If not empty, this body is the root of this region hierarchy.

parent_map: FxIndexMap<Scope, (Scope, ScopeDepth)>

Maps from a scope ID to the enclosing scope id; this is usually corresponding to the lexical nesting, though in the case of closures the parent scope is the innermost conditional expression or repeating block. (Note that the enclosing scope ID for the block associated with a closure is the closure itself.)

var_map: FxIndexMap<ItemLocalId, Scope>

Maps from a variable or binding ID to the block in which that variable is declared.

destruction_scopes: FxIndexMap<ItemLocalId, Scope>

Maps from a NodeId to the associated destruction scope (if any).

rvalue_candidates: FxHashMap<HirId, RvalueCandidateType>

Identifies expressions which, if captured into a temporary, ought to have a temporary whose lifetime extends to the end of the enclosing block, and not the enclosing statement. Expressions that are not present in this table are not rvalue candidates. The set of rvalue candidates is computed during type check based on a traversal of the AST.

yield_in_scope: FxHashMap<Scope, Vec<YieldData>>

If there are any yield nested within a scope, this map stores the Span of the last one and its index in the postorder of the Visitor traversal on the HIR.

HIR Visitor postorder indexes might seem like a peculiar thing to care about. but it turns out that HIR bindings and the temporary results of HIR expressions are never storage-live at the end of HIR nodes with postorder indexes lower than theirs, and therefore don’t need to be suspended at yield-points at these indexes.

For an example, suppose we have some code such as:

    foo(f(), yield y, bar(g()))

With the HIR tree (calls numbered for expository purposes)

    Call#0(foo, [Call#1(f), Yield(y), Call#2(bar, Call#3(g))])

Obviously, the result of f() was created before the yield (and therefore needs to be kept valid over the yield) while the result of g() occurs after the yield (and therefore doesn’t). If we want to infer that, we can look at the postorder traversal:

    `foo` `f` Call#1 `y` Yield `bar` `g` Call#3 Call#2 Call#0

In which we can easily see that Call#1 occurs before the yield, and Call#3 after it.

To see that this method works, consider:

Let D be our binding/temporary and U be our other HIR node, with HIR-postorder(U) < HIR-postorder(D). Suppose, as in our example, U is the yield and D is one of the calls. Let’s show that D is storage-dead at U.

Remember that storage-live/storage-dead refers to the state of the storage, and does not consider moves/drop flags.


  1. From the ordering guarantee of HIR visitors (see rustc_hir::intravisit), D does not dominate U.

  2. Therefore, D is potentially storage-dead at U (because we might visit U without ever getting to D).

  3. However, we guarantee that at each HIR point, each binding/temporary is always either always storage-live or always storage-dead. This is what is being guaranteed by terminating_scopes including all blocks where the count of executions is not guaranteed.

  4. By 2. and 3., D is statically storage-dead at U, QED.

This property ought to not on (3) in an essential way – it is probably still correct even if we have “unrestricted” terminating scopes. However, why use the complicated proof when a simple one works?

A subtle thing: box expressions, such as box (&x, yield 2, &y). It might seem that a box expression creates a Box<T> temporary when it starts executing, at HIR-preorder(BOX-EXPR). That might be true in the MIR desugaring, but it is not important in the semantics.

The reason is that semantically, until the box expression returns, the values are still owned by their containing expressions. So we’ll see that &x.

body_expr_count: FxHashMap<BodyId, usize>

The number of visit_expr and visit_pat calls done in the body. Used to sanity check visit_expr/visit_pat call count when calculating generator interiors.


Returns the narrowest scope that encloses id, if any.

Returns the lifetime of the local variable var_id, if any.

Returns true if subscope is equal to or is lexically nested inside superscope, and false otherwise.

Used by clippy.

Checks whether the given scope contains a yield. If so, returns Some(YieldData). If not, returns None.

Gives the number of expressions visited in a body. Used to sanity check visit_expr call count when calculating generator interiors.

Trait Implementations

Formats the value using the given formatter. Read more

Returns the “default value” for a type. Read more

Auto Trait Implementations

Blanket Implementations

Gets the TypeId of self. Read more

Immutably borrows from an owned value. Read more

Mutably borrows from an owned value. Read more

This method turns the parameters of a DepNodeConstructor into an opaque Fingerprint to be used in DepNode. Not all DepNodeParams support being turned into a Fingerprint (they don’t need to if the corresponding DepNode is anonymous). Read more

This method tries to recover the query key from the given DepNode, something which is needed when forcing DepNodes during red-green evaluation. The query system will only call this method if fingerprint_style() is not FingerprintStyle::Opaque. It is always valid to return None here, in which case incremental compilation will treat the query as having changed instead of forcing it. Read more

Returns the argument unchanged.

Calls U::from(self).

That is, this conversion is whatever the implementation of From<T> for U chooses to do.

The type returned in the event of a conversion error.

Performs the conversion.

The type returned in the event of a conversion error.

Performs the conversion.


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: 272 bytes