]> git.proxmox.com Git - rustc.git/blob - compiler/rustc_middle/src/middle/region.rs
New upstream version 1.50.0+dfsg1
[rustc.git] / compiler / rustc_middle / src / middle / region.rs
1 //! This file declares the `ScopeTree` type, which describes
2 //! the parent links in the region hierarchy.
3 //!
4 //! For more information about how MIR-based region-checking works,
5 //! see the [rustc dev guide].
6 //!
7 //! [rustc dev guide]: https://rustc-dev-guide.rust-lang.org/borrow_check.html
8
9 use crate::ich::{NodeIdHashingMode, StableHashingContext};
10 use crate::ty::TyCtxt;
11 use rustc_hir as hir;
12 use rustc_hir::Node;
13
14 use rustc_data_structures::fx::FxHashMap;
15 use rustc_data_structures::stable_hasher::{HashStable, StableHasher};
16 use rustc_macros::HashStable;
17 use rustc_span::{Span, DUMMY_SP};
18
19 use std::fmt;
20
21 /// Represents a statically-describable scope that can be used to
22 /// bound the lifetime/region for values.
23 ///
24 /// `Node(node_id)`: Any AST node that has any scope at all has the
25 /// `Node(node_id)` scope. Other variants represent special cases not
26 /// immediately derivable from the abstract syntax tree structure.
27 ///
28 /// `DestructionScope(node_id)` represents the scope of destructors
29 /// implicitly-attached to `node_id` that run immediately after the
30 /// expression for `node_id` itself. Not every AST node carries a
31 /// `DestructionScope`, but those that are `terminating_scopes` do;
32 /// see discussion with `ScopeTree`.
33 ///
34 /// `Remainder { block, statement_index }` represents
35 /// the scope of user code running immediately after the initializer
36 /// expression for the indexed statement, until the end of the block.
37 ///
38 /// So: the following code can be broken down into the scopes beneath:
39 ///
40 /// ```text
41 /// let a = f().g( 'b: { let x = d(); let y = d(); x.h(y) } ) ;
42 ///
43 /// +-+ (D12.)
44 /// +-+ (D11.)
45 /// +---------+ (R10.)
46 /// +-+ (D9.)
47 /// +----------+ (M8.)
48 /// +----------------------+ (R7.)
49 /// +-+ (D6.)
50 /// +----------+ (M5.)
51 /// +-----------------------------------+ (M4.)
52 /// +--------------------------------------------------+ (M3.)
53 /// +--+ (M2.)
54 /// +-----------------------------------------------------------+ (M1.)
55 ///
56 /// (M1.): Node scope of the whole `let a = ...;` statement.
57 /// (M2.): Node scope of the `f()` expression.
58 /// (M3.): Node scope of the `f().g(..)` expression.
59 /// (M4.): Node scope of the block labeled `'b:`.
60 /// (M5.): Node scope of the `let x = d();` statement
61 /// (D6.): DestructionScope for temporaries created during M5.
62 /// (R7.): Remainder scope for block `'b:`, stmt 0 (let x = ...).
63 /// (M8.): Node scope of the `let y = d();` statement.
64 /// (D9.): DestructionScope for temporaries created during M8.
65 /// (R10.): Remainder scope for block `'b:`, stmt 1 (let y = ...).
66 /// (D11.): DestructionScope for temporaries and bindings from block `'b:`.
67 /// (D12.): DestructionScope for temporaries created during M1 (e.g., f()).
68 /// ```
69 ///
70 /// Note that while the above picture shows the destruction scopes
71 /// as following their corresponding node scopes, in the internal
72 /// data structures of the compiler the destruction scopes are
73 /// represented as enclosing parents. This is sound because we use the
74 /// enclosing parent relationship just to ensure that referenced
75 /// values live long enough; phrased another way, the starting point
76 /// of each range is not really the important thing in the above
77 /// picture, but rather the ending point.
78 //
79 // FIXME(pnkfelix): this currently derives `PartialOrd` and `Ord` to
80 // placate the same deriving in `ty::FreeRegion`, but we may want to
81 // actually attach a more meaningful ordering to scopes than the one
82 // generated via deriving here.
83 #[derive(Clone, PartialEq, PartialOrd, Eq, Ord, Hash, Copy, TyEncodable, TyDecodable)]
84 #[derive(HashStable)]
85 pub struct Scope {
86 pub id: hir::ItemLocalId,
87 pub data: ScopeData,
88 }
89
90 impl fmt::Debug for Scope {
91 fn fmt(&self, fmt: &mut fmt::Formatter<'_>) -> fmt::Result {
92 match self.data {
93 ScopeData::Node => write!(fmt, "Node({:?})", self.id),
94 ScopeData::CallSite => write!(fmt, "CallSite({:?})", self.id),
95 ScopeData::Arguments => write!(fmt, "Arguments({:?})", self.id),
96 ScopeData::Destruction => write!(fmt, "Destruction({:?})", self.id),
97 ScopeData::Remainder(fsi) => write!(
98 fmt,
99 "Remainder {{ block: {:?}, first_statement_index: {}}}",
100 self.id,
101 fsi.as_u32(),
102 ),
103 }
104 }
105 }
106
107 #[derive(Clone, PartialEq, PartialOrd, Eq, Ord, Hash, Debug, Copy, TyEncodable, TyDecodable)]
108 #[derive(HashStable)]
109 pub enum ScopeData {
110 Node,
111
112 /// Scope of the call-site for a function or closure
113 /// (outlives the arguments as well as the body).
114 CallSite,
115
116 /// Scope of arguments passed to a function or closure
117 /// (they outlive its body).
118 Arguments,
119
120 /// Scope of destructors for temporaries of node-id.
121 Destruction,
122
123 /// Scope following a `let id = expr;` binding in a block.
124 Remainder(FirstStatementIndex),
125 }
126
127 rustc_index::newtype_index! {
128 /// Represents a subscope of `block` for a binding that is introduced
129 /// by `block.stmts[first_statement_index]`. Such subscopes represent
130 /// a suffix of the block. Note that each subscope does not include
131 /// the initializer expression, if any, for the statement indexed by
132 /// `first_statement_index`.
133 ///
134 /// For example, given `{ let (a, b) = EXPR_1; let c = EXPR_2; ... }`:
135 ///
136 /// * The subscope with `first_statement_index == 0` is scope of both
137 /// `a` and `b`; it does not include EXPR_1, but does include
138 /// everything after that first `let`. (If you want a scope that
139 /// includes EXPR_1 as well, then do not use `Scope::Remainder`,
140 /// but instead another `Scope` that encompasses the whole block,
141 /// e.g., `Scope::Node`.
142 ///
143 /// * The subscope with `first_statement_index == 1` is scope of `c`,
144 /// and thus does not include EXPR_2, but covers the `...`.
145 pub struct FirstStatementIndex {
146 derive [HashStable]
147 }
148 }
149
150 // compilation error if size of `ScopeData` is not the same as a `u32`
151 static_assert_size!(ScopeData, 4);
152
153 impl Scope {
154 /// Returns a item-local ID associated with this scope.
155 ///
156 /// N.B., likely to be replaced as API is refined; e.g., pnkfelix
157 /// anticipates `fn entry_node_id` and `fn each_exit_node_id`.
158 pub fn item_local_id(&self) -> hir::ItemLocalId {
159 self.id
160 }
161
162 pub fn hir_id(&self, scope_tree: &ScopeTree) -> Option<hir::HirId> {
163 scope_tree
164 .root_body
165 .map(|hir_id| hir::HirId { owner: hir_id.owner, local_id: self.item_local_id() })
166 }
167
168 /// Returns the span of this `Scope`. Note that in general the
169 /// returned span may not correspond to the span of any `NodeId` in
170 /// the AST.
171 pub fn span(&self, tcx: TyCtxt<'_>, scope_tree: &ScopeTree) -> Span {
172 let hir_id = match self.hir_id(scope_tree) {
173 Some(hir_id) => hir_id,
174 None => return DUMMY_SP,
175 };
176 let span = tcx.hir().span(hir_id);
177 if let ScopeData::Remainder(first_statement_index) = self.data {
178 if let Node::Block(ref blk) = tcx.hir().get(hir_id) {
179 // Want span for scope starting after the
180 // indexed statement and ending at end of
181 // `blk`; reuse span of `blk` and shift `lo`
182 // forward to end of indexed statement.
183 //
184 // (This is the special case alluded to in the
185 // doc-comment for this method)
186
187 let stmt_span = blk.stmts[first_statement_index.index()].span;
188
189 // To avoid issues with macro-generated spans, the span
190 // of the statement must be nested in that of the block.
191 if span.lo() <= stmt_span.lo() && stmt_span.lo() <= span.hi() {
192 return Span::new(stmt_span.lo(), span.hi(), span.ctxt());
193 }
194 }
195 }
196 span
197 }
198 }
199
200 pub type ScopeDepth = u32;
201
202 /// The region scope tree encodes information about region relationships.
203 #[derive(Default, Debug)]
204 pub struct ScopeTree {
205 /// If not empty, this body is the root of this region hierarchy.
206 pub root_body: Option<hir::HirId>,
207
208 /// The parent of the root body owner, if the latter is an
209 /// an associated const or method, as impls/traits can also
210 /// have lifetime parameters free in this body.
211 pub root_parent: Option<hir::HirId>,
212
213 /// Maps from a scope ID to the enclosing scope id;
214 /// this is usually corresponding to the lexical nesting, though
215 /// in the case of closures the parent scope is the innermost
216 /// conditional expression or repeating block. (Note that the
217 /// enclosing scope ID for the block associated with a closure is
218 /// the closure itself.)
219 pub parent_map: FxHashMap<Scope, (Scope, ScopeDepth)>,
220
221 /// Maps from a variable or binding ID to the block in which that
222 /// variable is declared.
223 var_map: FxHashMap<hir::ItemLocalId, Scope>,
224
225 /// Maps from a `NodeId` to the associated destruction scope (if any).
226 destruction_scopes: FxHashMap<hir::ItemLocalId, Scope>,
227
228 /// `rvalue_scopes` includes entries for those expressions whose
229 /// cleanup scope is larger than the default. The map goes from the
230 /// expression ID to the cleanup scope id. For rvalues not present in
231 /// this table, the appropriate cleanup scope is the innermost
232 /// enclosing statement, conditional expression, or repeating
233 /// block (see `terminating_scopes`).
234 /// In constants, None is used to indicate that certain expressions
235 /// escape into 'static and should have no local cleanup scope.
236 rvalue_scopes: FxHashMap<hir::ItemLocalId, Option<Scope>>,
237
238 /// Encodes the hierarchy of fn bodies. Every fn body (including
239 /// closures) forms its own distinct region hierarchy, rooted in
240 /// the block that is the fn body. This map points from the ID of
241 /// that root block to the ID of the root block for the enclosing
242 /// fn, if any. Thus the map structures the fn bodies into a
243 /// hierarchy based on their lexical mapping. This is used to
244 /// handle the relationships between regions in a fn and in a
245 /// closure defined by that fn. See the "Modeling closures"
246 /// section of the README in infer::region_constraints for
247 /// more details.
248 closure_tree: FxHashMap<hir::ItemLocalId, hir::ItemLocalId>,
249
250 /// If there are any `yield` nested within a scope, this map
251 /// stores the `Span` of the last one and its index in the
252 /// postorder of the Visitor traversal on the HIR.
253 ///
254 /// HIR Visitor postorder indexes might seem like a peculiar
255 /// thing to care about. but it turns out that HIR bindings
256 /// and the temporary results of HIR expressions are never
257 /// storage-live at the end of HIR nodes with postorder indexes
258 /// lower than theirs, and therefore don't need to be suspended
259 /// at yield-points at these indexes.
260 ///
261 /// For an example, suppose we have some code such as:
262 /// ```rust,ignore (example)
263 /// foo(f(), yield y, bar(g()))
264 /// ```
265 ///
266 /// With the HIR tree (calls numbered for expository purposes)
267 /// ```
268 /// Call#0(foo, [Call#1(f), Yield(y), Call#2(bar, Call#3(g))])
269 /// ```
270 ///
271 /// Obviously, the result of `f()` was created before the yield
272 /// (and therefore needs to be kept valid over the yield) while
273 /// the result of `g()` occurs after the yield (and therefore
274 /// doesn't). If we want to infer that, we can look at the
275 /// postorder traversal:
276 /// ```plain,ignore
277 /// `foo` `f` Call#1 `y` Yield `bar` `g` Call#3 Call#2 Call#0
278 /// ```
279 ///
280 /// In which we can easily see that `Call#1` occurs before the yield,
281 /// and `Call#3` after it.
282 ///
283 /// To see that this method works, consider:
284 ///
285 /// Let `D` be our binding/temporary and `U` be our other HIR node, with
286 /// `HIR-postorder(U) < HIR-postorder(D)`. Suppose, as in our example,
287 /// U is the yield and D is one of the calls.
288 /// Let's show that `D` is storage-dead at `U`.
289 ///
290 /// Remember that storage-live/storage-dead refers to the state of
291 /// the *storage*, and does not consider moves/drop flags.
292 ///
293 /// Then:
294 ///
295 /// 1. From the ordering guarantee of HIR visitors (see
296 /// `rustc_hir::intravisit`), `D` does not dominate `U`.
297 ///
298 /// 2. Therefore, `D` is *potentially* storage-dead at `U` (because
299 /// we might visit `U` without ever getting to `D`).
300 ///
301 /// 3. However, we guarantee that at each HIR point, each
302 /// binding/temporary is always either always storage-live
303 /// or always storage-dead. This is what is being guaranteed
304 /// by `terminating_scopes` including all blocks where the
305 /// count of executions is not guaranteed.
306 ///
307 /// 4. By `2.` and `3.`, `D` is *statically* storage-dead at `U`,
308 /// QED.
309 ///
310 /// This property ought to not on (3) in an essential way -- it
311 /// is probably still correct even if we have "unrestricted" terminating
312 /// scopes. However, why use the complicated proof when a simple one
313 /// works?
314 ///
315 /// A subtle thing: `box` expressions, such as `box (&x, yield 2, &y)`. It
316 /// might seem that a `box` expression creates a `Box<T>` temporary
317 /// when it *starts* executing, at `HIR-preorder(BOX-EXPR)`. That might
318 /// be true in the MIR desugaring, but it is not important in the semantics.
319 ///
320 /// The reason is that semantically, until the `box` expression returns,
321 /// the values are still owned by their containing expressions. So
322 /// we'll see that `&x`.
323 pub yield_in_scope: FxHashMap<Scope, YieldData>,
324
325 /// The number of visit_expr and visit_pat calls done in the body.
326 /// Used to sanity check visit_expr/visit_pat call count when
327 /// calculating generator interiors.
328 pub body_expr_count: FxHashMap<hir::BodyId, usize>,
329 }
330
331 #[derive(Debug, Copy, Clone, TyEncodable, TyDecodable, HashStable)]
332 pub struct YieldData {
333 /// The `Span` of the yield.
334 pub span: Span,
335 /// The number of expressions and patterns appearing before the `yield` in the body, plus one.
336 pub expr_and_pat_count: usize,
337 pub source: hir::YieldSource,
338 }
339
340 impl ScopeTree {
341 pub fn record_scope_parent(&mut self, child: Scope, parent: Option<(Scope, ScopeDepth)>) {
342 debug!("{:?}.parent = {:?}", child, parent);
343
344 if let Some(p) = parent {
345 let prev = self.parent_map.insert(child, p);
346 assert!(prev.is_none());
347 }
348
349 // Record the destruction scopes for later so we can query them.
350 if let ScopeData::Destruction = child.data {
351 self.destruction_scopes.insert(child.item_local_id(), child);
352 }
353 }
354
355 pub fn opt_destruction_scope(&self, n: hir::ItemLocalId) -> Option<Scope> {
356 self.destruction_scopes.get(&n).cloned()
357 }
358
359 /// Records that `sub_closure` is defined within `sup_closure`. These IDs
360 /// should be the ID of the block that is the fn body, which is
361 /// also the root of the region hierarchy for that fn.
362 pub fn record_closure_parent(
363 &mut self,
364 sub_closure: hir::ItemLocalId,
365 sup_closure: hir::ItemLocalId,
366 ) {
367 debug!(
368 "record_closure_parent(sub_closure={:?}, sup_closure={:?})",
369 sub_closure, sup_closure
370 );
371 assert!(sub_closure != sup_closure);
372 let previous = self.closure_tree.insert(sub_closure, sup_closure);
373 assert!(previous.is_none());
374 }
375
376 pub fn record_var_scope(&mut self, var: hir::ItemLocalId, lifetime: Scope) {
377 debug!("record_var_scope(sub={:?}, sup={:?})", var, lifetime);
378 assert!(var != lifetime.item_local_id());
379 self.var_map.insert(var, lifetime);
380 }
381
382 pub fn record_rvalue_scope(&mut self, var: hir::ItemLocalId, lifetime: Option<Scope>) {
383 debug!("record_rvalue_scope(sub={:?}, sup={:?})", var, lifetime);
384 if let Some(lifetime) = lifetime {
385 assert!(var != lifetime.item_local_id());
386 }
387 self.rvalue_scopes.insert(var, lifetime);
388 }
389
390 /// Returns the narrowest scope that encloses `id`, if any.
391 pub fn opt_encl_scope(&self, id: Scope) -> Option<Scope> {
392 self.parent_map.get(&id).cloned().map(|(p, _)| p)
393 }
394
395 /// Returns the lifetime of the local variable `var_id`
396 pub fn var_scope(&self, var_id: hir::ItemLocalId) -> Scope {
397 self.var_map
398 .get(&var_id)
399 .cloned()
400 .unwrap_or_else(|| bug!("no enclosing scope for id {:?}", var_id))
401 }
402
403 /// Returns the scope when the temp created by `expr_id` will be cleaned up.
404 pub fn temporary_scope(&self, expr_id: hir::ItemLocalId) -> Option<Scope> {
405 // Check for a designated rvalue scope.
406 if let Some(&s) = self.rvalue_scopes.get(&expr_id) {
407 debug!("temporary_scope({:?}) = {:?} [custom]", expr_id, s);
408 return s;
409 }
410
411 // Otherwise, locate the innermost terminating scope
412 // if there's one. Static items, for instance, won't
413 // have an enclosing scope, hence no scope will be
414 // returned.
415 let mut id = Scope { id: expr_id, data: ScopeData::Node };
416
417 while let Some(&(p, _)) = self.parent_map.get(&id) {
418 match p.data {
419 ScopeData::Destruction => {
420 debug!("temporary_scope({:?}) = {:?} [enclosing]", expr_id, id);
421 return Some(id);
422 }
423 _ => id = p,
424 }
425 }
426
427 debug!("temporary_scope({:?}) = None", expr_id);
428 None
429 }
430
431 /// Returns `true` if `subscope` is equal to or is lexically nested inside `superscope`, and
432 /// `false` otherwise.
433 pub fn is_subscope_of(&self, subscope: Scope, superscope: Scope) -> bool {
434 let mut s = subscope;
435 debug!("is_subscope_of({:?}, {:?})", subscope, superscope);
436 while superscope != s {
437 match self.opt_encl_scope(s) {
438 None => {
439 debug!("is_subscope_of({:?}, {:?}, s={:?})=false", subscope, superscope, s);
440 return false;
441 }
442 Some(scope) => s = scope,
443 }
444 }
445
446 debug!("is_subscope_of({:?}, {:?})=true", subscope, superscope);
447
448 true
449 }
450
451 /// Checks whether the given scope contains a `yield`. If so,
452 /// returns `Some(YieldData)`. If not, returns `None`.
453 pub fn yield_in_scope(&self, scope: Scope) -> Option<YieldData> {
454 self.yield_in_scope.get(&scope).cloned()
455 }
456
457 /// Gives the number of expressions visited in a body.
458 /// Used to sanity check visit_expr call count when
459 /// calculating generator interiors.
460 pub fn body_expr_count(&self, body_id: hir::BodyId) -> Option<usize> {
461 self.body_expr_count.get(&body_id).copied()
462 }
463 }
464
465 impl<'a> HashStable<StableHashingContext<'a>> for ScopeTree {
466 fn hash_stable(&self, hcx: &mut StableHashingContext<'a>, hasher: &mut StableHasher) {
467 let ScopeTree {
468 root_body,
469 root_parent,
470 ref body_expr_count,
471 ref parent_map,
472 ref var_map,
473 ref destruction_scopes,
474 ref rvalue_scopes,
475 ref closure_tree,
476 ref yield_in_scope,
477 } = *self;
478
479 hcx.with_node_id_hashing_mode(NodeIdHashingMode::HashDefPath, |hcx| {
480 root_body.hash_stable(hcx, hasher);
481 root_parent.hash_stable(hcx, hasher);
482 });
483
484 body_expr_count.hash_stable(hcx, hasher);
485 parent_map.hash_stable(hcx, hasher);
486 var_map.hash_stable(hcx, hasher);
487 destruction_scopes.hash_stable(hcx, hasher);
488 rvalue_scopes.hash_stable(hcx, hasher);
489 closure_tree.hash_stable(hcx, hasher);
490 yield_in_scope.hash_stable(hcx, hasher);
491 }
492 }