1 //! Types that pin data to its location in memory.
3 //! It is sometimes useful to have objects that are guaranteed not to move,
4 //! in the sense that their placement in memory does not change, and can thus be relied upon.
5 //! A prime example of such a scenario would be building self-referential structs,
6 //! as moving an object with pointers to itself will invalidate them, which could cause undefined
9 //! At a high level, a <code>[Pin]\<P></code> ensures that the pointee of any pointer type
10 //! `P` has a stable location in memory, meaning it cannot be moved elsewhere
11 //! and its memory cannot be deallocated until it gets dropped. We say that the
12 //! pointee is "pinned". Things get more subtle when discussing types that
13 //! combine pinned with non-pinned data; [see below](#projections-and-structural-pinning)
16 //! By default, all types in Rust are movable. Rust allows passing all types by-value,
17 //! and common smart-pointer types such as <code>[Box]\<T></code> and <code>[&mut] T</code> allow
18 //! replacing and moving the values they contain: you can move out of a <code>[Box]\<T></code>,
19 //! or you can use [`mem::swap`]. <code>[Pin]\<P></code> wraps a pointer type `P`, so
20 //! <code>[Pin]<[Box]\<T>></code> functions much like a regular <code>[Box]\<T></code>:
21 //! when a <code>[Pin]<[Box]\<T>></code> gets dropped, so do its contents, and the memory gets
22 //! deallocated. Similarly, <code>[Pin]<[&mut] T></code> is a lot like <code>[&mut] T</code>.
23 //! However, <code>[Pin]\<P></code> does not let clients actually obtain a <code>[Box]\<T></code>
24 //! or <code>[&mut] T</code> to pinned data, which implies that you cannot use operations such
28 //! use std::pin::Pin;
29 //! fn swap_pins<T>(x: Pin<&mut T>, y: Pin<&mut T>) {
30 //! // `mem::swap` needs `&mut T`, but we cannot get it.
31 //! // We are stuck, we cannot swap the contents of these references.
32 //! // We could use `Pin::get_unchecked_mut`, but that is unsafe for a reason:
33 //! // we are not allowed to use it for moving things out of the `Pin`.
37 //! It is worth reiterating that <code>[Pin]\<P></code> does *not* change the fact that a Rust
38 //! compiler considers all types movable. [`mem::swap`] remains callable for any `T`. Instead,
39 //! <code>[Pin]\<P></code> prevents certain *values* (pointed to by pointers wrapped in
40 //! <code>[Pin]\<P></code>) from being moved by making it impossible to call methods that require
41 //! <code>[&mut] T</code> on them (like [`mem::swap`]).
43 //! <code>[Pin]\<P></code> can be used to wrap any pointer type `P`, and as such it interacts with
44 //! [`Deref`] and [`DerefMut`]. A <code>[Pin]\<P></code> where <code>P: [Deref]</code> should be
45 //! considered as a "`P`-style pointer" to a pinned <code>P::[Target]</code> – so, a
46 //! <code>[Pin]<[Box]\<T>></code> is an owned pointer to a pinned `T`, and a
47 //! <code>[Pin]<[Rc]\<T>></code> is a reference-counted pointer to a pinned `T`.
48 //! For correctness, <code>[Pin]\<P></code> relies on the implementations of [`Deref`] and
49 //! [`DerefMut`] not to move out of their `self` parameter, and only ever to
50 //! return a pointer to pinned data when they are called on a pinned pointer.
54 //! Many types are always freely movable, even when pinned, because they do not
55 //! rely on having a stable address. This includes all the basic types (like
56 //! [`bool`], [`i32`], and references) as well as types consisting solely of these
57 //! types. Types that do not care about pinning implement the [`Unpin`]
58 //! auto-trait, which cancels the effect of <code>[Pin]\<P></code>. For <code>T: [Unpin]</code>,
59 //! <code>[Pin]<[Box]\<T>></code> and <code>[Box]\<T></code> function identically, as do
60 //! <code>[Pin]<[&mut] T></code> and <code>[&mut] T</code>.
62 //! Note that pinning and [`Unpin`] only affect the pointed-to type <code>P::[Target]</code>,
63 //! not the pointer type `P` itself that got wrapped in <code>[Pin]\<P></code>. For example,
64 //! whether or not <code>[Box]\<T></code> is [`Unpin`] has no effect on the behavior of
65 //! <code>[Pin]<[Box]\<T>></code> (here, `T` is the pointed-to type).
67 //! # Example: self-referential struct
69 //! Before we go into more details to explain the guarantees and choices
70 //! associated with <code>[Pin]\<P></code>, we discuss some examples for how it might be used.
71 //! Feel free to [skip to where the theoretical discussion continues](#drop-guarantee).
74 //! use std::pin::Pin;
75 //! use std::marker::PhantomPinned;
76 //! use std::ptr::NonNull;
78 //! // This is a self-referential struct because the slice field points to the data field.
79 //! // We cannot inform the compiler about that with a normal reference,
80 //! // as this pattern cannot be described with the usual borrowing rules.
81 //! // Instead we use a raw pointer, though one which is known not to be null,
82 //! // as we know it's pointing at the string.
83 //! struct Unmovable {
85 //! slice: NonNull<String>,
86 //! _pin: PhantomPinned,
90 //! // To ensure the data doesn't move when the function returns,
91 //! // we place it in the heap where it will stay for the lifetime of the object,
92 //! // and the only way to access it would be through a pointer to it.
93 //! fn new(data: String) -> Pin<Box<Self>> {
94 //! let res = Unmovable {
96 //! // we only create the pointer once the data is in place
97 //! // otherwise it will have already moved before we even started
98 //! slice: NonNull::dangling(),
99 //! _pin: PhantomPinned,
101 //! let mut boxed = Box::pin(res);
103 //! let slice = NonNull::from(&boxed.data);
104 //! // we know this is safe because modifying a field doesn't move the whole struct
106 //! let mut_ref: Pin<&mut Self> = Pin::as_mut(&mut boxed);
107 //! Pin::get_unchecked_mut(mut_ref).slice = slice;
113 //! let unmoved = Unmovable::new("hello".to_string());
114 //! // The pointer should point to the correct location,
115 //! // so long as the struct hasn't moved.
116 //! // Meanwhile, we are free to move the pointer around.
117 //! # #[allow(unused_mut)]
118 //! let mut still_unmoved = unmoved;
119 //! assert_eq!(still_unmoved.slice, NonNull::from(&still_unmoved.data));
121 //! // Since our type doesn't implement Unpin, this will fail to compile:
122 //! // let mut new_unmoved = Unmovable::new("world".to_string());
123 //! // std::mem::swap(&mut *still_unmoved, &mut *new_unmoved);
126 //! # Example: intrusive doubly-linked list
128 //! In an intrusive doubly-linked list, the collection does not actually allocate
129 //! the memory for the elements itself. Allocation is controlled by the clients,
130 //! and elements can live on a stack frame that lives shorter than the collection does.
132 //! To make this work, every element has pointers to its predecessor and successor in
133 //! the list. Elements can only be added when they are pinned, because moving the elements
134 //! around would invalidate the pointers. Moreover, the [`Drop`][Drop] implementation of a linked
135 //! list element will patch the pointers of its predecessor and successor to remove itself
138 //! Crucially, we have to be able to rely on [`drop`] being called. If an element
139 //! could be deallocated or otherwise invalidated without calling [`drop`], the pointers into it
140 //! from its neighboring elements would become invalid, which would break the data structure.
142 //! Therefore, pinning also comes with a [`drop`]-related guarantee.
144 //! # `Drop` guarantee
146 //! The purpose of pinning is to be able to rely on the placement of some data in memory.
147 //! To make this work, not just moving the data is restricted; deallocating, repurposing, or
148 //! otherwise invalidating the memory used to store the data is restricted, too.
149 //! Concretely, for pinned data you have to maintain the invariant
150 //! that *its memory will not get invalidated or repurposed from the moment it gets pinned until
151 //! when [`drop`] is called*. Only once [`drop`] returns or panics, the memory may be reused.
153 //! Memory can be "invalidated" by deallocation, but also by
154 //! replacing a <code>[Some]\(v)</code> by [`None`], or calling [`Vec::set_len`] to "kill" some
155 //! elements off of a vector. It can be repurposed by using [`ptr::write`] to overwrite it without
156 //! calling the destructor first. None of this is allowed for pinned data without calling [`drop`].
158 //! This is exactly the kind of guarantee that the intrusive linked list from the previous
159 //! section needs to function correctly.
161 //! Notice that this guarantee does *not* mean that memory does not leak! It is still
162 //! completely okay to not ever call [`drop`] on a pinned element (e.g., you can still
163 //! call [`mem::forget`] on a <code>[Pin]<[Box]\<T>></code>). In the example of the doubly-linked
164 //! list, that element would just stay in the list. However you must not free or reuse the storage
165 //! *without calling [`drop`]*.
167 //! # `Drop` implementation
169 //! If your type uses pinning (such as the two examples above), you have to be careful
170 //! when implementing [`Drop`][Drop]. The [`drop`] function takes <code>[&mut] self</code>, but this
171 //! is called *even if your type was previously pinned*! It is as if the
172 //! compiler automatically called [`Pin::get_unchecked_mut`].
174 //! This can never cause a problem in safe code because implementing a type that
175 //! relies on pinning requires unsafe code, but be aware that deciding to make
176 //! use of pinning in your type (for example by implementing some operation on
177 //! <code>[Pin]<[&]Self></code> or <code>[Pin]<[&mut] Self></code>) has consequences for your
178 //! [`Drop`][Drop] implementation as well: if an element of your type could have been pinned,
179 //! you must treat [`Drop`][Drop] as implicitly taking <code>[Pin]<[&mut] Self></code>.
181 //! For example, you could implement [`Drop`][Drop] as follows:
184 //! # use std::pin::Pin;
185 //! # struct Type { }
186 //! impl Drop for Type {
187 //! fn drop(&mut self) {
188 //! // `new_unchecked` is okay because we know this value is never used
189 //! // again after being dropped.
190 //! inner_drop(unsafe { Pin::new_unchecked(self)});
191 //! fn inner_drop(this: Pin<&mut Type>) {
192 //! // Actual drop code goes here.
198 //! The function `inner_drop` has the type that [`drop`] *should* have, so this makes sure that
199 //! you do not accidentally use `self`/`this` in a way that is in conflict with pinning.
201 //! Moreover, if your type is `#[repr(packed)]`, the compiler will automatically
202 //! move fields around to be able to drop them. It might even do
203 //! that for fields that happen to be sufficiently aligned. As a consequence, you cannot use
204 //! pinning with a `#[repr(packed)]` type.
206 //! # Projections and Structural Pinning
208 //! When working with pinned structs, the question arises how one can access the
209 //! fields of that struct in a method that takes just <code>[Pin]<[&mut] Struct></code>.
210 //! The usual approach is to write helper methods (so called *projections*)
211 //! that turn <code>[Pin]<[&mut] Struct></code> into a reference to the field, but what type should
212 //! that reference have? Is it <code>[Pin]<[&mut] Field></code> or <code>[&mut] Field</code>?
213 //! The same question arises with the fields of an `enum`, and also when considering
214 //! container/wrapper types such as <code>[Vec]\<T></code>, <code>[Box]\<T></code>,
215 //! or <code>[RefCell]\<T></code>. (This question applies to both mutable and shared references,
216 //! we just use the more common case of mutable references here for illustration.)
218 //! It turns out that it is actually up to the author of the data structure to decide whether
219 //! the pinned projection for a particular field turns <code>[Pin]<[&mut] Struct></code>
220 //! into <code>[Pin]<[&mut] Field></code> or <code>[&mut] Field</code>. There are some
221 //! constraints though, and the most important constraint is *consistency*:
222 //! every field can be *either* projected to a pinned reference, *or* have
223 //! pinning removed as part of the projection. If both are done for the same field,
224 //! that will likely be unsound!
226 //! As the author of a data structure you get to decide for each field whether pinning
227 //! "propagates" to this field or not. Pinning that propagates is also called "structural",
228 //! because it follows the structure of the type.
229 //! In the following subsections, we describe the considerations that have to be made
230 //! for either choice.
232 //! ## Pinning *is not* structural for `field`
234 //! It may seem counter-intuitive that the field of a pinned struct might not be pinned,
235 //! but that is actually the easiest choice: if a <code>[Pin]<[&mut] Field></code> is never created,
236 //! nothing can go wrong! So, if you decide that some field does not have structural pinning,
237 //! all you have to ensure is that you never create a pinned reference to that field.
239 //! Fields without structural pinning may have a projection method that turns
240 //! <code>[Pin]<[&mut] Struct></code> into <code>[&mut] Field</code>:
243 //! # use std::pin::Pin;
244 //! # type Field = i32;
245 //! # struct Struct { field: Field }
247 //! fn pin_get_field(self: Pin<&mut Self>) -> &mut Field {
248 //! // This is okay because `field` is never considered pinned.
249 //! unsafe { &mut self.get_unchecked_mut().field }
254 //! You may also <code>impl [Unpin] for Struct</code> *even if* the type of `field`
255 //! is not [`Unpin`]. What that type thinks about pinning is not relevant
256 //! when no <code>[Pin]<[&mut] Field></code> is ever created.
258 //! ## Pinning *is* structural for `field`
260 //! The other option is to decide that pinning is "structural" for `field`,
261 //! meaning that if the struct is pinned then so is the field.
263 //! This allows writing a projection that creates a <code>[Pin]<[&mut] Field></code>, thus
264 //! witnessing that the field is pinned:
267 //! # use std::pin::Pin;
268 //! # type Field = i32;
269 //! # struct Struct { field: Field }
271 //! fn pin_get_field(self: Pin<&mut Self>) -> Pin<&mut Field> {
272 //! // This is okay because `field` is pinned when `self` is.
273 //! unsafe { self.map_unchecked_mut(|s| &mut s.field) }
278 //! However, structural pinning comes with a few extra requirements:
280 //! 1. The struct must only be [`Unpin`] if all the structural fields are
281 //! [`Unpin`]. This is the default, but [`Unpin`] is a safe trait, so as the author of
282 //! the struct it is your responsibility *not* to add something like
283 //! <code>impl\<T> [Unpin] for Struct\<T></code>. (Notice that adding a projection operation
284 //! requires unsafe code, so the fact that [`Unpin`] is a safe trait does not break
285 //! the principle that you only have to worry about any of this if you use [`unsafe`].)
286 //! 2. The destructor of the struct must not move structural fields out of its argument. This
287 //! is the exact point that was raised in the [previous section][drop-impl]: [`drop`] takes
288 //! <code>[&mut] self</code>, but the struct (and hence its fields) might have been pinned
289 //! before. You have to guarantee that you do not move a field inside your [`Drop`][Drop]
290 //! implementation. In particular, as explained previously, this means that your struct
291 //! must *not* be `#[repr(packed)]`.
292 //! See that section for how to write [`drop`] in a way that the compiler can help you
293 //! not accidentally break pinning.
294 //! 3. You must make sure that you uphold the [`Drop` guarantee][drop-guarantee]:
295 //! once your struct is pinned, the memory that contains the
296 //! content is not overwritten or deallocated without calling the content's destructors.
297 //! This can be tricky, as witnessed by <code>[VecDeque]\<T></code>: the destructor of
298 //! <code>[VecDeque]\<T></code> can fail to call [`drop`] on all elements if one of the
299 //! destructors panics. This violates the [`Drop`][Drop] guarantee, because it can lead to
300 //! elements being deallocated without their destructor being called.
301 //! (<code>[VecDeque]\<T></code> has no pinning projections, so this
302 //! does not cause unsoundness.)
303 //! 4. You must not offer any other operations that could lead to data being moved out of
304 //! the structural fields when your type is pinned. For example, if the struct contains an
305 //! <code>[Option]\<T></code> and there is a [`take`][Option::take]-like operation with type
306 //! <code>fn([Pin]<[&mut] Struct\<T>>) -> [Option]\<T></code>,
307 //! that operation can be used to move a `T` out of a pinned `Struct<T>` – which means
308 //! pinning cannot be structural for the field holding this data.
310 //! For a more complex example of moving data out of a pinned type,
311 //! imagine if <code>[RefCell]\<T></code> had a method
312 //! <code>fn get_pin_mut(self: [Pin]<[&mut] Self>) -> [Pin]<[&mut] T></code>.
313 //! Then we could do the following:
315 //! fn exploit_ref_cell<T>(rc: Pin<&mut RefCell<T>>) {
316 //! { let p = rc.as_mut().get_pin_mut(); } // Here we get pinned access to the `T`.
317 //! let rc_shr: &RefCell<T> = rc.into_ref().get_ref();
318 //! let b = rc_shr.borrow_mut();
319 //! let content = &mut *b; // And here we have `&mut T` to the same data.
322 //! This is catastrophic, it means we can first pin the content of the
323 //! <code>[RefCell]\<T></code> (using <code>[RefCell]::get_pin_mut</code>) and then move that
324 //! content using the mutable reference we got later.
328 //! For a type like <code>[Vec]\<T></code>, both possibilities (structural pinning or not) make
329 //! sense. A <code>[Vec]\<T></code> with structural pinning could have `get_pin`/`get_pin_mut`
330 //! methods to get pinned references to elements. However, it could *not* allow calling
331 //! [`pop`][Vec::pop] on a pinned <code>[Vec]\<T></code> because that would move the (structurally
332 //! pinned) contents! Nor could it allow [`push`][Vec::push], which might reallocate and thus also
333 //! move the contents.
335 //! A <code>[Vec]\<T></code> without structural pinning could
336 //! <code>impl\<T> [Unpin] for [Vec]\<T></code>, because the contents are never pinned
337 //! and the <code>[Vec]\<T></code> itself is fine with being moved as well.
338 //! At that point pinning just has no effect on the vector at all.
340 //! In the standard library, pointer types generally do not have structural pinning,
341 //! and thus they do not offer pinning projections. This is why <code>[Box]\<T>: [Unpin]</code>
342 //! holds for all `T`. It makes sense to do this for pointer types, because moving the
343 //! <code>[Box]\<T></code> does not actually move the `T`: the <code>[Box]\<T></code> can be freely
344 //! movable (aka [`Unpin`]) even if the `T` is not. In fact, even <code>[Pin]<[Box]\<T>></code> and
345 //! <code>[Pin]<[&mut] T></code> are always [`Unpin`] themselves, for the same reason:
346 //! their contents (the `T`) are pinned, but the pointers themselves can be moved without moving
347 //! the pinned data. For both <code>[Box]\<T></code> and <code>[Pin]<[Box]\<T>></code>,
348 //! whether the content is pinned is entirely independent of whether the
349 //! pointer is pinned, meaning pinning is *not* structural.
351 //! When implementing a [`Future`] combinator, you will usually need structural pinning
352 //! for the nested futures, as you need to get pinned references to them to call [`poll`].
353 //! But if your combinator contains any other data that does not need to be pinned,
354 //! you can make those fields not structural and hence freely access them with a
355 //! mutable reference even when you just have <code>[Pin]<[&mut] Self></code> (such as in your own
356 //! [`poll`] implementation).
358 //! [Deref]: crate::ops::Deref "ops::Deref"
359 //! [`Deref`]: crate::ops::Deref "ops::Deref"
360 //! [Target]: crate::ops::Deref::Target "ops::Deref::Target"
361 //! [`DerefMut`]: crate::ops::DerefMut "ops::DerefMut"
362 //! [`mem::swap`]: crate::mem::swap "mem::swap"
363 //! [`mem::forget`]: crate::mem::forget "mem::forget"
364 //! [Vec]: ../../std/vec/struct.Vec.html "Vec"
365 //! [`Vec::set_len`]: ../../std/vec/struct.Vec.html#method.set_len "Vec::set_len"
366 //! [Box]: ../../std/boxed/struct.Box.html "Box"
367 //! [Vec::pop]: ../../std/vec/struct.Vec.html#method.pop "Vec::pop"
368 //! [Vec::push]: ../../std/vec/struct.Vec.html#method.push "Vec::push"
369 //! [Rc]: ../../std/rc/struct.Rc.html "rc::Rc"
370 //! [RefCell]: crate::cell::RefCell "cell::RefCell"
371 //! [`drop`]: Drop::drop
372 //! [VecDeque]: ../../std/collections/struct.VecDeque.html "collections::VecDeque"
373 //! [`ptr::write`]: crate::ptr::write "ptr::write"
374 //! [`Future`]: crate::future::Future "future::Future"
375 //! [drop-impl]: #drop-implementation
376 //! [drop-guarantee]: #drop-guarantee
377 //! [`poll`]: crate::future::Future::poll "future::Future::poll"
378 //! [&]: reference "shared reference"
379 //! [&mut]: reference "mutable reference"
380 //! [`unsafe`]: ../../std/keyword.unsafe.html "keyword unsafe"
382 #![stable(feature = "pin", since = "1.33.0")]
384 use crate::cmp::{self, PartialEq, PartialOrd};
386 use crate::hash::{Hash, Hasher};
387 use crate::marker::{Sized, Unpin};
388 use crate::ops::{CoerceUnsized, Deref, DerefMut, DispatchFromDyn, Receiver};
390 /// A pinned pointer.
392 /// This is a wrapper around a kind of pointer which makes that pointer "pin" its
393 /// value in place, preventing the value referenced by that pointer from being moved
394 /// unless it implements [`Unpin`].
396 /// *See the [`pin` module] documentation for an explanation of pinning.*
398 /// [`pin` module]: self
400 // Note: the `Clone` derive below causes unsoundness as it's possible to implement
401 // `Clone` for mutable references.
402 // See <https://internals.rust-lang.org/t/unsoundness-in-pin/11311> for more details.
403 #[stable(feature = "pin", since = "1.33.0")]
407 #[derive(Copy, Clone)]
409 // FIXME(#93176): this field is made `#[unstable] #[doc(hidden)] pub` to:
410 // - deter downstream users from accessing it (which would be unsound!),
411 // - let the `pin!` macro access it (such a macro requires using struct
412 // literal syntax in order to benefit from lifetime extension).
413 // Long-term, `unsafe` fields or macro hygiene are expected to offer more robust alternatives.
414 #[unstable(feature = "unsafe_pin_internals", issue = "none")]
419 // The following implementations aren't derived in order to avoid soundness
420 // issues. `&self.pointer` should not be accessible to untrusted trait
423 // See <https://internals.rust-lang.org/t/unsoundness-in-pin/11311/73> for more details.
425 #[stable(feature = "pin_trait_impls", since = "1.41.0")]
426 impl<P: Deref, Q: Deref> PartialEq<Pin<Q>> for Pin<P>
428 P::Target: PartialEq<Q::Target>,
430 fn eq(&self, other: &Pin<Q>) -> bool {
431 P::Target::eq(self, other)
434 fn ne(&self, other: &Pin<Q>) -> bool {
435 P::Target::ne(self, other)
439 #[stable(feature = "pin_trait_impls", since = "1.41.0")]
440 impl<P: Deref<Target: Eq>> Eq for Pin<P> {}
442 #[stable(feature = "pin_trait_impls", since = "1.41.0")]
443 impl<P: Deref, Q: Deref> PartialOrd<Pin<Q>> for Pin<P>
445 P::Target: PartialOrd<Q::Target>,
447 fn partial_cmp(&self, other: &Pin<Q>) -> Option<cmp::Ordering> {
448 P::Target::partial_cmp(self, other)
451 fn lt(&self, other: &Pin<Q>) -> bool {
452 P::Target::lt(self, other)
455 fn le(&self, other: &Pin<Q>) -> bool {
456 P::Target::le(self, other)
459 fn gt(&self, other: &Pin<Q>) -> bool {
460 P::Target::gt(self, other)
463 fn ge(&self, other: &Pin<Q>) -> bool {
464 P::Target::ge(self, other)
468 #[stable(feature = "pin_trait_impls", since = "1.41.0")]
469 impl<P: Deref<Target: Ord>> Ord for Pin<P> {
470 fn cmp(&self, other: &Self) -> cmp::Ordering {
471 P::Target::cmp(self, other)
475 #[stable(feature = "pin_trait_impls", since = "1.41.0")]
476 impl<P: Deref<Target: Hash>> Hash for Pin<P> {
477 fn hash<H: Hasher>(&self, state: &mut H) {
478 P::Target::hash(self, state);
482 impl<P: Deref<Target: Unpin>> Pin<P> {
483 /// Construct a new `Pin<P>` around a pointer to some data of a type that
484 /// implements [`Unpin`].
486 /// Unlike `Pin::new_unchecked`, this method is safe because the pointer
487 /// `P` dereferences to an [`Unpin`] type, which cancels the pinning guarantees.
489 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
490 #[stable(feature = "pin", since = "1.33.0")]
491 pub const fn new(pointer: P) -> Pin<P> {
492 // SAFETY: the value pointed to is `Unpin`, and so has no requirements
494 unsafe { Pin::new_unchecked(pointer) }
497 /// Unwraps this `Pin<P>` returning the underlying pointer.
499 /// This requires that the data inside this `Pin` is [`Unpin`] so that we
500 /// can ignore the pinning invariants when unwrapping it.
502 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
503 #[stable(feature = "pin_into_inner", since = "1.39.0")]
504 pub const fn into_inner(pin: Pin<P>) -> P {
509 impl<P: Deref> Pin<P> {
510 /// Construct a new `Pin<P>` around a reference to some data of a type that
511 /// may or may not implement `Unpin`.
513 /// If `pointer` dereferences to an `Unpin` type, `Pin::new` should be used
518 /// This constructor is unsafe because we cannot guarantee that the data
519 /// pointed to by `pointer` is pinned, meaning that the data will not be moved or
520 /// its storage invalidated until it gets dropped. If the constructed `Pin<P>` does
521 /// not guarantee that the data `P` points to is pinned, that is a violation of
522 /// the API contract and may lead to undefined behavior in later (safe) operations.
524 /// By using this method, you are making a promise about the `P::Deref` and
525 /// `P::DerefMut` implementations, if they exist. Most importantly, they
526 /// must not move out of their `self` arguments: `Pin::as_mut` and `Pin::as_ref`
527 /// will call `DerefMut::deref_mut` and `Deref::deref` *on the pinned pointer*
528 /// and expect these methods to uphold the pinning invariants.
529 /// Moreover, by calling this method you promise that the reference `P`
530 /// dereferences to will not be moved out of again; in particular, it
531 /// must not be possible to obtain a `&mut P::Target` and then
532 /// move out of that reference (using, for example [`mem::swap`]).
534 /// For example, calling `Pin::new_unchecked` on an `&'a mut T` is unsafe because
535 /// while you are able to pin it for the given lifetime `'a`, you have no control
536 /// over whether it is kept pinned once `'a` ends:
539 /// use std::pin::Pin;
541 /// fn move_pinned_ref<T>(mut a: T, mut b: T) {
543 /// let p: Pin<&mut T> = Pin::new_unchecked(&mut a);
544 /// // This should mean the pointee `a` can never move again.
546 /// mem::swap(&mut a, &mut b);
547 /// // The address of `a` changed to `b`'s stack slot, so `a` got moved even
548 /// // though we have previously pinned it! We have violated the pinning API contract.
551 /// A value, once pinned, must remain pinned forever (unless its type implements `Unpin`).
553 /// Similarly, calling `Pin::new_unchecked` on an `Rc<T>` is unsafe because there could be
554 /// aliases to the same data that are not subject to the pinning restrictions:
557 /// use std::pin::Pin;
559 /// fn move_pinned_rc<T>(mut x: Rc<T>) {
560 /// let pinned = unsafe { Pin::new_unchecked(Rc::clone(&x)) };
562 /// let p: Pin<&T> = pinned.as_ref();
563 /// // This should mean the pointee can never move again.
566 /// let content = Rc::get_mut(&mut x).unwrap();
567 /// // Now, if `x` was the only reference, we have a mutable reference to
568 /// // data that we pinned above, which we could use to move it as we have
569 /// // seen in the previous example. We have violated the pinning API contract.
573 /// [`mem::swap`]: crate::mem::swap
574 #[lang = "new_unchecked"]
576 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
577 #[stable(feature = "pin", since = "1.33.0")]
578 pub const unsafe fn new_unchecked(pointer: P) -> Pin<P> {
582 /// Gets a pinned shared reference from this pinned pointer.
584 /// This is a generic method to go from `&Pin<Pointer<T>>` to `Pin<&T>`.
585 /// It is safe because, as part of the contract of `Pin::new_unchecked`,
586 /// the pointee cannot move after `Pin<Pointer<T>>` got created.
587 /// "Malicious" implementations of `Pointer::Deref` are likewise
588 /// ruled out by the contract of `Pin::new_unchecked`.
589 #[stable(feature = "pin", since = "1.33.0")]
591 pub fn as_ref(&self) -> Pin<&P::Target> {
592 // SAFETY: see documentation on this function
593 unsafe { Pin::new_unchecked(&*self.pointer) }
596 /// Unwraps this `Pin<P>` returning the underlying pointer.
600 /// This function is unsafe. You must guarantee that you will continue to
601 /// treat the pointer `P` as pinned after you call this function, so that
602 /// the invariants on the `Pin` type can be upheld. If the code using the
603 /// resulting `P` does not continue to maintain the pinning invariants that
604 /// is a violation of the API contract and may lead to undefined behavior in
605 /// later (safe) operations.
607 /// If the underlying data is [`Unpin`], [`Pin::into_inner`] should be used
610 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
611 #[stable(feature = "pin_into_inner", since = "1.39.0")]
612 pub const unsafe fn into_inner_unchecked(pin: Pin<P>) -> P {
617 impl<P: DerefMut> Pin<P> {
618 /// Gets a pinned mutable reference from this pinned pointer.
620 /// This is a generic method to go from `&mut Pin<Pointer<T>>` to `Pin<&mut T>`.
621 /// It is safe because, as part of the contract of `Pin::new_unchecked`,
622 /// the pointee cannot move after `Pin<Pointer<T>>` got created.
623 /// "Malicious" implementations of `Pointer::DerefMut` are likewise
624 /// ruled out by the contract of `Pin::new_unchecked`.
626 /// This method is useful when doing multiple calls to functions that consume the pinned type.
631 /// use std::pin::Pin;
635 /// fn method(self: Pin<&mut Self>) {
639 /// fn call_method_twice(mut self: Pin<&mut Self>) {
640 /// // `method` consumes `self`, so reborrow the `Pin<&mut Self>` via `as_mut`.
641 /// self.as_mut().method();
642 /// self.as_mut().method();
646 #[stable(feature = "pin", since = "1.33.0")]
648 pub fn as_mut(&mut self) -> Pin<&mut P::Target> {
649 // SAFETY: see documentation on this function
650 unsafe { Pin::new_unchecked(&mut *self.pointer) }
653 /// Assigns a new value to the memory behind the pinned reference.
655 /// This overwrites pinned data, but that is okay: its destructor gets
656 /// run before being overwritten, so no pinning guarantee is violated.
657 #[stable(feature = "pin", since = "1.33.0")]
659 pub fn set(&mut self, value: P::Target)
663 *(self.pointer) = value;
667 impl<'a, T: ?Sized> Pin<&'a T> {
668 /// Constructs a new pin by mapping the interior value.
670 /// For example, if you wanted to get a `Pin` of a field of something,
671 /// you could use this to get access to that field in one line of code.
672 /// However, there are several gotchas with these "pinning projections";
673 /// see the [`pin` module] documentation for further details on that topic.
677 /// This function is unsafe. You must guarantee that the data you return
678 /// will not move so long as the argument value does not move (for example,
679 /// because it is one of the fields of that value), and also that you do
680 /// not move out of the argument you receive to the interior function.
682 /// [`pin` module]: self#projections-and-structural-pinning
683 #[stable(feature = "pin", since = "1.33.0")]
684 pub unsafe fn map_unchecked<U, F>(self, func: F) -> Pin<&'a U>
689 let pointer = &*self.pointer;
690 let new_pointer = func(pointer);
692 // SAFETY: the safety contract for `new_unchecked` must be
693 // upheld by the caller.
694 unsafe { Pin::new_unchecked(new_pointer) }
697 /// Gets a shared reference out of a pin.
699 /// This is safe because it is not possible to move out of a shared reference.
700 /// It may seem like there is an issue here with interior mutability: in fact,
701 /// it *is* possible to move a `T` out of a `&RefCell<T>`. However, this is
702 /// not a problem as long as there does not also exist a `Pin<&T>` pointing
703 /// to the same data, and `RefCell<T>` does not let you create a pinned reference
704 /// to its contents. See the discussion on ["pinning projections"] for further
707 /// Note: `Pin` also implements `Deref` to the target, which can be used
708 /// to access the inner value. However, `Deref` only provides a reference
709 /// that lives for as long as the borrow of the `Pin`, not the lifetime of
710 /// the `Pin` itself. This method allows turning the `Pin` into a reference
711 /// with the same lifetime as the original `Pin`.
713 /// ["pinning projections"]: self#projections-and-structural-pinning
716 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
717 #[stable(feature = "pin", since = "1.33.0")]
718 pub const fn get_ref(self) -> &'a T {
723 impl<'a, T: ?Sized> Pin<&'a mut T> {
724 /// Converts this `Pin<&mut T>` into a `Pin<&T>` with the same lifetime.
726 #[must_use = "`self` will be dropped if the result is not used"]
727 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
728 #[stable(feature = "pin", since = "1.33.0")]
729 pub const fn into_ref(self) -> Pin<&'a T> {
730 Pin { pointer: self.pointer }
733 /// Gets a mutable reference to the data inside of this `Pin`.
735 /// This requires that the data inside this `Pin` is `Unpin`.
737 /// Note: `Pin` also implements `DerefMut` to the data, which can be used
738 /// to access the inner value. However, `DerefMut` only provides a reference
739 /// that lives for as long as the borrow of the `Pin`, not the lifetime of
740 /// the `Pin` itself. This method allows turning the `Pin` into a reference
741 /// with the same lifetime as the original `Pin`.
743 #[must_use = "`self` will be dropped if the result is not used"]
744 #[stable(feature = "pin", since = "1.33.0")]
745 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
746 pub const fn get_mut(self) -> &'a mut T
753 /// Gets a mutable reference to the data inside of this `Pin`.
757 /// This function is unsafe. You must guarantee that you will never move
758 /// the data out of the mutable reference you receive when you call this
759 /// function, so that the invariants on the `Pin` type can be upheld.
761 /// If the underlying data is `Unpin`, `Pin::get_mut` should be used
764 #[must_use = "`self` will be dropped if the result is not used"]
765 #[stable(feature = "pin", since = "1.33.0")]
766 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
767 pub const unsafe fn get_unchecked_mut(self) -> &'a mut T {
771 /// Construct a new pin by mapping the interior value.
773 /// For example, if you wanted to get a `Pin` of a field of something,
774 /// you could use this to get access to that field in one line of code.
775 /// However, there are several gotchas with these "pinning projections";
776 /// see the [`pin` module] documentation for further details on that topic.
780 /// This function is unsafe. You must guarantee that the data you return
781 /// will not move so long as the argument value does not move (for example,
782 /// because it is one of the fields of that value), and also that you do
783 /// not move out of the argument you receive to the interior function.
785 /// [`pin` module]: self#projections-and-structural-pinning
786 #[must_use = "`self` will be dropped if the result is not used"]
787 #[stable(feature = "pin", since = "1.33.0")]
788 pub unsafe fn map_unchecked_mut<U, F>(self, func: F) -> Pin<&'a mut U>
791 F: FnOnce(&mut T) -> &mut U,
793 // SAFETY: the caller is responsible for not moving the
794 // value out of this reference.
795 let pointer = unsafe { Pin::get_unchecked_mut(self) };
796 let new_pointer = func(pointer);
797 // SAFETY: as the value of `this` is guaranteed to not have
798 // been moved out, this call to `new_unchecked` is safe.
799 unsafe { Pin::new_unchecked(new_pointer) }
803 impl<T: ?Sized> Pin<&'static T> {
804 /// Get a pinned reference from a static reference.
806 /// This is safe, because `T` is borrowed for the `'static` lifetime, which
808 #[stable(feature = "pin_static_ref", since = "1.61.0")]
809 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
810 pub const fn static_ref(r: &'static T) -> Pin<&'static T> {
811 // SAFETY: The 'static borrow guarantees the data will not be
812 // moved/invalidated until it gets dropped (which is never).
813 unsafe { Pin::new_unchecked(r) }
817 impl<'a, P: DerefMut> Pin<&'a mut Pin<P>> {
818 /// Gets a pinned mutable reference from this nested pinned pointer.
820 /// This is a generic method to go from `Pin<&mut Pin<Pointer<T>>>` to `Pin<&mut T>`. It is
821 /// safe because the existence of a `Pin<Pointer<T>>` ensures that the pointee, `T`, cannot
822 /// move in the future, and this method does not enable the pointee to move. "Malicious"
823 /// implementations of `P::DerefMut` are likewise ruled out by the contract of
824 /// `Pin::new_unchecked`.
825 #[unstable(feature = "pin_deref_mut", issue = "86918")]
826 #[must_use = "`self` will be dropped if the result is not used"]
828 pub fn as_deref_mut(self) -> Pin<&'a mut P::Target> {
829 // SAFETY: What we're asserting here is that going from
835 // Pin<&mut P::Target>
839 // We need to ensure that two things hold for that to be the case:
841 // 1) Once we give out a `Pin<&mut P::Target>`, an `&mut P::Target` will not be given out.
842 // 2) By giving out a `Pin<&mut P::Target>`, we do not risk of violating `Pin<&mut Pin<P>>`
844 // The existence of `Pin<P>` is sufficient to guarantee #1: since we already have a
845 // `Pin<P>`, it must already uphold the pinning guarantees, which must mean that
846 // `Pin<&mut P::Target>` does as well, since `Pin::as_mut` is safe. We do not have to rely
847 // on the fact that P is _also_ pinned.
849 // For #2, we need to ensure that code given a `Pin<&mut P::Target>` cannot cause the
850 // `Pin<P>` to move? That is not possible, since `Pin<&mut P::Target>` no longer retains
851 // any access to the `P` itself, much less the `Pin<P>`.
852 unsafe { self.get_unchecked_mut() }.as_mut()
856 impl<T: ?Sized> Pin<&'static mut T> {
857 /// Get a pinned mutable reference from a static mutable reference.
859 /// This is safe, because `T` is borrowed for the `'static` lifetime, which
861 #[stable(feature = "pin_static_ref", since = "1.61.0")]
862 #[rustc_const_unstable(feature = "const_pin", issue = "76654")]
863 pub const fn static_mut(r: &'static mut T) -> Pin<&'static mut T> {
864 // SAFETY: The 'static borrow guarantees the data will not be
865 // moved/invalidated until it gets dropped (which is never).
866 unsafe { Pin::new_unchecked(r) }
870 #[stable(feature = "pin", since = "1.33.0")]
871 impl<P: Deref> Deref for Pin<P> {
872 type Target = P::Target;
873 fn deref(&self) -> &P::Target {
874 Pin::get_ref(Pin::as_ref(self))
878 #[stable(feature = "pin", since = "1.33.0")]
879 impl<P: DerefMut<Target: Unpin>> DerefMut for Pin<P> {
880 fn deref_mut(&mut self) -> &mut P::Target {
881 Pin::get_mut(Pin::as_mut(self))
885 #[unstable(feature = "receiver_trait", issue = "none")]
886 impl<P: Receiver> Receiver for Pin<P> {}
888 #[stable(feature = "pin", since = "1.33.0")]
889 impl<P: fmt::Debug> fmt::Debug for Pin<P> {
890 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
891 fmt::Debug::fmt(&self.pointer, f)
895 #[stable(feature = "pin", since = "1.33.0")]
896 impl<P: fmt::Display> fmt::Display for Pin<P> {
897 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
898 fmt::Display::fmt(&self.pointer, f)
902 #[stable(feature = "pin", since = "1.33.0")]
903 impl<P: fmt::Pointer> fmt::Pointer for Pin<P> {
904 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
905 fmt::Pointer::fmt(&self.pointer, f)
909 // Note: this means that any impl of `CoerceUnsized` that allows coercing from
910 // a type that impls `Deref<Target=impl !Unpin>` to a type that impls
911 // `Deref<Target=Unpin>` is unsound. Any such impl would probably be unsound
912 // for other reasons, though, so we just need to take care not to allow such
913 // impls to land in std.
914 #[stable(feature = "pin", since = "1.33.0")]
915 impl<P, U> CoerceUnsized<Pin<U>> for Pin<P> where P: CoerceUnsized<U> {}
917 #[stable(feature = "pin", since = "1.33.0")]
918 impl<P, U> DispatchFromDyn<Pin<U>> for Pin<P> where P: DispatchFromDyn<U> {}
920 /// Constructs a <code>[Pin]<[&mut] T></code>, by pinning[^1] a `value: T` _locally_[^2].
922 /// Unlike [`Box::pin`], this does not involve a heap allocation.
924 /// [^1]: If the (type `T` of the) given value does not implement [`Unpin`], then this
925 /// effectively pins the `value` in memory, where it will be unable to be moved.
926 /// Otherwise, <code>[Pin]<[&mut] T></code> behaves like <code>[&mut] T</code>, and operations such
927 /// as [`mem::replace()`][crate::mem::replace] will allow extracting that value, and therefore,
929 /// See [the `Unpin` section of the `pin` module][self#unpin] for more info.
931 /// [^2]: This is usually dubbed "stack"-pinning. And whilst local values are almost always located
932 /// in the stack (_e.g._, when within the body of a non-`async` function), the truth is that inside
933 /// the body of an `async fn` or block —more generally, the body of a generator— any locals crossing
934 /// an `.await` point —a `yield` point— end up being part of the state captured by the `Future` —by
935 /// the `Generator`—, and thus will be stored wherever that one is.
942 /// #![feature(pin_macro)]
943 /// # use core::marker::PhantomPinned as Foo;
944 /// use core::pin::{pin, Pin};
946 /// fn stuff(foo: Pin<&mut Foo>) {
951 /// let pinned_foo = pin!(Foo { /* … */ });
952 /// stuff(pinned_foo);
954 /// stuff(pin!(Foo { /* … */ }));
957 /// ### Manually polling a `Future` (without `Unpin` bounds)
960 /// #![feature(pin_macro)]
964 /// task::{Context, Poll},
967 /// # use std::{sync::Arc, task::Wake, thread::Thread};
969 /// # /// A waker that wakes up the current thread when called.
970 /// # struct ThreadWaker(Thread);
972 /// # impl Wake for ThreadWaker {
973 /// # fn wake(self: Arc<Self>) {
974 /// # self.0.unpark();
978 /// /// Runs a future to completion.
979 /// fn block_on<Fut: Future>(fut: Fut) -> Fut::Output {
980 /// let waker_that_unparks_thread = // …
981 /// # Arc::new(ThreadWaker(thread::current())).into();
982 /// let mut cx = Context::from_waker(&waker_that_unparks_thread);
983 /// // Pin the future so it can be polled.
984 /// let mut pinned_fut = pin!(fut);
986 /// match pinned_fut.as_mut().poll(&mut cx) {
987 /// Poll::Pending => thread::park(),
988 /// Poll::Ready(res) => return res,
993 /// # assert_eq!(42, block_on(async { 42 }));
996 /// ### With `Generator`s
999 /// #![feature(generators, generator_trait, pin_macro)]
1001 /// ops::{Generator, GeneratorState},
1005 /// fn generator_fn() -> impl Generator<Yield = usize, Return = ()> /* not Unpin */ {
1006 /// // Allow generator to be self-referential (not `Unpin`)
1007 /// // vvvvvv so that locals can cross yield points.
1009 /// let foo = String::from("foo");
1010 /// let foo_ref = &foo; // ------+
1011 /// yield 0; // | <- crosses yield point!
1012 /// println!("{foo_ref}"); // <--+
1013 /// yield foo.len();
1018 /// let mut generator = pin!(generator_fn());
1019 /// match generator.as_mut().resume(()) {
1020 /// GeneratorState::Yielded(0) => {},
1021 /// _ => unreachable!(),
1023 /// match generator.as_mut().resume(()) {
1024 /// GeneratorState::Yielded(3) => {},
1025 /// _ => unreachable!(),
1027 /// match generator.resume(()) {
1028 /// GeneratorState::Yielded(_) => unreachable!(),
1029 /// GeneratorState::Complete(()) => {},
1036 /// Precisely because a value is pinned to local storage, the resulting <code>[Pin]<[&mut] T></code>
1037 /// reference ends up borrowing a local tied to that block: it can't escape it.
1039 /// The following, for instance, fails to compile:
1041 /// ```rust,compile_fail
1042 /// #![feature(pin_macro)]
1043 /// use core::pin::{pin, Pin};
1044 /// # use core::{marker::PhantomPinned as Foo, mem::drop as stuff};
1046 /// let x: Pin<&mut Foo> = {
1047 /// let x: Pin<&mut Foo> = pin!(Foo { /* … */ });
1049 /// }; // <- Foo is dropped
1050 /// stuff(x); // Error: use of dropped value
1053 /// <details><summary>Error message</summary>
1056 /// error[E0716]: temporary value dropped while borrowed
1057 /// --> src/main.rs:9:28
1059 /// 8 | let x: Pin<&mut Foo> = {
1060 /// | - borrow later stored here
1061 /// 9 | let x: Pin<&mut Foo> = pin!(Foo { /* … */ });
1062 /// | ^^^^^^^^^^^^^^^^^^^^^ creates a temporary value which is freed while still in use
1064 /// 11 | }; // <- Foo is dropped
1065 /// | - temporary value is freed at the end of this statement
1067 /// = note: consider using a `let` binding to create a longer lived value
1072 /// This makes [`pin!`] **unsuitable to pin values when intending to _return_ them**. Instead, the
1073 /// value is expected to be passed around _unpinned_ until the point where it is to be consumed,
1074 /// where it is then useful and even sensible to pin the value locally using [`pin!`].
1076 /// If you really need to return a pinned value, consider using [`Box::pin`] instead.
1078 /// On the other hand, pinning to the stack[<sup>2</sup>](#fn2) using [`pin!`] is likely to be
1079 /// cheaper than pinning into a fresh heap allocation using [`Box::pin`]. Moreover, by virtue of not
1080 /// even needing an allocator, [`pin!`] is the main non-`unsafe` `#![no_std]`-compatible [`Pin`]
1083 /// [`Box::pin`]: ../../std/boxed/struct.Box.html#method.pin
1084 #[unstable(feature = "pin_macro", issue = "93178")]
1085 #[rustc_macro_transparency = "semitransparent"]
1086 #[allow_internal_unstable(unsafe_pin_internals)]
1087 pub macro pin($value:expr $(,)?) {
1088 // This is `Pin::new_unchecked(&mut { $value })`, so, for starters, let's
1089 // review such a hypothetical macro (that any user-code could define):
1092 // macro_rules! pin {( $value:expr ) => (
1093 // match &mut { $value } { at_value => unsafe { // Do not wrap `$value` in an `unsafe` block.
1094 // $crate::pin::Pin::<&mut _>::new_unchecked(at_value)
1100 // - `type P = &mut _`. There are thus no pathological `Deref{,Mut}` impls
1101 // that would break `Pin`'s invariants.
1102 // - `{ $value }` is braced, making it a _block expression_, thus **moving**
1103 // the given `$value`, and making it _become an **anonymous** temporary_.
1104 // By virtue of being anonymous, it can no longer be accessed, thus
1105 // preventing any attempts to `mem::replace` it or `mem::forget` it, _etc._
1107 // This gives us a `pin!` definition that is sound, and which works, but only
1108 // in certain scenarios:
1109 // - If the `pin!(value)` expression is _directly_ fed to a function call:
1110 // `let poll = pin!(fut).poll(cx);`
1111 // - If the `pin!(value)` expression is part of a scrutinee:
1113 // match pin!(fut) { pinned_fut => {
1114 // pinned_fut.as_mut().poll(...);
1115 // pinned_fut.as_mut().poll(...);
1116 // }} // <- `fut` is dropped here.
1118 // Alas, it doesn't work for the more straight-forward use-case: `let` bindings.
1120 // let pinned_fut = pin!(fut); // <- temporary value is freed at the end of this statement
1121 // pinned_fut.poll(...) // error[E0716]: temporary value dropped while borrowed
1122 // // note: consider using a `let` binding to create a longer lived value
1124 // - Issues such as this one are the ones motivating https://github.com/rust-lang/rfcs/pull/66
1126 // This makes such a macro incredibly unergonomic in practice, and the reason most macros
1127 // out there had to take the path of being a statement/binding macro (_e.g._, `pin!(future);`)
1128 // instead of featuring the more intuitive ergonomics of an expression macro.
1130 // Luckily, there is a way to avoid the problem. Indeed, the problem stems from the fact that a
1131 // temporary is dropped at the end of its enclosing statement when it is part of the parameters
1132 // given to function call, which has precisely been the case with our `Pin::new_unchecked()`!
1135 // let p = Pin::new_unchecked(&mut <temporary>);
1139 // let p = { let mut anon = <temporary>; &mut anon };
1142 // However, when using a literal braced struct to construct the value, references to temporaries
1143 // can then be taken. This makes Rust change the lifespan of such temporaries so that they are,
1144 // instead, dropped _at the end of the enscoping block_.
1147 // let p = Pin { pointer: &mut <temporary> };
1151 // let mut anon = <temporary>;
1152 // let p = Pin { pointer: &mut anon };
1154 // which is *exactly* what we want.
1156 // See https://doc.rust-lang.org/1.58.1/reference/destructors.html#temporary-lifetime-extension
1158 $crate::pin::Pin::<&mut _> { pointer: &mut { $value } }