1 | //! A contiguous growable array type with heap-allocated contents, written |
2 | //! `Vec<T>`. |
3 | //! |
4 | //! Vectors have *O*(1) indexing, amortized *O*(1) push (to the end) and |
5 | //! *O*(1) pop (from the end). |
6 | //! |
7 | //! Vectors ensure they never allocate more than `isize::MAX` bytes. |
8 | //! |
9 | //! # Examples |
10 | //! |
11 | //! You can explicitly create a [`Vec`] with [`Vec::new`]: |
12 | //! |
13 | //! ``` |
14 | //! let v: Vec<i32> = Vec::new(); |
15 | //! ``` |
16 | //! |
17 | //! ...or by using the [`vec!`] macro: |
18 | //! |
19 | //! ``` |
20 | //! let v: Vec<i32> = vec![]; |
21 | //! |
22 | //! let v = vec![1, 2, 3, 4, 5]; |
23 | //! |
24 | //! let v = vec![0; 10]; // ten zeroes |
25 | //! ``` |
26 | //! |
27 | //! You can [`push`] values onto the end of a vector (which will grow the vector |
28 | //! as needed): |
29 | //! |
30 | //! ``` |
31 | //! let mut v = vec![1, 2]; |
32 | //! |
33 | //! v.push(3); |
34 | //! ``` |
35 | //! |
36 | //! Popping values works in much the same way: |
37 | //! |
38 | //! ``` |
39 | //! let mut v = vec![1, 2]; |
40 | //! |
41 | //! let two = v.pop(); |
42 | //! ``` |
43 | //! |
44 | //! Vectors also support indexing (through the [`Index`] and [`IndexMut`] traits): |
45 | //! |
46 | //! ``` |
47 | //! let mut v = vec![1, 2, 3]; |
48 | //! let three = v[2]; |
49 | //! v[1] = v[1] + 5; |
50 | //! ``` |
51 | //! |
52 | //! [`push`]: Vec::push |
53 | |
54 | #[cfg (not(no_global_oom_handling))] |
55 | use core::cmp; |
56 | use core::cmp::Ordering; |
57 | use core::convert::TryFrom; |
58 | use core::fmt; |
59 | use core::hash::{Hash, Hasher}; |
60 | #[cfg (not(no_global_oom_handling))] |
61 | use core::iter; |
62 | #[cfg (not(no_global_oom_handling))] |
63 | use core::iter::FromIterator; |
64 | use core::marker::PhantomData; |
65 | use core::mem::{self, size_of, ManuallyDrop, MaybeUninit}; |
66 | use core::ops::{self, Bound, Index, IndexMut, Range, RangeBounds}; |
67 | use core::ptr::{self, NonNull}; |
68 | use core::slice::{self, SliceIndex}; |
69 | |
70 | use super::{ |
71 | alloc::{Allocator, Global}, |
72 | assume, |
73 | boxed::Box, |
74 | raw_vec::{RawVec, TryReserveError}, |
75 | }; |
76 | |
77 | #[cfg (not(no_global_oom_handling))] |
78 | pub use self::splice::Splice; |
79 | |
80 | #[cfg (not(no_global_oom_handling))] |
81 | mod splice; |
82 | |
83 | pub use self::drain::Drain; |
84 | |
85 | mod drain; |
86 | |
87 | pub use self::into_iter::IntoIter; |
88 | |
89 | mod into_iter; |
90 | |
91 | mod partial_eq; |
92 | |
93 | #[cfg (not(no_global_oom_handling))] |
94 | mod set_len_on_drop; |
95 | |
96 | #[cfg (not(no_global_oom_handling))] |
97 | use self::set_len_on_drop::SetLenOnDrop; |
98 | |
99 | /// A contiguous growable array type, written as `Vec<T>`, short for 'vector'. |
100 | /// |
101 | /// # Examples |
102 | /// |
103 | /// ``` |
104 | /// let mut vec = Vec::new(); |
105 | /// vec.push(1); |
106 | /// vec.push(2); |
107 | /// |
108 | /// assert_eq!(vec.len(), 2); |
109 | /// assert_eq!(vec[0], 1); |
110 | /// |
111 | /// assert_eq!(vec.pop(), Some(2)); |
112 | /// assert_eq!(vec.len(), 1); |
113 | /// |
114 | /// vec[0] = 7; |
115 | /// assert_eq!(vec[0], 7); |
116 | /// |
117 | /// vec.extend([1, 2, 3].iter().copied()); |
118 | /// |
119 | /// for x in &vec { |
120 | /// println!("{x}" ); |
121 | /// } |
122 | /// assert_eq!(vec, [7, 1, 2, 3]); |
123 | /// ``` |
124 | /// |
125 | /// The [`vec!`] macro is provided for convenient initialization: |
126 | /// |
127 | /// ``` |
128 | /// let mut vec1 = vec![1, 2, 3]; |
129 | /// vec1.push(4); |
130 | /// let vec2 = Vec::from([1, 2, 3, 4]); |
131 | /// assert_eq!(vec1, vec2); |
132 | /// ``` |
133 | /// |
134 | /// It can also initialize each element of a `Vec<T>` with a given value. |
135 | /// This may be more efficient than performing allocation and initialization |
136 | /// in separate steps, especially when initializing a vector of zeros: |
137 | /// |
138 | /// ``` |
139 | /// let vec = vec![0; 5]; |
140 | /// assert_eq!(vec, [0, 0, 0, 0, 0]); |
141 | /// |
142 | /// // The following is equivalent, but potentially slower: |
143 | /// let mut vec = Vec::with_capacity(5); |
144 | /// vec.resize(5, 0); |
145 | /// assert_eq!(vec, [0, 0, 0, 0, 0]); |
146 | /// ``` |
147 | /// |
148 | /// For more information, see |
149 | /// [Capacity and Reallocation](#capacity-and-reallocation). |
150 | /// |
151 | /// Use a `Vec<T>` as an efficient stack: |
152 | /// |
153 | /// ``` |
154 | /// let mut stack = Vec::new(); |
155 | /// |
156 | /// stack.push(1); |
157 | /// stack.push(2); |
158 | /// stack.push(3); |
159 | /// |
160 | /// while let Some(top) = stack.pop() { |
161 | /// // Prints 3, 2, 1 |
162 | /// println!("{top}" ); |
163 | /// } |
164 | /// ``` |
165 | /// |
166 | /// # Indexing |
167 | /// |
168 | /// The `Vec` type allows to access values by index, because it implements the |
169 | /// [`Index`] trait. An example will be more explicit: |
170 | /// |
171 | /// ``` |
172 | /// let v = vec![0, 2, 4, 6]; |
173 | /// println!("{}" , v[1]); // it will display '2' |
174 | /// ``` |
175 | /// |
176 | /// However be careful: if you try to access an index which isn't in the `Vec`, |
177 | /// your software will panic! You cannot do this: |
178 | /// |
179 | /// ```should_panic |
180 | /// let v = vec![0, 2, 4, 6]; |
181 | /// println!("{}" , v[6]); // it will panic! |
182 | /// ``` |
183 | /// |
184 | /// Use [`get`] and [`get_mut`] if you want to check whether the index is in |
185 | /// the `Vec`. |
186 | /// |
187 | /// # Slicing |
188 | /// |
189 | /// A `Vec` can be mutable. On the other hand, slices are read-only objects. |
190 | /// To get a [slice][prim@slice], use [`&`]. Example: |
191 | /// |
192 | /// ``` |
193 | /// fn read_slice(slice: &[usize]) { |
194 | /// // ... |
195 | /// } |
196 | /// |
197 | /// let v = vec![0, 1]; |
198 | /// read_slice(&v); |
199 | /// |
200 | /// // ... and that's all! |
201 | /// // you can also do it like this: |
202 | /// let u: &[usize] = &v; |
203 | /// // or like this: |
204 | /// let u: &[_] = &v; |
205 | /// ``` |
206 | /// |
207 | /// In Rust, it's more common to pass slices as arguments rather than vectors |
208 | /// when you just want to provide read access. The same goes for [`String`] and |
209 | /// [`&str`]. |
210 | /// |
211 | /// # Capacity and reallocation |
212 | /// |
213 | /// The capacity of a vector is the amount of space allocated for any future |
214 | /// elements that will be added onto the vector. This is not to be confused with |
215 | /// the *length* of a vector, which specifies the number of actual elements |
216 | /// within the vector. If a vector's length exceeds its capacity, its capacity |
217 | /// will automatically be increased, but its elements will have to be |
218 | /// reallocated. |
219 | /// |
220 | /// For example, a vector with capacity 10 and length 0 would be an empty vector |
221 | /// with space for 10 more elements. Pushing 10 or fewer elements onto the |
222 | /// vector will not change its capacity or cause reallocation to occur. However, |
223 | /// if the vector's length is increased to 11, it will have to reallocate, which |
224 | /// can be slow. For this reason, it is recommended to use [`Vec::with_capacity`] |
225 | /// whenever possible to specify how big the vector is expected to get. |
226 | /// |
227 | /// # Guarantees |
228 | /// |
229 | /// Due to its incredibly fundamental nature, `Vec` makes a lot of guarantees |
230 | /// about its design. This ensures that it's as low-overhead as possible in |
231 | /// the general case, and can be correctly manipulated in primitive ways |
232 | /// by unsafe code. Note that these guarantees refer to an unqualified `Vec<T>`. |
233 | /// If additional type parameters are added (e.g., to support custom allocators), |
234 | /// overriding their defaults may change the behavior. |
235 | /// |
236 | /// Most fundamentally, `Vec` is and always will be a (pointer, capacity, length) |
237 | /// triplet. No more, no less. The order of these fields is completely |
238 | /// unspecified, and you should use the appropriate methods to modify these. |
239 | /// The pointer will never be null, so this type is null-pointer-optimized. |
240 | /// |
241 | /// However, the pointer might not actually point to allocated memory. In particular, |
242 | /// if you construct a `Vec` with capacity 0 via [`Vec::new`], [`vec![]`][`vec!`], |
243 | /// [`Vec::with_capacity(0)`][`Vec::with_capacity`], or by calling [`shrink_to_fit`] |
244 | /// on an empty Vec, it will not allocate memory. Similarly, if you store zero-sized |
245 | /// types inside a `Vec`, it will not allocate space for them. *Note that in this case |
246 | /// the `Vec` might not report a [`capacity`] of 0*. `Vec` will allocate if and only |
247 | /// if <code>[mem::size_of::\<T>]\() * [capacity]\() > 0</code>. In general, `Vec`'s allocation |
248 | /// details are very subtle --- if you intend to allocate memory using a `Vec` |
249 | /// and use it for something else (either to pass to unsafe code, or to build your |
250 | /// own memory-backed collection), be sure to deallocate this memory by using |
251 | /// `from_raw_parts` to recover the `Vec` and then dropping it. |
252 | /// |
253 | /// If a `Vec` *has* allocated memory, then the memory it points to is on the heap |
254 | /// (as defined by the allocator Rust is configured to use by default), and its |
255 | /// pointer points to [`len`] initialized, contiguous elements in order (what |
256 | /// you would see if you coerced it to a slice), followed by <code>[capacity] - [len]</code> |
257 | /// logically uninitialized, contiguous elements. |
258 | /// |
259 | /// A vector containing the elements `'a'` and `'b'` with capacity 4 can be |
260 | /// visualized as below. The top part is the `Vec` struct, it contains a |
261 | /// pointer to the head of the allocation in the heap, length and capacity. |
262 | /// The bottom part is the allocation on the heap, a contiguous memory block. |
263 | /// |
264 | /// ```text |
265 | /// ptr len capacity |
266 | /// +--------+--------+--------+ |
267 | /// | 0x0123 | 2 | 4 | |
268 | /// +--------+--------+--------+ |
269 | /// | |
270 | /// v |
271 | /// Heap +--------+--------+--------+--------+ |
272 | /// | 'a' | 'b' | uninit | uninit | |
273 | /// +--------+--------+--------+--------+ |
274 | /// ``` |
275 | /// |
276 | /// - **uninit** represents memory that is not initialized, see [`MaybeUninit`]. |
277 | /// - Note: the ABI is not stable and `Vec` makes no guarantees about its memory |
278 | /// layout (including the order of fields). |
279 | /// |
280 | /// `Vec` will never perform a "small optimization" where elements are actually |
281 | /// stored on the stack for two reasons: |
282 | /// |
283 | /// * It would make it more difficult for unsafe code to correctly manipulate |
284 | /// a `Vec`. The contents of a `Vec` wouldn't have a stable address if it were |
285 | /// only moved, and it would be more difficult to determine if a `Vec` had |
286 | /// actually allocated memory. |
287 | /// |
288 | /// * It would penalize the general case, incurring an additional branch |
289 | /// on every access. |
290 | /// |
291 | /// `Vec` will never automatically shrink itself, even if completely empty. This |
292 | /// ensures no unnecessary allocations or deallocations occur. Emptying a `Vec` |
293 | /// and then filling it back up to the same [`len`] should incur no calls to |
294 | /// the allocator. If you wish to free up unused memory, use |
295 | /// [`shrink_to_fit`] or [`shrink_to`]. |
296 | /// |
297 | /// [`push`] and [`insert`] will never (re)allocate if the reported capacity is |
298 | /// sufficient. [`push`] and [`insert`] *will* (re)allocate if |
299 | /// <code>[len] == [capacity]</code>. That is, the reported capacity is completely |
300 | /// accurate, and can be relied on. It can even be used to manually free the memory |
301 | /// allocated by a `Vec` if desired. Bulk insertion methods *may* reallocate, even |
302 | /// when not necessary. |
303 | /// |
304 | /// `Vec` does not guarantee any particular growth strategy when reallocating |
305 | /// when full, nor when [`reserve`] is called. The current strategy is basic |
306 | /// and it may prove desirable to use a non-constant growth factor. Whatever |
307 | /// strategy is used will of course guarantee *O*(1) amortized [`push`]. |
308 | /// |
309 | /// `vec![x; n]`, `vec![a, b, c, d]`, and |
310 | /// [`Vec::with_capacity(n)`][`Vec::with_capacity`], will all produce a `Vec` |
311 | /// with exactly the requested capacity. If <code>[len] == [capacity]</code>, |
312 | /// (as is the case for the [`vec!`] macro), then a `Vec<T>` can be converted to |
313 | /// and from a [`Box<[T]>`][owned slice] without reallocating or moving the elements. |
314 | /// |
315 | /// `Vec` will not specifically overwrite any data that is removed from it, |
316 | /// but also won't specifically preserve it. Its uninitialized memory is |
317 | /// scratch space that it may use however it wants. It will generally just do |
318 | /// whatever is most efficient or otherwise easy to implement. Do not rely on |
319 | /// removed data to be erased for security purposes. Even if you drop a `Vec`, its |
320 | /// buffer may simply be reused by another allocation. Even if you zero a `Vec`'s memory |
321 | /// first, that might not actually happen because the optimizer does not consider |
322 | /// this a side-effect that must be preserved. There is one case which we will |
323 | /// not break, however: using `unsafe` code to write to the excess capacity, |
324 | /// and then increasing the length to match, is always valid. |
325 | /// |
326 | /// Currently, `Vec` does not guarantee the order in which elements are dropped. |
327 | /// The order has changed in the past and may change again. |
328 | /// |
329 | /// [`get`]: ../../std/vec/struct.Vec.html#method.get |
330 | /// [`get_mut`]: ../../std/vec/struct.Vec.html#method.get_mut |
331 | /// [`String`]: alloc_crate::string::String |
332 | /// [`&str`]: type@str |
333 | /// [`shrink_to_fit`]: Vec::shrink_to_fit |
334 | /// [`shrink_to`]: Vec::shrink_to |
335 | /// [capacity]: Vec::capacity |
336 | /// [`capacity`]: Vec::capacity |
337 | /// [mem::size_of::\<T>]: core::mem::size_of |
338 | /// [len]: Vec::len |
339 | /// [`len`]: Vec::len |
340 | /// [`push`]: Vec::push |
341 | /// [`insert`]: Vec::insert |
342 | /// [`reserve`]: Vec::reserve |
343 | /// [`MaybeUninit`]: core::mem::MaybeUninit |
344 | /// [owned slice]: Box |
345 | pub struct Vec<T, A: Allocator = Global> { |
346 | buf: RawVec<T, A>, |
347 | len: usize, |
348 | } |
349 | |
350 | //////////////////////////////////////////////////////////////////////////////// |
351 | // Inherent methods |
352 | //////////////////////////////////////////////////////////////////////////////// |
353 | |
354 | impl<T> Vec<T> { |
355 | /// Constructs a new, empty `Vec<T>`. |
356 | /// |
357 | /// The vector will not allocate until elements are pushed onto it. |
358 | /// |
359 | /// # Examples |
360 | /// |
361 | /// ``` |
362 | /// # #![allow (unused_mut)] |
363 | /// let mut vec: Vec<i32> = Vec::new(); |
364 | /// ``` |
365 | #[inline (always)] |
366 | #[must_use ] |
367 | pub const fn new() -> Self { |
368 | Vec { |
369 | buf: RawVec::new(), |
370 | len: 0, |
371 | } |
372 | } |
373 | |
374 | /// Constructs a new, empty `Vec<T>` with at least the specified capacity. |
375 | /// |
376 | /// The vector will be able to hold at least `capacity` elements without |
377 | /// reallocating. This method is allowed to allocate for more elements than |
378 | /// `capacity`. If `capacity` is 0, the vector will not allocate. |
379 | /// |
380 | /// It is important to note that although the returned vector has the |
381 | /// minimum *capacity* specified, the vector will have a zero *length*. For |
382 | /// an explanation of the difference between length and capacity, see |
383 | /// *[Capacity and reallocation]*. |
384 | /// |
385 | /// If it is important to know the exact allocated capacity of a `Vec`, |
386 | /// always use the [`capacity`] method after construction. |
387 | /// |
388 | /// For `Vec<T>` where `T` is a zero-sized type, there will be no allocation |
389 | /// and the capacity will always be `usize::MAX`. |
390 | /// |
391 | /// [Capacity and reallocation]: #capacity-and-reallocation |
392 | /// [`capacity`]: Vec::capacity |
393 | /// |
394 | /// # Panics |
395 | /// |
396 | /// Panics if the new capacity exceeds `isize::MAX` bytes. |
397 | /// |
398 | /// # Examples |
399 | /// |
400 | /// ``` |
401 | /// let mut vec = Vec::with_capacity(10); |
402 | /// |
403 | /// // The vector contains no items, even though it has capacity for more |
404 | /// assert_eq!(vec.len(), 0); |
405 | /// assert!(vec.capacity() >= 10); |
406 | /// |
407 | /// // These are all done without reallocating... |
408 | /// for i in 0..10 { |
409 | /// vec.push(i); |
410 | /// } |
411 | /// assert_eq!(vec.len(), 10); |
412 | /// assert!(vec.capacity() >= 10); |
413 | /// |
414 | /// // ...but this may make the vector reallocate |
415 | /// vec.push(11); |
416 | /// assert_eq!(vec.len(), 11); |
417 | /// assert!(vec.capacity() >= 11); |
418 | /// |
419 | /// // A vector of a zero-sized type will always over-allocate, since no |
420 | /// // allocation is necessary |
421 | /// let vec_units = Vec::<()>::with_capacity(10); |
422 | /// assert_eq!(vec_units.capacity(), usize::MAX); |
423 | /// ``` |
424 | #[cfg (not(no_global_oom_handling))] |
425 | #[inline (always)] |
426 | #[must_use ] |
427 | pub fn with_capacity(capacity: usize) -> Self { |
428 | Self::with_capacity_in(capacity, Global) |
429 | } |
430 | |
431 | /// Creates a `Vec<T>` directly from a pointer, a capacity, and a length. |
432 | /// |
433 | /// # Safety |
434 | /// |
435 | /// This is highly unsafe, due to the number of invariants that aren't |
436 | /// checked: |
437 | /// |
438 | /// * `T` needs to have the same alignment as what `ptr` was allocated with. |
439 | /// (`T` having a less strict alignment is not sufficient, the alignment really |
440 | /// needs to be equal to satisfy the [`dealloc`] requirement that memory must be |
441 | /// allocated and deallocated with the same layout.) |
442 | /// * The size of `T` times the `capacity` (ie. the allocated size in bytes) needs |
443 | /// to be the same size as the pointer was allocated with. (Because similar to |
444 | /// alignment, [`dealloc`] must be called with the same layout `size`.) |
445 | /// * `length` needs to be less than or equal to `capacity`. |
446 | /// * The first `length` values must be properly initialized values of type `T`. |
447 | /// * `capacity` needs to be the capacity that the pointer was allocated with. |
448 | /// * The allocated size in bytes must be no larger than `isize::MAX`. |
449 | /// See the safety documentation of [`pointer::offset`](https://doc.rust-lang.org/nightly/std/primitive.pointer.html#method.offset). |
450 | /// |
451 | /// These requirements are always upheld by any `ptr` that has been allocated |
452 | /// via `Vec<T>`. Other allocation sources are allowed if the invariants are |
453 | /// upheld. |
454 | /// |
455 | /// Violating these may cause problems like corrupting the allocator's |
456 | /// internal data structures. For example it is normally **not** safe |
457 | /// to build a `Vec<u8>` from a pointer to a C `char` array with length |
458 | /// `size_t`, doing so is only safe if the array was initially allocated by |
459 | /// a `Vec` or `String`. |
460 | /// It's also not safe to build one from a `Vec<u16>` and its length, because |
461 | /// the allocator cares about the alignment, and these two types have different |
462 | /// alignments. The buffer was allocated with alignment 2 (for `u16`), but after |
463 | /// turning it into a `Vec<u8>` it'll be deallocated with alignment 1. To avoid |
464 | /// these issues, it is often preferable to do casting/transmuting using |
465 | /// [`slice::from_raw_parts`] instead. |
466 | /// |
467 | /// The ownership of `ptr` is effectively transferred to the |
468 | /// `Vec<T>` which may then deallocate, reallocate or change the |
469 | /// contents of memory pointed to by the pointer at will. Ensure |
470 | /// that nothing else uses the pointer after calling this |
471 | /// function. |
472 | /// |
473 | /// [`String`]: alloc_crate::string::String |
474 | /// [`dealloc`]: crate::alloc::GlobalAlloc::dealloc |
475 | /// |
476 | /// # Examples |
477 | /// |
478 | /// ``` |
479 | /// use std::ptr; |
480 | /// use std::mem; |
481 | /// |
482 | /// let v = vec![1, 2, 3]; |
483 | /// |
484 | // FIXME Update this when vec_into_raw_parts is stabilized |
485 | /// // Prevent running `v`'s destructor so we are in complete control |
486 | /// // of the allocation. |
487 | /// let mut v = mem::ManuallyDrop::new(v); |
488 | /// |
489 | /// // Pull out the various important pieces of information about `v` |
490 | /// let p = v.as_mut_ptr(); |
491 | /// let len = v.len(); |
492 | /// let cap = v.capacity(); |
493 | /// |
494 | /// unsafe { |
495 | /// // Overwrite memory with 4, 5, 6 |
496 | /// for i in 0..len { |
497 | /// ptr::write(p.add(i), 4 + i); |
498 | /// } |
499 | /// |
500 | /// // Put everything back together into a Vec |
501 | /// let rebuilt = Vec::from_raw_parts(p, len, cap); |
502 | /// assert_eq!(rebuilt, [4, 5, 6]); |
503 | /// } |
504 | /// ``` |
505 | /// |
506 | /// Using memory that was allocated elsewhere: |
507 | /// |
508 | /// ```rust |
509 | /// #![feature(allocator_api)] |
510 | /// |
511 | /// use std::alloc::{AllocError, Allocator, Global, Layout}; |
512 | /// |
513 | /// fn main() { |
514 | /// let layout = Layout::array::<u32>(16).expect("overflow cannot happen" ); |
515 | /// |
516 | /// let vec = unsafe { |
517 | /// let mem = match Global.allocate(layout) { |
518 | /// Ok(mem) => mem.cast::<u32>().as_ptr(), |
519 | /// Err(AllocError) => return, |
520 | /// }; |
521 | /// |
522 | /// mem.write(1_000_000); |
523 | /// |
524 | /// Vec::from_raw_parts_in(mem, 1, 16, Global) |
525 | /// }; |
526 | /// |
527 | /// assert_eq!(vec, &[1_000_000]); |
528 | /// assert_eq!(vec.capacity(), 16); |
529 | /// } |
530 | /// ``` |
531 | #[inline (always)] |
532 | pub unsafe fn from_raw_parts(ptr: *mut T, length: usize, capacity: usize) -> Self { |
533 | unsafe { Self::from_raw_parts_in(ptr, length, capacity, Global) } |
534 | } |
535 | } |
536 | |
537 | impl<T, A: Allocator> Vec<T, A> { |
538 | /// Constructs a new, empty `Vec<T, A>`. |
539 | /// |
540 | /// The vector will not allocate until elements are pushed onto it. |
541 | /// |
542 | /// # Examples |
543 | /// |
544 | /// ``` |
545 | /// use std::alloc::System; |
546 | /// |
547 | /// # #[allow (unused_mut)] |
548 | /// let mut vec: Vec<i32, _> = Vec::new_in(System); |
549 | /// ``` |
550 | #[inline (always)] |
551 | pub const fn new_in(alloc: A) -> Self { |
552 | Vec { |
553 | buf: RawVec::new_in(alloc), |
554 | len: 0, |
555 | } |
556 | } |
557 | |
558 | /// Constructs a new, empty `Vec<T, A>` with at least the specified capacity |
559 | /// with the provided allocator. |
560 | /// |
561 | /// The vector will be able to hold at least `capacity` elements without |
562 | /// reallocating. This method is allowed to allocate for more elements than |
563 | /// `capacity`. If `capacity` is 0, the vector will not allocate. |
564 | /// |
565 | /// It is important to note that although the returned vector has the |
566 | /// minimum *capacity* specified, the vector will have a zero *length*. For |
567 | /// an explanation of the difference between length and capacity, see |
568 | /// *[Capacity and reallocation]*. |
569 | /// |
570 | /// If it is important to know the exact allocated capacity of a `Vec`, |
571 | /// always use the [`capacity`] method after construction. |
572 | /// |
573 | /// For `Vec<T, A>` where `T` is a zero-sized type, there will be no allocation |
574 | /// and the capacity will always be `usize::MAX`. |
575 | /// |
576 | /// [Capacity and reallocation]: #capacity-and-reallocation |
577 | /// [`capacity`]: Vec::capacity |
578 | /// |
579 | /// # Panics |
580 | /// |
581 | /// Panics if the new capacity exceeds `isize::MAX` bytes. |
582 | /// |
583 | /// # Examples |
584 | /// |
585 | /// ``` |
586 | /// use std::alloc::System; |
587 | /// |
588 | /// let mut vec = Vec::with_capacity_in(10, System); |
589 | /// |
590 | /// // The vector contains no items, even though it has capacity for more |
591 | /// assert_eq!(vec.len(), 0); |
592 | /// assert_eq!(vec.capacity(), 10); |
593 | /// |
594 | /// // These are all done without reallocating... |
595 | /// for i in 0..10 { |
596 | /// vec.push(i); |
597 | /// } |
598 | /// assert_eq!(vec.len(), 10); |
599 | /// assert_eq!(vec.capacity(), 10); |
600 | /// |
601 | /// // ...but this may make the vector reallocate |
602 | /// vec.push(11); |
603 | /// assert_eq!(vec.len(), 11); |
604 | /// assert!(vec.capacity() >= 11); |
605 | /// |
606 | /// // A vector of a zero-sized type will always over-allocate, since no |
607 | /// // allocation is necessary |
608 | /// let vec_units = Vec::<(), System>::with_capacity_in(10, System); |
609 | /// assert_eq!(vec_units.capacity(), usize::MAX); |
610 | /// ``` |
611 | #[cfg (not(no_global_oom_handling))] |
612 | #[inline (always)] |
613 | pub fn with_capacity_in(capacity: usize, alloc: A) -> Self { |
614 | Vec { |
615 | buf: RawVec::with_capacity_in(capacity, alloc), |
616 | len: 0, |
617 | } |
618 | } |
619 | |
620 | /// Creates a `Vec<T, A>` directly from a pointer, a capacity, a length, |
621 | /// and an allocator. |
622 | /// |
623 | /// # Safety |
624 | /// |
625 | /// This is highly unsafe, due to the number of invariants that aren't |
626 | /// checked: |
627 | /// |
628 | /// * `T` needs to have the same alignment as what `ptr` was allocated with. |
629 | /// (`T` having a less strict alignment is not sufficient, the alignment really |
630 | /// needs to be equal to satisfy the [`dealloc`] requirement that memory must be |
631 | /// allocated and deallocated with the same layout.) |
632 | /// * The size of `T` times the `capacity` (ie. the allocated size in bytes) needs |
633 | /// to be the same size as the pointer was allocated with. (Because similar to |
634 | /// alignment, [`dealloc`] must be called with the same layout `size`.) |
635 | /// * `length` needs to be less than or equal to `capacity`. |
636 | /// * The first `length` values must be properly initialized values of type `T`. |
637 | /// * `capacity` needs to [*fit*] the layout size that the pointer was allocated with. |
638 | /// * The allocated size in bytes must be no larger than `isize::MAX`. |
639 | /// See the safety documentation of [`pointer::offset`](https://doc.rust-lang.org/nightly/std/primitive.pointer.html#method.offset). |
640 | /// |
641 | /// These requirements are always upheld by any `ptr` that has been allocated |
642 | /// via `Vec<T, A>`. Other allocation sources are allowed if the invariants are |
643 | /// upheld. |
644 | /// |
645 | /// Violating these may cause problems like corrupting the allocator's |
646 | /// internal data structures. For example it is **not** safe |
647 | /// to build a `Vec<u8>` from a pointer to a C `char` array with length `size_t`. |
648 | /// It's also not safe to build one from a `Vec<u16>` and its length, because |
649 | /// the allocator cares about the alignment, and these two types have different |
650 | /// alignments. The buffer was allocated with alignment 2 (for `u16`), but after |
651 | /// turning it into a `Vec<u8>` it'll be deallocated with alignment 1. |
652 | /// |
653 | /// The ownership of `ptr` is effectively transferred to the |
654 | /// `Vec<T>` which may then deallocate, reallocate or change the |
655 | /// contents of memory pointed to by the pointer at will. Ensure |
656 | /// that nothing else uses the pointer after calling this |
657 | /// function. |
658 | /// |
659 | /// [`String`]: alloc_crate::string::String |
660 | /// [`dealloc`]: crate::alloc::GlobalAlloc::dealloc |
661 | /// [*fit*]: crate::alloc::Allocator#memory-fitting |
662 | /// |
663 | /// # Examples |
664 | /// |
665 | /// ``` |
666 | /// use std::alloc::System; |
667 | /// |
668 | /// use std::ptr; |
669 | /// use std::mem; |
670 | /// |
671 | /// |
672 | /// # use allocator_api2::vec::Vec; |
673 | /// let mut v = Vec::with_capacity_in(3, System); |
674 | /// v.push(1); |
675 | /// v.push(2); |
676 | /// v.push(3); |
677 | /// |
678 | // FIXME Update this when vec_into_raw_parts is stabilized |
679 | /// // Prevent running `v`'s destructor so we are in complete control |
680 | /// // of the allocation. |
681 | /// let mut v = mem::ManuallyDrop::new(v); |
682 | /// |
683 | /// // Pull out the various important pieces of information about `v` |
684 | /// let p = v.as_mut_ptr(); |
685 | /// let len = v.len(); |
686 | /// let cap = v.capacity(); |
687 | /// let alloc = v.allocator(); |
688 | /// |
689 | /// unsafe { |
690 | /// // Overwrite memory with 4, 5, 6 |
691 | /// for i in 0..len { |
692 | /// ptr::write(p.add(i), 4 + i); |
693 | /// } |
694 | /// |
695 | /// // Put everything back together into a Vec |
696 | /// let rebuilt = Vec::from_raw_parts_in(p, len, cap, alloc.clone()); |
697 | /// assert_eq!(rebuilt, [4, 5, 6]); |
698 | /// } |
699 | /// ``` |
700 | /// |
701 | /// Using memory that was allocated elsewhere: |
702 | /// |
703 | /// ```rust |
704 | /// use std::alloc::{alloc, Layout}; |
705 | /// |
706 | /// fn main() { |
707 | /// let layout = Layout::array::<u32>(16).expect("overflow cannot happen" ); |
708 | /// let vec = unsafe { |
709 | /// let mem = alloc(layout).cast::<u32>(); |
710 | /// if mem.is_null() { |
711 | /// return; |
712 | /// } |
713 | /// |
714 | /// mem.write(1_000_000); |
715 | /// |
716 | /// Vec::from_raw_parts(mem, 1, 16) |
717 | /// }; |
718 | /// |
719 | /// assert_eq!(vec, &[1_000_000]); |
720 | /// assert_eq!(vec.capacity(), 16); |
721 | /// } |
722 | /// ``` |
723 | #[inline (always)] |
724 | pub unsafe fn from_raw_parts_in(ptr: *mut T, length: usize, capacity: usize, alloc: A) -> Self { |
725 | unsafe { |
726 | Vec { |
727 | buf: RawVec::from_raw_parts_in(ptr, capacity, alloc), |
728 | len: length, |
729 | } |
730 | } |
731 | } |
732 | |
733 | /// Decomposes a `Vec<T>` into its raw components. |
734 | /// |
735 | /// Returns the raw pointer to the underlying data, the length of |
736 | /// the vector (in elements), and the allocated capacity of the |
737 | /// data (in elements). These are the same arguments in the same |
738 | /// order as the arguments to [`from_raw_parts`]. |
739 | /// |
740 | /// After calling this function, the caller is responsible for the |
741 | /// memory previously managed by the `Vec`. The only way to do |
742 | /// this is to convert the raw pointer, length, and capacity back |
743 | /// into a `Vec` with the [`from_raw_parts`] function, allowing |
744 | /// the destructor to perform the cleanup. |
745 | /// |
746 | /// [`from_raw_parts`]: Vec::from_raw_parts |
747 | /// |
748 | /// # Examples |
749 | /// |
750 | /// ``` |
751 | /// #![feature(vec_into_raw_parts)] |
752 | /// let v: Vec<i32> = vec![-1, 0, 1]; |
753 | /// |
754 | /// let (ptr, len, cap) = v.into_raw_parts(); |
755 | /// |
756 | /// let rebuilt = unsafe { |
757 | /// // We can now make changes to the components, such as |
758 | /// // transmuting the raw pointer to a compatible type. |
759 | /// let ptr = ptr as *mut u32; |
760 | /// |
761 | /// Vec::from_raw_parts(ptr, len, cap) |
762 | /// }; |
763 | /// assert_eq!(rebuilt, [4294967295, 0, 1]); |
764 | /// ``` |
765 | pub fn into_raw_parts(self) -> (*mut T, usize, usize) { |
766 | let mut me = ManuallyDrop::new(self); |
767 | (me.as_mut_ptr(), me.len(), me.capacity()) |
768 | } |
769 | |
770 | /// Decomposes a `Vec<T>` into its raw components. |
771 | /// |
772 | /// Returns the raw pointer to the underlying data, the length of the vector (in elements), |
773 | /// the allocated capacity of the data (in elements), and the allocator. These are the same |
774 | /// arguments in the same order as the arguments to [`from_raw_parts_in`]. |
775 | /// |
776 | /// After calling this function, the caller is responsible for the |
777 | /// memory previously managed by the `Vec`. The only way to do |
778 | /// this is to convert the raw pointer, length, and capacity back |
779 | /// into a `Vec` with the [`from_raw_parts_in`] function, allowing |
780 | /// the destructor to perform the cleanup. |
781 | /// |
782 | /// [`from_raw_parts_in`]: Vec::from_raw_parts_in |
783 | /// |
784 | /// # Examples |
785 | /// |
786 | /// ``` |
787 | /// #![feature(allocator_api, vec_into_raw_parts)] |
788 | /// |
789 | /// use std::alloc::System; |
790 | /// |
791 | /// let mut v: Vec<i32, System> = Vec::new_in(System); |
792 | /// v.push(-1); |
793 | /// v.push(0); |
794 | /// v.push(1); |
795 | /// |
796 | /// let (ptr, len, cap, alloc) = v.into_raw_parts_with_alloc(); |
797 | /// |
798 | /// let rebuilt = unsafe { |
799 | /// // We can now make changes to the components, such as |
800 | /// // transmuting the raw pointer to a compatible type. |
801 | /// let ptr = ptr as *mut u32; |
802 | /// |
803 | /// Vec::from_raw_parts_in(ptr, len, cap, alloc) |
804 | /// }; |
805 | /// assert_eq!(rebuilt, [4294967295, 0, 1]); |
806 | /// ``` |
807 | // #[unstable(feature = "vec_into_raw_parts", reason = "new API", issue = "65816")] |
808 | pub fn into_raw_parts_with_alloc(self) -> (*mut T, usize, usize, A) { |
809 | let mut me = ManuallyDrop::new(self); |
810 | let len = me.len(); |
811 | let capacity = me.capacity(); |
812 | let ptr = me.as_mut_ptr(); |
813 | let alloc = unsafe { ptr::read(me.allocator()) }; |
814 | (ptr, len, capacity, alloc) |
815 | } |
816 | |
817 | /// Returns the total number of elements the vector can hold without |
818 | /// reallocating. |
819 | /// |
820 | /// # Examples |
821 | /// |
822 | /// ``` |
823 | /// let mut vec: Vec<i32> = Vec::with_capacity(10); |
824 | /// vec.push(42); |
825 | /// assert_eq!(vec.capacity(), 10); |
826 | /// ``` |
827 | #[inline (always)] |
828 | pub fn capacity(&self) -> usize { |
829 | self.buf.capacity() |
830 | } |
831 | |
832 | /// Reserves capacity for at least `additional` more elements to be inserted |
833 | /// in the given `Vec<T>`. The collection may reserve more space to |
834 | /// speculatively avoid frequent reallocations. After calling `reserve`, |
835 | /// capacity will be greater than or equal to `self.len() + additional`. |
836 | /// Does nothing if capacity is already sufficient. |
837 | /// |
838 | /// # Panics |
839 | /// |
840 | /// Panics if the new capacity exceeds `isize::MAX` bytes. |
841 | /// |
842 | /// # Examples |
843 | /// |
844 | /// ``` |
845 | /// let mut vec = vec![1]; |
846 | /// vec.reserve(10); |
847 | /// assert!(vec.capacity() >= 11); |
848 | /// ``` |
849 | #[cfg (not(no_global_oom_handling))] |
850 | #[inline (always)] |
851 | pub fn reserve(&mut self, additional: usize) { |
852 | self.buf.reserve(self.len, additional); |
853 | } |
854 | |
855 | /// Reserves the minimum capacity for at least `additional` more elements to |
856 | /// be inserted in the given `Vec<T>`. Unlike [`reserve`], this will not |
857 | /// deliberately over-allocate to speculatively avoid frequent allocations. |
858 | /// After calling `reserve_exact`, capacity will be greater than or equal to |
859 | /// `self.len() + additional`. Does nothing if the capacity is already |
860 | /// sufficient. |
861 | /// |
862 | /// Note that the allocator may give the collection more space than it |
863 | /// requests. Therefore, capacity can not be relied upon to be precisely |
864 | /// minimal. Prefer [`reserve`] if future insertions are expected. |
865 | /// |
866 | /// [`reserve`]: Vec::reserve |
867 | /// |
868 | /// # Panics |
869 | /// |
870 | /// Panics if the new capacity exceeds `isize::MAX` bytes. |
871 | /// |
872 | /// # Examples |
873 | /// |
874 | /// ``` |
875 | /// let mut vec = vec![1]; |
876 | /// vec.reserve_exact(10); |
877 | /// assert!(vec.capacity() >= 11); |
878 | /// ``` |
879 | #[cfg (not(no_global_oom_handling))] |
880 | #[inline (always)] |
881 | pub fn reserve_exact(&mut self, additional: usize) { |
882 | self.buf.reserve_exact(self.len, additional); |
883 | } |
884 | |
885 | /// Tries to reserve capacity for at least `additional` more elements to be inserted |
886 | /// in the given `Vec<T>`. The collection may reserve more space to speculatively avoid |
887 | /// frequent reallocations. After calling `try_reserve`, capacity will be |
888 | /// greater than or equal to `self.len() + additional` if it returns |
889 | /// `Ok(())`. Does nothing if capacity is already sufficient. This method |
890 | /// preserves the contents even if an error occurs. |
891 | /// |
892 | /// # Errors |
893 | /// |
894 | /// If the capacity overflows, or the allocator reports a failure, then an error |
895 | /// is returned. |
896 | /// |
897 | /// # Examples |
898 | /// |
899 | /// ``` |
900 | /// use std::collections::TryReserveError; |
901 | /// |
902 | /// fn process_data(data: &[u32]) -> Result<Vec<u32>, TryReserveError> { |
903 | /// let mut output = Vec::new(); |
904 | /// |
905 | /// // Pre-reserve the memory, exiting if we can't |
906 | /// output.try_reserve(data.len())?; |
907 | /// |
908 | /// // Now we know this can't OOM in the middle of our complex work |
909 | /// output.extend(data.iter().map(|&val| { |
910 | /// val * 2 + 5 // very complicated |
911 | /// })); |
912 | /// |
913 | /// Ok(output) |
914 | /// } |
915 | /// # process_data(&[1, 2, 3]).expect("why is the test harness OOMing on 12 bytes?" ); |
916 | /// ``` |
917 | #[inline (always)] |
918 | pub fn try_reserve(&mut self, additional: usize) -> Result<(), TryReserveError> { |
919 | self.buf.try_reserve(self.len, additional) |
920 | } |
921 | |
922 | /// Tries to reserve the minimum capacity for at least `additional` |
923 | /// elements to be inserted in the given `Vec<T>`. Unlike [`try_reserve`], |
924 | /// this will not deliberately over-allocate to speculatively avoid frequent |
925 | /// allocations. After calling `try_reserve_exact`, capacity will be greater |
926 | /// than or equal to `self.len() + additional` if it returns `Ok(())`. |
927 | /// Does nothing if the capacity is already sufficient. |
928 | /// |
929 | /// Note that the allocator may give the collection more space than it |
930 | /// requests. Therefore, capacity can not be relied upon to be precisely |
931 | /// minimal. Prefer [`try_reserve`] if future insertions are expected. |
932 | /// |
933 | /// [`try_reserve`]: Vec::try_reserve |
934 | /// |
935 | /// # Errors |
936 | /// |
937 | /// If the capacity overflows, or the allocator reports a failure, then an error |
938 | /// is returned. |
939 | /// |
940 | /// # Examples |
941 | /// |
942 | /// ``` |
943 | /// use std::collections::TryReserveError; |
944 | /// |
945 | /// fn process_data(data: &[u32]) -> Result<Vec<u32>, TryReserveError> { |
946 | /// let mut output = Vec::new(); |
947 | /// |
948 | /// // Pre-reserve the memory, exiting if we can't |
949 | /// output.try_reserve_exact(data.len())?; |
950 | /// |
951 | /// // Now we know this can't OOM in the middle of our complex work |
952 | /// output.extend(data.iter().map(|&val| { |
953 | /// val * 2 + 5 // very complicated |
954 | /// })); |
955 | /// |
956 | /// Ok(output) |
957 | /// } |
958 | /// # process_data(&[1, 2, 3]).expect("why is the test harness OOMing on 12 bytes?" ); |
959 | /// ``` |
960 | #[inline (always)] |
961 | pub fn try_reserve_exact(&mut self, additional: usize) -> Result<(), TryReserveError> { |
962 | self.buf.try_reserve_exact(self.len, additional) |
963 | } |
964 | |
965 | /// Shrinks the capacity of the vector as much as possible. |
966 | /// |
967 | /// It will drop down as close as possible to the length but the allocator |
968 | /// may still inform the vector that there is space for a few more elements. |
969 | /// |
970 | /// # Examples |
971 | /// |
972 | /// ``` |
973 | /// let mut vec = Vec::with_capacity(10); |
974 | /// vec.extend([1, 2, 3]); |
975 | /// assert_eq!(vec.capacity(), 10); |
976 | /// vec.shrink_to_fit(); |
977 | /// assert!(vec.capacity() >= 3); |
978 | /// ``` |
979 | #[cfg (not(no_global_oom_handling))] |
980 | #[inline (always)] |
981 | pub fn shrink_to_fit(&mut self) { |
982 | // The capacity is never less than the length, and there's nothing to do when |
983 | // they are equal, so we can avoid the panic case in `RawVec::shrink_to_fit` |
984 | // by only calling it with a greater capacity. |
985 | if self.capacity() > self.len { |
986 | self.buf.shrink_to_fit(self.len); |
987 | } |
988 | } |
989 | |
990 | /// Shrinks the capacity of the vector with a lower bound. |
991 | /// |
992 | /// The capacity will remain at least as large as both the length |
993 | /// and the supplied value. |
994 | /// |
995 | /// If the current capacity is less than the lower limit, this is a no-op. |
996 | /// |
997 | /// # Examples |
998 | /// |
999 | /// ``` |
1000 | /// let mut vec = Vec::with_capacity(10); |
1001 | /// vec.extend([1, 2, 3]); |
1002 | /// assert_eq!(vec.capacity(), 10); |
1003 | /// vec.shrink_to(4); |
1004 | /// assert!(vec.capacity() >= 4); |
1005 | /// vec.shrink_to(0); |
1006 | /// assert!(vec.capacity() >= 3); |
1007 | /// ``` |
1008 | #[cfg (not(no_global_oom_handling))] |
1009 | #[inline (always)] |
1010 | pub fn shrink_to(&mut self, min_capacity: usize) { |
1011 | if self.capacity() > min_capacity { |
1012 | self.buf.shrink_to_fit(cmp::max(self.len, min_capacity)); |
1013 | } |
1014 | } |
1015 | |
1016 | /// Converts the vector into [`Box<[T]>`][owned slice]. |
1017 | /// |
1018 | /// If the vector has excess capacity, its items will be moved into a |
1019 | /// newly-allocated buffer with exactly the right capacity. |
1020 | /// |
1021 | /// [owned slice]: Box |
1022 | /// |
1023 | /// # Examples |
1024 | /// |
1025 | /// ``` |
1026 | /// let v = vec![1, 2, 3]; |
1027 | /// |
1028 | /// let slice = v.into_boxed_slice(); |
1029 | /// ``` |
1030 | /// |
1031 | /// Any excess capacity is removed: |
1032 | /// |
1033 | /// ``` |
1034 | /// let mut vec = Vec::with_capacity(10); |
1035 | /// vec.extend([1, 2, 3]); |
1036 | /// |
1037 | /// assert_eq!(vec.capacity(), 10); |
1038 | /// let slice = vec.into_boxed_slice(); |
1039 | /// assert_eq!(slice.into_vec().capacity(), 3); |
1040 | /// ``` |
1041 | #[cfg (not(no_global_oom_handling))] |
1042 | #[inline (always)] |
1043 | pub fn into_boxed_slice(mut self) -> Box<[T], A> { |
1044 | unsafe { |
1045 | self.shrink_to_fit(); |
1046 | let me = ManuallyDrop::new(self); |
1047 | let buf = ptr::read(&me.buf); |
1048 | let len = me.len(); |
1049 | buf.into_box(len).assume_init() |
1050 | } |
1051 | } |
1052 | |
1053 | /// Shortens the vector, keeping the first `len` elements and dropping |
1054 | /// the rest. |
1055 | /// |
1056 | /// If `len` is greater than the vector's current length, this has no |
1057 | /// effect. |
1058 | /// |
1059 | /// The [`drain`] method can emulate `truncate`, but causes the excess |
1060 | /// elements to be returned instead of dropped. |
1061 | /// |
1062 | /// Note that this method has no effect on the allocated capacity |
1063 | /// of the vector. |
1064 | /// |
1065 | /// # Examples |
1066 | /// |
1067 | /// Truncating a five element vector to two elements: |
1068 | /// |
1069 | /// ``` |
1070 | /// let mut vec = vec![1, 2, 3, 4, 5]; |
1071 | /// vec.truncate(2); |
1072 | /// assert_eq!(vec, [1, 2]); |
1073 | /// ``` |
1074 | /// |
1075 | /// No truncation occurs when `len` is greater than the vector's current |
1076 | /// length: |
1077 | /// |
1078 | /// ``` |
1079 | /// let mut vec = vec![1, 2, 3]; |
1080 | /// vec.truncate(8); |
1081 | /// assert_eq!(vec, [1, 2, 3]); |
1082 | /// ``` |
1083 | /// |
1084 | /// Truncating when `len == 0` is equivalent to calling the [`clear`] |
1085 | /// method. |
1086 | /// |
1087 | /// ``` |
1088 | /// let mut vec = vec![1, 2, 3]; |
1089 | /// vec.truncate(0); |
1090 | /// assert_eq!(vec, []); |
1091 | /// ``` |
1092 | /// |
1093 | /// [`clear`]: Vec::clear |
1094 | /// [`drain`]: Vec::drain |
1095 | #[inline (always)] |
1096 | pub fn truncate(&mut self, len: usize) { |
1097 | // This is safe because: |
1098 | // |
1099 | // * the slice passed to `drop_in_place` is valid; the `len > self.len` |
1100 | // case avoids creating an invalid slice, and |
1101 | // * the `len` of the vector is shrunk before calling `drop_in_place`, |
1102 | // such that no value will be dropped twice in case `drop_in_place` |
1103 | // were to panic once (if it panics twice, the program aborts). |
1104 | unsafe { |
1105 | // Note: It's intentional that this is `>` and not `>=`. |
1106 | // Changing it to `>=` has negative performance |
1107 | // implications in some cases. See #78884 for more. |
1108 | if len > self.len { |
1109 | return; |
1110 | } |
1111 | let remaining_len = self.len - len; |
1112 | let s = ptr::slice_from_raw_parts_mut(self.as_mut_ptr().add(len), remaining_len); |
1113 | self.len = len; |
1114 | ptr::drop_in_place(s); |
1115 | } |
1116 | } |
1117 | |
1118 | /// Extracts a slice containing the entire vector. |
1119 | /// |
1120 | /// Equivalent to `&s[..]`. |
1121 | /// |
1122 | /// # Examples |
1123 | /// |
1124 | /// ``` |
1125 | /// use std::io::{self, Write}; |
1126 | /// let buffer = vec![1, 2, 3, 5, 8]; |
1127 | /// io::sink().write(buffer.as_slice()).unwrap(); |
1128 | /// ``` |
1129 | #[inline (always)] |
1130 | pub fn as_slice(&self) -> &[T] { |
1131 | self |
1132 | } |
1133 | |
1134 | /// Extracts a mutable slice of the entire vector. |
1135 | /// |
1136 | /// Equivalent to `&mut s[..]`. |
1137 | /// |
1138 | /// # Examples |
1139 | /// |
1140 | /// ``` |
1141 | /// use std::io::{self, Read}; |
1142 | /// let mut buffer = vec![0; 3]; |
1143 | /// io::repeat(0b101).read_exact(buffer.as_mut_slice()).unwrap(); |
1144 | /// ``` |
1145 | #[inline (always)] |
1146 | pub fn as_mut_slice(&mut self) -> &mut [T] { |
1147 | self |
1148 | } |
1149 | |
1150 | /// Returns a raw pointer to the vector's buffer, or a dangling raw pointer |
1151 | /// valid for zero sized reads if the vector didn't allocate. |
1152 | /// |
1153 | /// The caller must ensure that the vector outlives the pointer this |
1154 | /// function returns, or else it will end up pointing to garbage. |
1155 | /// Modifying the vector may cause its buffer to be reallocated, |
1156 | /// which would also make any pointers to it invalid. |
1157 | /// |
1158 | /// The caller must also ensure that the memory the pointer (non-transitively) points to |
1159 | /// is never written to (except inside an `UnsafeCell`) using this pointer or any pointer |
1160 | /// derived from it. If you need to mutate the contents of the slice, use [`as_mut_ptr`]. |
1161 | /// |
1162 | /// # Examples |
1163 | /// |
1164 | /// ``` |
1165 | /// let x = vec![1, 2, 4]; |
1166 | /// let x_ptr = x.as_ptr(); |
1167 | /// |
1168 | /// unsafe { |
1169 | /// for i in 0..x.len() { |
1170 | /// assert_eq!(*x_ptr.add(i), 1 << i); |
1171 | /// } |
1172 | /// } |
1173 | /// ``` |
1174 | /// |
1175 | /// [`as_mut_ptr`]: Vec::as_mut_ptr |
1176 | #[inline (always)] |
1177 | pub fn as_ptr(&self) -> *const T { |
1178 | // We shadow the slice method of the same name to avoid going through |
1179 | // `deref`, which creates an intermediate reference. |
1180 | let ptr = self.buf.ptr(); |
1181 | unsafe { |
1182 | assume(!ptr.is_null()); |
1183 | } |
1184 | ptr |
1185 | } |
1186 | |
1187 | /// Returns an unsafe mutable pointer to the vector's buffer, or a dangling |
1188 | /// raw pointer valid for zero sized reads if the vector didn't allocate. |
1189 | /// |
1190 | /// The caller must ensure that the vector outlives the pointer this |
1191 | /// function returns, or else it will end up pointing to garbage. |
1192 | /// Modifying the vector may cause its buffer to be reallocated, |
1193 | /// which would also make any pointers to it invalid. |
1194 | /// |
1195 | /// # Examples |
1196 | /// |
1197 | /// ``` |
1198 | /// // Allocate vector big enough for 4 elements. |
1199 | /// let size = 4; |
1200 | /// let mut x: Vec<i32> = Vec::with_capacity(size); |
1201 | /// let x_ptr = x.as_mut_ptr(); |
1202 | /// |
1203 | /// // Initialize elements via raw pointer writes, then set length. |
1204 | /// unsafe { |
1205 | /// for i in 0..size { |
1206 | /// *x_ptr.add(i) = i as i32; |
1207 | /// } |
1208 | /// x.set_len(size); |
1209 | /// } |
1210 | /// assert_eq!(&*x, &[0, 1, 2, 3]); |
1211 | /// ``` |
1212 | #[inline (always)] |
1213 | pub fn as_mut_ptr(&mut self) -> *mut T { |
1214 | // We shadow the slice method of the same name to avoid going through |
1215 | // `deref_mut`, which creates an intermediate reference. |
1216 | let ptr = self.buf.ptr(); |
1217 | unsafe { |
1218 | assume(!ptr.is_null()); |
1219 | } |
1220 | ptr |
1221 | } |
1222 | |
1223 | /// Returns a reference to the underlying allocator. |
1224 | #[inline (always)] |
1225 | pub fn allocator(&self) -> &A { |
1226 | self.buf.allocator() |
1227 | } |
1228 | |
1229 | /// Forces the length of the vector to `new_len`. |
1230 | /// |
1231 | /// This is a low-level operation that maintains none of the normal |
1232 | /// invariants of the type. Normally changing the length of a vector |
1233 | /// is done using one of the safe operations instead, such as |
1234 | /// [`truncate`], [`resize`], [`extend`], or [`clear`]. |
1235 | /// |
1236 | /// [`truncate`]: Vec::truncate |
1237 | /// [`resize`]: Vec::resize |
1238 | /// [`extend`]: Extend::extend |
1239 | /// [`clear`]: Vec::clear |
1240 | /// |
1241 | /// # Safety |
1242 | /// |
1243 | /// - `new_len` must be less than or equal to [`capacity()`]. |
1244 | /// - The elements at `old_len..new_len` must be initialized. |
1245 | /// |
1246 | /// [`capacity()`]: Vec::capacity |
1247 | /// |
1248 | /// # Examples |
1249 | /// |
1250 | /// This method can be useful for situations in which the vector |
1251 | /// is serving as a buffer for other code, particularly over FFI: |
1252 | /// |
1253 | /// ```no_run |
1254 | /// # #![allow (dead_code)] |
1255 | /// # // This is just a minimal skeleton for the doc example; |
1256 | /// # // don't use this as a starting point for a real library. |
1257 | /// # pub struct StreamWrapper { strm: *mut std::ffi::c_void } |
1258 | /// # const Z_OK: i32 = 0; |
1259 | /// # extern "C" { |
1260 | /// # fn deflateGetDictionary( |
1261 | /// # strm: *mut std::ffi::c_void, |
1262 | /// # dictionary: *mut u8, |
1263 | /// # dictLength: *mut usize, |
1264 | /// # ) -> i32; |
1265 | /// # } |
1266 | /// # impl StreamWrapper { |
1267 | /// pub fn get_dictionary(&self) -> Option<Vec<u8>> { |
1268 | /// // Per the FFI method's docs, "32768 bytes is always enough". |
1269 | /// let mut dict = Vec::with_capacity(32_768); |
1270 | /// let mut dict_length = 0; |
1271 | /// // SAFETY: When `deflateGetDictionary` returns `Z_OK`, it holds that: |
1272 | /// // 1. `dict_length` elements were initialized. |
1273 | /// // 2. `dict_length` <= the capacity (32_768) |
1274 | /// // which makes `set_len` safe to call. |
1275 | /// unsafe { |
1276 | /// // Make the FFI call... |
1277 | /// let r = deflateGetDictionary(self.strm, dict.as_mut_ptr(), &mut dict_length); |
1278 | /// if r == Z_OK { |
1279 | /// // ...and update the length to what was initialized. |
1280 | /// dict.set_len(dict_length); |
1281 | /// Some(dict) |
1282 | /// } else { |
1283 | /// None |
1284 | /// } |
1285 | /// } |
1286 | /// } |
1287 | /// # } |
1288 | /// ``` |
1289 | /// |
1290 | /// While the following example is sound, there is a memory leak since |
1291 | /// the inner vectors were not freed prior to the `set_len` call: |
1292 | /// |
1293 | /// ``` |
1294 | /// let mut vec = vec![vec![1, 0, 0], |
1295 | /// vec![0, 1, 0], |
1296 | /// vec![0, 0, 1]]; |
1297 | /// // SAFETY: |
1298 | /// // 1. `old_len..0` is empty so no elements need to be initialized. |
1299 | /// // 2. `0 <= capacity` always holds whatever `capacity` is. |
1300 | /// unsafe { |
1301 | /// vec.set_len(0); |
1302 | /// } |
1303 | /// ``` |
1304 | /// |
1305 | /// Normally, here, one would use [`clear`] instead to correctly drop |
1306 | /// the contents and thus not leak memory. |
1307 | #[inline (always)] |
1308 | pub unsafe fn set_len(&mut self, new_len: usize) { |
1309 | debug_assert!(new_len <= self.capacity()); |
1310 | |
1311 | self.len = new_len; |
1312 | } |
1313 | |
1314 | /// Removes an element from the vector and returns it. |
1315 | /// |
1316 | /// The removed element is replaced by the last element of the vector. |
1317 | /// |
1318 | /// This does not preserve ordering, but is *O*(1). |
1319 | /// If you need to preserve the element order, use [`remove`] instead. |
1320 | /// |
1321 | /// [`remove`]: Vec::remove |
1322 | /// |
1323 | /// # Panics |
1324 | /// |
1325 | /// Panics if `index` is out of bounds. |
1326 | /// |
1327 | /// # Examples |
1328 | /// |
1329 | /// ``` |
1330 | /// let mut v = vec!["foo" , "bar" , "baz" , "qux" ]; |
1331 | /// |
1332 | /// assert_eq!(v.swap_remove(1), "bar" ); |
1333 | /// assert_eq!(v, ["foo" , "qux" , "baz" ]); |
1334 | /// |
1335 | /// assert_eq!(v.swap_remove(0), "foo" ); |
1336 | /// assert_eq!(v, ["baz" , "qux" ]); |
1337 | /// ``` |
1338 | #[inline (always)] |
1339 | pub fn swap_remove(&mut self, index: usize) -> T { |
1340 | #[cold ] |
1341 | #[inline (never)] |
1342 | fn assert_failed(index: usize, len: usize) -> ! { |
1343 | panic!( |
1344 | "swap_remove index (is {}) should be < len (is {})" , |
1345 | index, len |
1346 | ); |
1347 | } |
1348 | |
1349 | let len = self.len(); |
1350 | if index >= len { |
1351 | assert_failed(index, len); |
1352 | } |
1353 | unsafe { |
1354 | // We replace self[index] with the last element. Note that if the |
1355 | // bounds check above succeeds there must be a last element (which |
1356 | // can be self[index] itself). |
1357 | let value = ptr::read(self.as_ptr().add(index)); |
1358 | let base_ptr = self.as_mut_ptr(); |
1359 | ptr::copy(base_ptr.add(len - 1), base_ptr.add(index), 1); |
1360 | self.set_len(len - 1); |
1361 | value |
1362 | } |
1363 | } |
1364 | |
1365 | /// Inserts an element at position `index` within the vector, shifting all |
1366 | /// elements after it to the right. |
1367 | /// |
1368 | /// # Panics |
1369 | /// |
1370 | /// Panics if `index > len`. |
1371 | /// |
1372 | /// # Examples |
1373 | /// |
1374 | /// ``` |
1375 | /// let mut vec = vec![1, 2, 3]; |
1376 | /// vec.insert(1, 4); |
1377 | /// assert_eq!(vec, [1, 4, 2, 3]); |
1378 | /// vec.insert(4, 5); |
1379 | /// assert_eq!(vec, [1, 4, 2, 3, 5]); |
1380 | /// ``` |
1381 | #[cfg (not(no_global_oom_handling))] |
1382 | pub fn insert(&mut self, index: usize, element: T) { |
1383 | #[cold ] |
1384 | #[inline (never)] |
1385 | fn assert_failed(index: usize, len: usize) -> ! { |
1386 | panic!( |
1387 | "insertion index (is {}) should be <= len (is {})" , |
1388 | index, len |
1389 | ); |
1390 | } |
1391 | |
1392 | let len = self.len(); |
1393 | |
1394 | // space for the new element |
1395 | if len == self.buf.capacity() { |
1396 | self.reserve(1); |
1397 | } |
1398 | |
1399 | unsafe { |
1400 | // infallible |
1401 | // The spot to put the new value |
1402 | { |
1403 | let p = self.as_mut_ptr().add(index); |
1404 | match cmp::Ord::cmp(&index, &len) { |
1405 | Ordering::Less => { |
1406 | // Shift everything over to make space. (Duplicating the |
1407 | // `index`th element into two consecutive places.) |
1408 | ptr::copy(p, p.add(1), len - index); |
1409 | } |
1410 | Ordering::Equal => { |
1411 | // No elements need shifting. |
1412 | } |
1413 | Ordering::Greater => { |
1414 | assert_failed(index, len); |
1415 | } |
1416 | } |
1417 | // Write it in, overwriting the first copy of the `index`th |
1418 | // element. |
1419 | ptr::write(p, element); |
1420 | } |
1421 | self.set_len(len + 1); |
1422 | } |
1423 | } |
1424 | |
1425 | /// Removes and returns the element at position `index` within the vector, |
1426 | /// shifting all elements after it to the left. |
1427 | /// |
1428 | /// Note: Because this shifts over the remaining elements, it has a |
1429 | /// worst-case performance of *O*(*n*). If you don't need the order of elements |
1430 | /// to be preserved, use [`swap_remove`] instead. If you'd like to remove |
1431 | /// elements from the beginning of the `Vec`, consider using |
1432 | /// [`VecDeque::pop_front`] instead. |
1433 | /// |
1434 | /// [`swap_remove`]: Vec::swap_remove |
1435 | /// [`VecDeque::pop_front`]: alloc_crate::collections::VecDeque::pop_front |
1436 | /// |
1437 | /// # Panics |
1438 | /// |
1439 | /// Panics if `index` is out of bounds. |
1440 | /// |
1441 | /// # Examples |
1442 | /// |
1443 | /// ``` |
1444 | /// let mut v = vec![1, 2, 3]; |
1445 | /// assert_eq!(v.remove(1), 2); |
1446 | /// assert_eq!(v, [1, 3]); |
1447 | /// ``` |
1448 | #[track_caller ] |
1449 | #[inline (always)] |
1450 | pub fn remove(&mut self, index: usize) -> T { |
1451 | #[cold ] |
1452 | #[inline (never)] |
1453 | #[track_caller ] |
1454 | fn assert_failed(index: usize, len: usize) -> ! { |
1455 | panic!("removal index (is {}) should be < len (is {})" , index, len); |
1456 | } |
1457 | |
1458 | let len = self.len(); |
1459 | if index >= len { |
1460 | assert_failed(index, len); |
1461 | } |
1462 | unsafe { |
1463 | // infallible |
1464 | let ret; |
1465 | { |
1466 | // the place we are taking from. |
1467 | let ptr = self.as_mut_ptr().add(index); |
1468 | // copy it out, unsafely having a copy of the value on |
1469 | // the stack and in the vector at the same time. |
1470 | ret = ptr::read(ptr); |
1471 | |
1472 | // Shift everything down to fill in that spot. |
1473 | ptr::copy(ptr.add(1), ptr, len - index - 1); |
1474 | } |
1475 | self.set_len(len - 1); |
1476 | ret |
1477 | } |
1478 | } |
1479 | |
1480 | /// Retains only the elements specified by the predicate. |
1481 | /// |
1482 | /// In other words, remove all elements `e` for which `f(&e)` returns `false`. |
1483 | /// This method operates in place, visiting each element exactly once in the |
1484 | /// original order, and preserves the order of the retained elements. |
1485 | /// |
1486 | /// # Examples |
1487 | /// |
1488 | /// ``` |
1489 | /// let mut vec = vec![1, 2, 3, 4]; |
1490 | /// vec.retain(|&x| x % 2 == 0); |
1491 | /// assert_eq!(vec, [2, 4]); |
1492 | /// ``` |
1493 | /// |
1494 | /// Because the elements are visited exactly once in the original order, |
1495 | /// external state may be used to decide which elements to keep. |
1496 | /// |
1497 | /// ``` |
1498 | /// let mut vec = vec![1, 2, 3, 4, 5]; |
1499 | /// let keep = [false, true, true, false, true]; |
1500 | /// let mut iter = keep.iter(); |
1501 | /// vec.retain(|_| *iter.next().unwrap()); |
1502 | /// assert_eq!(vec, [2, 3, 5]); |
1503 | /// ``` |
1504 | #[inline (always)] |
1505 | pub fn retain<F>(&mut self, mut f: F) |
1506 | where |
1507 | F: FnMut(&T) -> bool, |
1508 | { |
1509 | self.retain_mut(|elem| f(elem)); |
1510 | } |
1511 | |
1512 | /// Retains only the elements specified by the predicate, passing a mutable reference to it. |
1513 | /// |
1514 | /// In other words, remove all elements `e` such that `f(&mut e)` returns `false`. |
1515 | /// This method operates in place, visiting each element exactly once in the |
1516 | /// original order, and preserves the order of the retained elements. |
1517 | /// |
1518 | /// # Examples |
1519 | /// |
1520 | /// ``` |
1521 | /// let mut vec = vec![1, 2, 3, 4]; |
1522 | /// vec.retain_mut(|x| if *x <= 3 { |
1523 | /// *x += 1; |
1524 | /// true |
1525 | /// } else { |
1526 | /// false |
1527 | /// }); |
1528 | /// assert_eq!(vec, [2, 3, 4]); |
1529 | /// ``` |
1530 | #[inline ] |
1531 | pub fn retain_mut<F>(&mut self, mut f: F) |
1532 | where |
1533 | F: FnMut(&mut T) -> bool, |
1534 | { |
1535 | let original_len = self.len(); |
1536 | // Avoid double drop if the drop guard is not executed, |
1537 | // since we may make some holes during the process. |
1538 | unsafe { self.set_len(0) }; |
1539 | |
1540 | // Vec: [Kept, Kept, Hole, Hole, Hole, Hole, Unchecked, Unchecked] |
1541 | // |<- processed len ->| ^- next to check |
1542 | // |<- deleted cnt ->| |
1543 | // |<- original_len ->| |
1544 | // Kept: Elements which predicate returns true on. |
1545 | // Hole: Moved or dropped element slot. |
1546 | // Unchecked: Unchecked valid elements. |
1547 | // |
1548 | // This drop guard will be invoked when predicate or `drop` of element panicked. |
1549 | // It shifts unchecked elements to cover holes and `set_len` to the correct length. |
1550 | // In cases when predicate and `drop` never panick, it will be optimized out. |
1551 | struct BackshiftOnDrop<'a, T, A: Allocator> { |
1552 | v: &'a mut Vec<T, A>, |
1553 | processed_len: usize, |
1554 | deleted_cnt: usize, |
1555 | original_len: usize, |
1556 | } |
1557 | |
1558 | impl<T, A: Allocator> Drop for BackshiftOnDrop<'_, T, A> { |
1559 | fn drop(&mut self) { |
1560 | if self.deleted_cnt > 0 { |
1561 | // SAFETY: Trailing unchecked items must be valid since we never touch them. |
1562 | unsafe { |
1563 | ptr::copy( |
1564 | self.v.as_ptr().add(self.processed_len), |
1565 | self.v |
1566 | .as_mut_ptr() |
1567 | .add(self.processed_len - self.deleted_cnt), |
1568 | self.original_len - self.processed_len, |
1569 | ); |
1570 | } |
1571 | } |
1572 | // SAFETY: After filling holes, all items are in contiguous memory. |
1573 | unsafe { |
1574 | self.v.set_len(self.original_len - self.deleted_cnt); |
1575 | } |
1576 | } |
1577 | } |
1578 | |
1579 | let mut g = BackshiftOnDrop { |
1580 | v: self, |
1581 | processed_len: 0, |
1582 | deleted_cnt: 0, |
1583 | original_len, |
1584 | }; |
1585 | |
1586 | fn process_loop<F, T, A: Allocator, const DELETED: bool>( |
1587 | original_len: usize, |
1588 | f: &mut F, |
1589 | g: &mut BackshiftOnDrop<'_, T, A>, |
1590 | ) where |
1591 | F: FnMut(&mut T) -> bool, |
1592 | { |
1593 | while g.processed_len != original_len { |
1594 | // SAFETY: Unchecked element must be valid. |
1595 | let cur = unsafe { &mut *g.v.as_mut_ptr().add(g.processed_len) }; |
1596 | if !f(cur) { |
1597 | // Advance early to avoid double drop if `drop_in_place` panicked. |
1598 | g.processed_len += 1; |
1599 | g.deleted_cnt += 1; |
1600 | // SAFETY: We never touch this element again after dropped. |
1601 | unsafe { ptr::drop_in_place(cur) }; |
1602 | // We already advanced the counter. |
1603 | if DELETED { |
1604 | continue; |
1605 | } else { |
1606 | break; |
1607 | } |
1608 | } |
1609 | if DELETED { |
1610 | // SAFETY: `deleted_cnt` > 0, so the hole slot must not overlap with current element. |
1611 | // We use copy for move, and never touch this element again. |
1612 | unsafe { |
1613 | let hole_slot = g.v.as_mut_ptr().add(g.processed_len - g.deleted_cnt); |
1614 | ptr::copy_nonoverlapping(cur, hole_slot, 1); |
1615 | } |
1616 | } |
1617 | g.processed_len += 1; |
1618 | } |
1619 | } |
1620 | |
1621 | // Stage 1: Nothing was deleted. |
1622 | process_loop::<F, T, A, false>(original_len, &mut f, &mut g); |
1623 | |
1624 | // Stage 2: Some elements were deleted. |
1625 | process_loop::<F, T, A, true>(original_len, &mut f, &mut g); |
1626 | |
1627 | // All item are processed. This can be optimized to `set_len` by LLVM. |
1628 | drop(g); |
1629 | } |
1630 | |
1631 | /// Removes all but the first of consecutive elements in the vector that resolve to the same |
1632 | /// key. |
1633 | /// |
1634 | /// If the vector is sorted, this removes all duplicates. |
1635 | /// |
1636 | /// # Examples |
1637 | /// |
1638 | /// ``` |
1639 | /// let mut vec = vec![10, 20, 21, 30, 20]; |
1640 | /// |
1641 | /// vec.dedup_by_key(|i| *i / 10); |
1642 | /// |
1643 | /// assert_eq!(vec, [10, 20, 30, 20]); |
1644 | /// ``` |
1645 | #[inline (always)] |
1646 | pub fn dedup_by_key<F, K>(&mut self, mut key: F) |
1647 | where |
1648 | F: FnMut(&mut T) -> K, |
1649 | K: PartialEq, |
1650 | { |
1651 | self.dedup_by(|a, b| key(a) == key(b)) |
1652 | } |
1653 | |
1654 | /// Removes all but the first of consecutive elements in the vector satisfying a given equality |
1655 | /// relation. |
1656 | /// |
1657 | /// The `same_bucket` function is passed references to two elements from the vector and |
1658 | /// must determine if the elements compare equal. The elements are passed in opposite order |
1659 | /// from their order in the slice, so if `same_bucket(a, b)` returns `true`, `a` is removed. |
1660 | /// |
1661 | /// If the vector is sorted, this removes all duplicates. |
1662 | /// |
1663 | /// # Examples |
1664 | /// |
1665 | /// ``` |
1666 | /// let mut vec = vec!["foo" , "bar" , "Bar" , "baz" , "bar" ]; |
1667 | /// |
1668 | /// vec.dedup_by(|a, b| a.eq_ignore_ascii_case(b)); |
1669 | /// |
1670 | /// assert_eq!(vec, ["foo" , "bar" , "baz" , "bar" ]); |
1671 | /// ``` |
1672 | #[inline ] |
1673 | pub fn dedup_by<F>(&mut self, mut same_bucket: F) |
1674 | where |
1675 | F: FnMut(&mut T, &mut T) -> bool, |
1676 | { |
1677 | let len = self.len(); |
1678 | if len <= 1 { |
1679 | return; |
1680 | } |
1681 | |
1682 | /* INVARIANT: vec.len() > read >= write > write-1 >= 0 */ |
1683 | struct FillGapOnDrop<'a, T, A: Allocator> { |
1684 | /* Offset of the element we want to check if it is duplicate */ |
1685 | read: usize, |
1686 | |
1687 | /* Offset of the place where we want to place the non-duplicate |
1688 | * when we find it. */ |
1689 | write: usize, |
1690 | |
1691 | /* The Vec that would need correction if `same_bucket` panicked */ |
1692 | vec: &'a mut Vec<T, A>, |
1693 | } |
1694 | |
1695 | impl<'a, T, A: Allocator> Drop for FillGapOnDrop<'a, T, A> { |
1696 | fn drop(&mut self) { |
1697 | /* This code gets executed when `same_bucket` panics */ |
1698 | |
1699 | /* SAFETY: invariant guarantees that `read - write` |
1700 | * and `len - read` never overflow and that the copy is always |
1701 | * in-bounds. */ |
1702 | unsafe { |
1703 | let ptr = self.vec.as_mut_ptr(); |
1704 | let len = self.vec.len(); |
1705 | |
1706 | /* How many items were left when `same_bucket` panicked. |
1707 | * Basically vec[read..].len() */ |
1708 | let items_left = len.wrapping_sub(self.read); |
1709 | |
1710 | /* Pointer to first item in vec[write..write+items_left] slice */ |
1711 | let dropped_ptr = ptr.add(self.write); |
1712 | /* Pointer to first item in vec[read..] slice */ |
1713 | let valid_ptr = ptr.add(self.read); |
1714 | |
1715 | /* Copy `vec[read..]` to `vec[write..write+items_left]`. |
1716 | * The slices can overlap, so `copy_nonoverlapping` cannot be used */ |
1717 | ptr::copy(valid_ptr, dropped_ptr, items_left); |
1718 | |
1719 | /* How many items have been already dropped |
1720 | * Basically vec[read..write].len() */ |
1721 | let dropped = self.read.wrapping_sub(self.write); |
1722 | |
1723 | self.vec.set_len(len - dropped); |
1724 | } |
1725 | } |
1726 | } |
1727 | |
1728 | let mut gap = FillGapOnDrop { |
1729 | read: 1, |
1730 | write: 1, |
1731 | vec: self, |
1732 | }; |
1733 | let ptr = gap.vec.as_mut_ptr(); |
1734 | |
1735 | /* Drop items while going through Vec, it should be more efficient than |
1736 | * doing slice partition_dedup + truncate */ |
1737 | |
1738 | /* SAFETY: Because of the invariant, read_ptr, prev_ptr and write_ptr |
1739 | * are always in-bounds and read_ptr never aliases prev_ptr */ |
1740 | unsafe { |
1741 | while gap.read < len { |
1742 | let read_ptr = ptr.add(gap.read); |
1743 | let prev_ptr = ptr.add(gap.write.wrapping_sub(1)); |
1744 | |
1745 | if same_bucket(&mut *read_ptr, &mut *prev_ptr) { |
1746 | // Increase `gap.read` now since the drop may panic. |
1747 | gap.read += 1; |
1748 | /* We have found duplicate, drop it in-place */ |
1749 | ptr::drop_in_place(read_ptr); |
1750 | } else { |
1751 | let write_ptr = ptr.add(gap.write); |
1752 | |
1753 | /* Because `read_ptr` can be equal to `write_ptr`, we either |
1754 | * have to use `copy` or conditional `copy_nonoverlapping`. |
1755 | * Looks like the first option is faster. */ |
1756 | ptr::copy(read_ptr, write_ptr, 1); |
1757 | |
1758 | /* We have filled that place, so go further */ |
1759 | gap.write += 1; |
1760 | gap.read += 1; |
1761 | } |
1762 | } |
1763 | |
1764 | /* Technically we could let `gap` clean up with its Drop, but |
1765 | * when `same_bucket` is guaranteed to not panic, this bloats a little |
1766 | * the codegen, so we just do it manually */ |
1767 | gap.vec.set_len(gap.write); |
1768 | mem::forget(gap); |
1769 | } |
1770 | } |
1771 | |
1772 | /// Appends an element to the back of a collection. |
1773 | /// |
1774 | /// # Panics |
1775 | /// |
1776 | /// Panics if the new capacity exceeds `isize::MAX` bytes. |
1777 | /// |
1778 | /// # Examples |
1779 | /// |
1780 | /// ``` |
1781 | /// let mut vec = vec![1, 2]; |
1782 | /// vec.push(3); |
1783 | /// assert_eq!(vec, [1, 2, 3]); |
1784 | /// ``` |
1785 | #[cfg (not(no_global_oom_handling))] |
1786 | #[inline (always)] |
1787 | pub fn push(&mut self, value: T) { |
1788 | // This will panic or abort if we would allocate > isize::MAX bytes |
1789 | // or if the length increment would overflow for zero-sized types. |
1790 | if self.len == self.buf.capacity() { |
1791 | self.buf.reserve_for_push(self.len); |
1792 | } |
1793 | unsafe { |
1794 | let end = self.as_mut_ptr().add(self.len); |
1795 | ptr::write(end, value); |
1796 | self.len += 1; |
1797 | } |
1798 | } |
1799 | |
1800 | /// Appends an element if there is sufficient spare capacity, otherwise an error is returned |
1801 | /// with the element. |
1802 | /// |
1803 | /// Unlike [`push`] this method will not reallocate when there's insufficient capacity. |
1804 | /// The caller should use [`reserve`] or [`try_reserve`] to ensure that there is enough capacity. |
1805 | /// |
1806 | /// [`push`]: Vec::push |
1807 | /// [`reserve`]: Vec::reserve |
1808 | /// [`try_reserve`]: Vec::try_reserve |
1809 | /// |
1810 | /// # Examples |
1811 | /// |
1812 | /// A manual, panic-free alternative to [`FromIterator`]: |
1813 | /// |
1814 | /// ``` |
1815 | /// #![feature(vec_push_within_capacity)] |
1816 | /// |
1817 | /// use std::collections::TryReserveError; |
1818 | /// fn from_iter_fallible<T>(iter: impl Iterator<Item=T>) -> Result<Vec<T>, TryReserveError> { |
1819 | /// let mut vec = Vec::new(); |
1820 | /// for value in iter { |
1821 | /// if let Err(value) = vec.push_within_capacity(value) { |
1822 | /// vec.try_reserve(1)?; |
1823 | /// // this cannot fail, the previous line either returned or added at least 1 free slot |
1824 | /// let _ = vec.push_within_capacity(value); |
1825 | /// } |
1826 | /// } |
1827 | /// Ok(vec) |
1828 | /// } |
1829 | /// assert_eq!(from_iter_fallible(0..100), Ok(Vec::from_iter(0..100))); |
1830 | /// ``` |
1831 | #[inline (always)] |
1832 | pub fn push_within_capacity(&mut self, value: T) -> Result<(), T> { |
1833 | if self.len == self.buf.capacity() { |
1834 | return Err(value); |
1835 | } |
1836 | unsafe { |
1837 | let end = self.as_mut_ptr().add(self.len); |
1838 | ptr::write(end, value); |
1839 | self.len += 1; |
1840 | } |
1841 | Ok(()) |
1842 | } |
1843 | |
1844 | /// Removes the last element from a vector and returns it, or [`None`] if it |
1845 | /// is empty. |
1846 | /// |
1847 | /// If you'd like to pop the first element, consider using |
1848 | /// [`VecDeque::pop_front`] instead. |
1849 | /// |
1850 | /// [`VecDeque::pop_front`]: alloc_crate::collections::VecDeque::pop_front |
1851 | /// |
1852 | /// # Examples |
1853 | /// |
1854 | /// ``` |
1855 | /// let mut vec = vec![1, 2, 3]; |
1856 | /// assert_eq!(vec.pop(), Some(3)); |
1857 | /// assert_eq!(vec, [1, 2]); |
1858 | /// ``` |
1859 | #[inline (always)] |
1860 | pub fn pop(&mut self) -> Option<T> { |
1861 | if self.len == 0 { |
1862 | None |
1863 | } else { |
1864 | unsafe { |
1865 | self.len -= 1; |
1866 | Some(ptr::read(self.as_ptr().add(self.len()))) |
1867 | } |
1868 | } |
1869 | } |
1870 | |
1871 | /// Moves all the elements of `other` into `self`, leaving `other` empty. |
1872 | /// |
1873 | /// # Panics |
1874 | /// |
1875 | /// Panics if the new capacity exceeds `isize::MAX` bytes. |
1876 | /// |
1877 | /// # Examples |
1878 | /// |
1879 | /// ``` |
1880 | /// let mut vec = vec![1, 2, 3]; |
1881 | /// let mut vec2 = vec![4, 5, 6]; |
1882 | /// vec.append(&mut vec2); |
1883 | /// assert_eq!(vec, [1, 2, 3, 4, 5, 6]); |
1884 | /// assert_eq!(vec2, []); |
1885 | /// ``` |
1886 | #[cfg (not(no_global_oom_handling))] |
1887 | #[inline (always)] |
1888 | pub fn append(&mut self, other: &mut Self) { |
1889 | unsafe { |
1890 | self.append_elements(other.as_slice() as _); |
1891 | other.set_len(0); |
1892 | } |
1893 | } |
1894 | |
1895 | /// Appends elements to `self` from other buffer. |
1896 | #[cfg (not(no_global_oom_handling))] |
1897 | #[inline (always)] |
1898 | unsafe fn append_elements(&mut self, other: *const [T]) { |
1899 | let count = unsafe { (*other).len() }; |
1900 | self.reserve(count); |
1901 | let len = self.len(); |
1902 | unsafe { ptr::copy_nonoverlapping(other as *const T, self.as_mut_ptr().add(len), count) }; |
1903 | self.len += count; |
1904 | } |
1905 | |
1906 | /// Removes the specified range from the vector in bulk, returning all |
1907 | /// removed elements as an iterator. If the iterator is dropped before |
1908 | /// being fully consumed, it drops the remaining removed elements. |
1909 | /// |
1910 | /// The returned iterator keeps a mutable borrow on the vector to optimize |
1911 | /// its implementation. |
1912 | /// |
1913 | /// # Panics |
1914 | /// |
1915 | /// Panics if the starting point is greater than the end point or if |
1916 | /// the end point is greater than the length of the vector. |
1917 | /// |
1918 | /// # Leaking |
1919 | /// |
1920 | /// If the returned iterator goes out of scope without being dropped (due to |
1921 | /// [`mem::forget`], for example), the vector may have lost and leaked |
1922 | /// elements arbitrarily, including elements outside the range. |
1923 | /// |
1924 | /// # Examples |
1925 | /// |
1926 | /// ``` |
1927 | /// let mut v = vec![1, 2, 3]; |
1928 | /// let u: Vec<_> = v.drain(1..).collect(); |
1929 | /// assert_eq!(v, &[1]); |
1930 | /// assert_eq!(u, &[2, 3]); |
1931 | /// |
1932 | /// // A full range clears the vector, like `clear()` does |
1933 | /// v.drain(..); |
1934 | /// assert_eq!(v, &[]); |
1935 | /// ``` |
1936 | #[inline (always)] |
1937 | pub fn drain<R>(&mut self, range: R) -> Drain<'_, T, A> |
1938 | where |
1939 | R: RangeBounds<usize>, |
1940 | { |
1941 | // Memory safety |
1942 | // |
1943 | // When the Drain is first created, it shortens the length of |
1944 | // the source vector to make sure no uninitialized or moved-from elements |
1945 | // are accessible at all if the Drain's destructor never gets to run. |
1946 | // |
1947 | // Drain will ptr::read out the values to remove. |
1948 | // When finished, remaining tail of the vec is copied back to cover |
1949 | // the hole, and the vector length is restored to the new length. |
1950 | // |
1951 | let len = self.len(); |
1952 | |
1953 | // Replaced by code below |
1954 | // let Range { start, end } = slice::range(range, ..len); |
1955 | |
1956 | // Panics if range is out of bounds |
1957 | let _ = &self.as_slice()[(range.start_bound().cloned(), range.end_bound().cloned())]; |
1958 | |
1959 | let start = match range.start_bound() { |
1960 | Bound::Included(&n) => n, |
1961 | Bound::Excluded(&n) => n + 1, |
1962 | Bound::Unbounded => 0, |
1963 | }; |
1964 | let end = match range.end_bound() { |
1965 | Bound::Included(&n) => n + 1, |
1966 | Bound::Excluded(&n) => n, |
1967 | Bound::Unbounded => len, |
1968 | }; |
1969 | |
1970 | unsafe { |
1971 | // set self.vec length's to start, to be safe in case Drain is leaked |
1972 | self.set_len(start); |
1973 | let range_slice = slice::from_raw_parts(self.as_ptr().add(start), end - start); |
1974 | Drain { |
1975 | tail_start: end, |
1976 | tail_len: len - end, |
1977 | iter: range_slice.iter(), |
1978 | vec: NonNull::from(self), |
1979 | } |
1980 | } |
1981 | } |
1982 | |
1983 | /// Clears the vector, removing all values. |
1984 | /// |
1985 | /// Note that this method has no effect on the allocated capacity |
1986 | /// of the vector. |
1987 | /// |
1988 | /// # Examples |
1989 | /// |
1990 | /// ``` |
1991 | /// let mut v = vec![1, 2, 3]; |
1992 | /// |
1993 | /// v.clear(); |
1994 | /// |
1995 | /// assert!(v.is_empty()); |
1996 | /// ``` |
1997 | #[inline (always)] |
1998 | pub fn clear(&mut self) { |
1999 | let elems: *mut [T] = self.as_mut_slice(); |
2000 | |
2001 | // SAFETY: |
2002 | // - `elems` comes directly from `as_mut_slice` and is therefore valid. |
2003 | // - Setting `self.len` before calling `drop_in_place` means that, |
2004 | // if an element's `Drop` impl panics, the vector's `Drop` impl will |
2005 | // do nothing (leaking the rest of the elements) instead of dropping |
2006 | // some twice. |
2007 | unsafe { |
2008 | self.len = 0; |
2009 | ptr::drop_in_place(elems); |
2010 | } |
2011 | } |
2012 | |
2013 | /// Returns the number of elements in the vector, also referred to |
2014 | /// as its 'length'. |
2015 | /// |
2016 | /// # Examples |
2017 | /// |
2018 | /// ``` |
2019 | /// let a = vec![1, 2, 3]; |
2020 | /// assert_eq!(a.len(), 3); |
2021 | /// ``` |
2022 | #[inline (always)] |
2023 | pub fn len(&self) -> usize { |
2024 | self.len |
2025 | } |
2026 | |
2027 | /// Returns `true` if the vector contains no elements. |
2028 | /// |
2029 | /// # Examples |
2030 | /// |
2031 | /// ``` |
2032 | /// let mut v = Vec::new(); |
2033 | /// assert!(v.is_empty()); |
2034 | /// |
2035 | /// v.push(1); |
2036 | /// assert!(!v.is_empty()); |
2037 | /// ``` |
2038 | #[inline (always)] |
2039 | pub fn is_empty(&self) -> bool { |
2040 | self.len() == 0 |
2041 | } |
2042 | |
2043 | /// Splits the collection into two at the given index. |
2044 | /// |
2045 | /// Returns a newly allocated vector containing the elements in the range |
2046 | /// `[at, len)`. After the call, the original vector will be left containing |
2047 | /// the elements `[0, at)` with its previous capacity unchanged. |
2048 | /// |
2049 | /// # Panics |
2050 | /// |
2051 | /// Panics if `at > len`. |
2052 | /// |
2053 | /// # Examples |
2054 | /// |
2055 | /// ``` |
2056 | /// let mut vec = vec![1, 2, 3]; |
2057 | /// let vec2 = vec.split_off(1); |
2058 | /// assert_eq!(vec, [1]); |
2059 | /// assert_eq!(vec2, [2, 3]); |
2060 | /// ``` |
2061 | #[cfg (not(no_global_oom_handling))] |
2062 | #[inline (always)] |
2063 | #[must_use = "use `.truncate()` if you don't need the other half" ] |
2064 | pub fn split_off(&mut self, at: usize) -> Self |
2065 | where |
2066 | A: Clone, |
2067 | { |
2068 | #[cold ] |
2069 | #[inline (never)] |
2070 | fn assert_failed(at: usize, len: usize) -> ! { |
2071 | panic!("`at` split index (is {}) should be <= len (is {})" , at, len); |
2072 | } |
2073 | |
2074 | if at > self.len() { |
2075 | assert_failed(at, self.len()); |
2076 | } |
2077 | |
2078 | if at == 0 { |
2079 | // the new vector can take over the original buffer and avoid the copy |
2080 | return mem::replace( |
2081 | self, |
2082 | Vec::with_capacity_in(self.capacity(), self.allocator().clone()), |
2083 | ); |
2084 | } |
2085 | |
2086 | let other_len = self.len - at; |
2087 | let mut other = Vec::with_capacity_in(other_len, self.allocator().clone()); |
2088 | |
2089 | // Unsafely `set_len` and copy items to `other`. |
2090 | unsafe { |
2091 | self.set_len(at); |
2092 | other.set_len(other_len); |
2093 | |
2094 | ptr::copy_nonoverlapping(self.as_ptr().add(at), other.as_mut_ptr(), other.len()); |
2095 | } |
2096 | other |
2097 | } |
2098 | |
2099 | /// Resizes the `Vec` in-place so that `len` is equal to `new_len`. |
2100 | /// |
2101 | /// If `new_len` is greater than `len`, the `Vec` is extended by the |
2102 | /// difference, with each additional slot filled with the result of |
2103 | /// calling the closure `f`. The return values from `f` will end up |
2104 | /// in the `Vec` in the order they have been generated. |
2105 | /// |
2106 | /// If `new_len` is less than `len`, the `Vec` is simply truncated. |
2107 | /// |
2108 | /// This method uses a closure to create new values on every push. If |
2109 | /// you'd rather [`Clone`] a given value, use [`Vec::resize`]. If you |
2110 | /// want to use the [`Default`] trait to generate values, you can |
2111 | /// pass [`Default::default`] as the second argument. |
2112 | /// |
2113 | /// # Examples |
2114 | /// |
2115 | /// ``` |
2116 | /// let mut vec = vec![1, 2, 3]; |
2117 | /// vec.resize_with(5, Default::default); |
2118 | /// assert_eq!(vec, [1, 2, 3, 0, 0]); |
2119 | /// |
2120 | /// let mut vec = vec![]; |
2121 | /// let mut p = 1; |
2122 | /// vec.resize_with(4, || { p *= 2; p }); |
2123 | /// assert_eq!(vec, [2, 4, 8, 16]); |
2124 | /// ``` |
2125 | #[cfg (not(no_global_oom_handling))] |
2126 | #[inline (always)] |
2127 | pub fn resize_with<F>(&mut self, new_len: usize, f: F) |
2128 | where |
2129 | F: FnMut() -> T, |
2130 | { |
2131 | let len = self.len(); |
2132 | if new_len > len { |
2133 | self.extend(iter::repeat_with(f).take(new_len - len)); |
2134 | } else { |
2135 | self.truncate(new_len); |
2136 | } |
2137 | } |
2138 | |
2139 | /// Consumes and leaks the `Vec`, returning a mutable reference to the contents, |
2140 | /// `&'a mut [T]`. Note that the type `T` must outlive the chosen lifetime |
2141 | /// `'a`. If the type has only static references, or none at all, then this |
2142 | /// may be chosen to be `'static`. |
2143 | /// |
2144 | /// As of Rust 1.57, this method does not reallocate or shrink the `Vec`, |
2145 | /// so the leaked allocation may include unused capacity that is not part |
2146 | /// of the returned slice. |
2147 | /// |
2148 | /// This function is mainly useful for data that lives for the remainder of |
2149 | /// the program's life. Dropping the returned reference will cause a memory |
2150 | /// leak. |
2151 | /// |
2152 | /// # Examples |
2153 | /// |
2154 | /// Simple usage: |
2155 | /// |
2156 | /// ``` |
2157 | /// let x = vec![1, 2, 3]; |
2158 | /// let static_ref: &'static mut [usize] = x.leak(); |
2159 | /// static_ref[0] += 1; |
2160 | /// assert_eq!(static_ref, &[2, 2, 3]); |
2161 | /// ``` |
2162 | #[inline (always)] |
2163 | pub fn leak<'a>(self) -> &'a mut [T] |
2164 | where |
2165 | A: 'a, |
2166 | { |
2167 | let mut me = ManuallyDrop::new(self); |
2168 | unsafe { slice::from_raw_parts_mut(me.as_mut_ptr(), me.len) } |
2169 | } |
2170 | |
2171 | /// Returns the remaining spare capacity of the vector as a slice of |
2172 | /// `MaybeUninit<T>`. |
2173 | /// |
2174 | /// The returned slice can be used to fill the vector with data (e.g. by |
2175 | /// reading from a file) before marking the data as initialized using the |
2176 | /// [`set_len`] method. |
2177 | /// |
2178 | /// [`set_len`]: Vec::set_len |
2179 | /// |
2180 | /// # Examples |
2181 | /// |
2182 | /// ``` |
2183 | /// // Allocate vector big enough for 10 elements. |
2184 | /// let mut v = Vec::with_capacity(10); |
2185 | /// |
2186 | /// // Fill in the first 3 elements. |
2187 | /// let uninit = v.spare_capacity_mut(); |
2188 | /// uninit[0].write(0); |
2189 | /// uninit[1].write(1); |
2190 | /// uninit[2].write(2); |
2191 | /// |
2192 | /// // Mark the first 3 elements of the vector as being initialized. |
2193 | /// unsafe { |
2194 | /// v.set_len(3); |
2195 | /// } |
2196 | /// |
2197 | /// assert_eq!(&v, &[0, 1, 2]); |
2198 | /// ``` |
2199 | #[inline (always)] |
2200 | pub fn spare_capacity_mut(&mut self) -> &mut [MaybeUninit<T>] { |
2201 | // Note: |
2202 | // This method is not implemented in terms of `split_at_spare_mut`, |
2203 | // to prevent invalidation of pointers to the buffer. |
2204 | unsafe { |
2205 | slice::from_raw_parts_mut( |
2206 | self.as_mut_ptr().add(self.len) as *mut MaybeUninit<T>, |
2207 | self.buf.capacity() - self.len, |
2208 | ) |
2209 | } |
2210 | } |
2211 | |
2212 | /// Returns vector content as a slice of `T`, along with the remaining spare |
2213 | /// capacity of the vector as a slice of `MaybeUninit<T>`. |
2214 | /// |
2215 | /// The returned spare capacity slice can be used to fill the vector with data |
2216 | /// (e.g. by reading from a file) before marking the data as initialized using |
2217 | /// the [`set_len`] method. |
2218 | /// |
2219 | /// [`set_len`]: Vec::set_len |
2220 | /// |
2221 | /// Note that this is a low-level API, which should be used with care for |
2222 | /// optimization purposes. If you need to append data to a `Vec` |
2223 | /// you can use [`push`], [`extend`], [`extend_from_slice`], |
2224 | /// [`extend_from_within`], [`insert`], [`append`], [`resize`] or |
2225 | /// [`resize_with`], depending on your exact needs. |
2226 | /// |
2227 | /// [`push`]: Vec::push |
2228 | /// [`extend`]: Vec::extend |
2229 | /// [`extend_from_slice`]: Vec::extend_from_slice |
2230 | /// [`extend_from_within`]: Vec::extend_from_within |
2231 | /// [`insert`]: Vec::insert |
2232 | /// [`append`]: Vec::append |
2233 | /// [`resize`]: Vec::resize |
2234 | /// [`resize_with`]: Vec::resize_with |
2235 | /// |
2236 | /// # Examples |
2237 | /// |
2238 | /// ``` |
2239 | /// #![feature(vec_split_at_spare)] |
2240 | /// |
2241 | /// let mut v = vec![1, 1, 2]; |
2242 | /// |
2243 | /// // Reserve additional space big enough for 10 elements. |
2244 | /// v.reserve(10); |
2245 | /// |
2246 | /// let (init, uninit) = v.split_at_spare_mut(); |
2247 | /// let sum = init.iter().copied().sum::<u32>(); |
2248 | /// |
2249 | /// // Fill in the next 4 elements. |
2250 | /// uninit[0].write(sum); |
2251 | /// uninit[1].write(sum * 2); |
2252 | /// uninit[2].write(sum * 3); |
2253 | /// uninit[3].write(sum * 4); |
2254 | /// |
2255 | /// // Mark the 4 elements of the vector as being initialized. |
2256 | /// unsafe { |
2257 | /// let len = v.len(); |
2258 | /// v.set_len(len + 4); |
2259 | /// } |
2260 | /// |
2261 | /// assert_eq!(&v, &[1, 1, 2, 4, 8, 12, 16]); |
2262 | /// ``` |
2263 | #[inline (always)] |
2264 | pub fn split_at_spare_mut(&mut self) -> (&mut [T], &mut [MaybeUninit<T>]) { |
2265 | // SAFETY: |
2266 | // - len is ignored and so never changed |
2267 | let (init, spare, _) = unsafe { self.split_at_spare_mut_with_len() }; |
2268 | (init, spare) |
2269 | } |
2270 | |
2271 | /// Safety: changing returned .2 (&mut usize) is considered the same as calling `.set_len(_)`. |
2272 | /// |
2273 | /// This method provides unique access to all vec parts at once in `extend_from_within`. |
2274 | unsafe fn split_at_spare_mut_with_len( |
2275 | &mut self, |
2276 | ) -> (&mut [T], &mut [MaybeUninit<T>], &mut usize) { |
2277 | let ptr = self.as_mut_ptr(); |
2278 | // SAFETY: |
2279 | // - `ptr` is guaranteed to be valid for `self.len` elements |
2280 | // - but the allocation extends out to `self.buf.capacity()` elements, possibly |
2281 | // uninitialized |
2282 | let spare_ptr = unsafe { ptr.add(self.len) }; |
2283 | let spare_ptr = spare_ptr.cast::<MaybeUninit<T>>(); |
2284 | let spare_len = self.buf.capacity() - self.len; |
2285 | |
2286 | // SAFETY: |
2287 | // - `ptr` is guaranteed to be valid for `self.len` elements |
2288 | // - `spare_ptr` is pointing one element past the buffer, so it doesn't overlap with `initialized` |
2289 | unsafe { |
2290 | let initialized = slice::from_raw_parts_mut(ptr, self.len); |
2291 | let spare = slice::from_raw_parts_mut(spare_ptr, spare_len); |
2292 | |
2293 | (initialized, spare, &mut self.len) |
2294 | } |
2295 | } |
2296 | } |
2297 | |
2298 | impl<T: Clone, A: Allocator> Vec<T, A> { |
2299 | /// Resizes the `Vec` in-place so that `len` is equal to `new_len`. |
2300 | /// |
2301 | /// If `new_len` is greater than `len`, the `Vec` is extended by the |
2302 | /// difference, with each additional slot filled with `value`. |
2303 | /// If `new_len` is less than `len`, the `Vec` is simply truncated. |
2304 | /// |
2305 | /// This method requires `T` to implement [`Clone`], |
2306 | /// in order to be able to clone the passed value. |
2307 | /// If you need more flexibility (or want to rely on [`Default`] instead of |
2308 | /// [`Clone`]), use [`Vec::resize_with`]. |
2309 | /// If you only need to resize to a smaller size, use [`Vec::truncate`]. |
2310 | /// |
2311 | /// # Examples |
2312 | /// |
2313 | /// ``` |
2314 | /// let mut vec = vec!["hello" ]; |
2315 | /// vec.resize(3, "world" ); |
2316 | /// assert_eq!(vec, ["hello" , "world" , "world" ]); |
2317 | /// |
2318 | /// let mut vec = vec![1, 2, 3, 4]; |
2319 | /// vec.resize(2, 0); |
2320 | /// assert_eq!(vec, [1, 2]); |
2321 | /// ``` |
2322 | #[cfg (not(no_global_oom_handling))] |
2323 | #[inline (always)] |
2324 | pub fn resize(&mut self, new_len: usize, value: T) { |
2325 | let len = self.len(); |
2326 | |
2327 | if new_len > len { |
2328 | self.extend_with(new_len - len, ExtendElement(value)) |
2329 | } else { |
2330 | self.truncate(new_len); |
2331 | } |
2332 | } |
2333 | |
2334 | /// Clones and appends all elements in a slice to the `Vec`. |
2335 | /// |
2336 | /// Iterates over the slice `other`, clones each element, and then appends |
2337 | /// it to this `Vec`. The `other` slice is traversed in-order. |
2338 | /// |
2339 | /// Note that this function is same as [`extend`] except that it is |
2340 | /// specialized to work with slices instead. If and when Rust gets |
2341 | /// specialization this function will likely be deprecated (but still |
2342 | /// available). |
2343 | /// |
2344 | /// # Examples |
2345 | /// |
2346 | /// ``` |
2347 | /// let mut vec = vec![1]; |
2348 | /// vec.extend_from_slice(&[2, 3, 4]); |
2349 | /// assert_eq!(vec, [1, 2, 3, 4]); |
2350 | /// ``` |
2351 | /// |
2352 | /// [`extend`]: Vec::extend |
2353 | #[cfg (not(no_global_oom_handling))] |
2354 | #[inline (always)] |
2355 | pub fn extend_from_slice(&mut self, other: &[T]) { |
2356 | self.extend(other.iter().cloned()) |
2357 | } |
2358 | |
2359 | /// Copies elements from `src` range to the end of the vector. |
2360 | /// |
2361 | /// # Panics |
2362 | /// |
2363 | /// Panics if the starting point is greater than the end point or if |
2364 | /// the end point is greater than the length of the vector. |
2365 | /// |
2366 | /// # Examples |
2367 | /// |
2368 | /// ``` |
2369 | /// let mut vec = vec![0, 1, 2, 3, 4]; |
2370 | /// |
2371 | /// vec.extend_from_within(2..); |
2372 | /// assert_eq!(vec, [0, 1, 2, 3, 4, 2, 3, 4]); |
2373 | /// |
2374 | /// vec.extend_from_within(..2); |
2375 | /// assert_eq!(vec, [0, 1, 2, 3, 4, 2, 3, 4, 0, 1]); |
2376 | /// |
2377 | /// vec.extend_from_within(4..8); |
2378 | /// assert_eq!(vec, [0, 1, 2, 3, 4, 2, 3, 4, 0, 1, 4, 2, 3, 4]); |
2379 | /// ``` |
2380 | #[cfg (not(no_global_oom_handling))] |
2381 | #[inline (always)] |
2382 | pub fn extend_from_within<R>(&mut self, src: R) |
2383 | where |
2384 | R: RangeBounds<usize>, |
2385 | { |
2386 | // let range = slice::range(src, ..self.len()); |
2387 | |
2388 | let _ = &self.as_slice()[(src.start_bound().cloned(), src.end_bound().cloned())]; |
2389 | |
2390 | let len = self.len(); |
2391 | |
2392 | let start: ops::Bound<&usize> = src.start_bound(); |
2393 | let start = match start { |
2394 | ops::Bound::Included(&start) => start, |
2395 | ops::Bound::Excluded(start) => start + 1, |
2396 | ops::Bound::Unbounded => 0, |
2397 | }; |
2398 | |
2399 | let end: ops::Bound<&usize> = src.end_bound(); |
2400 | let end = match end { |
2401 | ops::Bound::Included(end) => end + 1, |
2402 | ops::Bound::Excluded(&end) => end, |
2403 | ops::Bound::Unbounded => len, |
2404 | }; |
2405 | |
2406 | let range = start..end; |
2407 | |
2408 | self.reserve(range.len()); |
2409 | |
2410 | // SAFETY: |
2411 | // - len is increased only after initializing elements |
2412 | let (this, spare, len) = unsafe { self.split_at_spare_mut_with_len() }; |
2413 | |
2414 | // SAFETY: |
2415 | // - caller guarantees that src is a valid index |
2416 | let to_clone = unsafe { this.get_unchecked(range) }; |
2417 | |
2418 | iter::zip(to_clone, spare) |
2419 | .map(|(src, dst)| dst.write(src.clone())) |
2420 | // Note: |
2421 | // - Element was just initialized with `MaybeUninit::write`, so it's ok to increase len |
2422 | // - len is increased after each element to prevent leaks (see issue #82533) |
2423 | .for_each(|_| *len += 1); |
2424 | } |
2425 | } |
2426 | |
2427 | impl<T, A: Allocator, const N: usize> Vec<[T; N], A> { |
2428 | /// Takes a `Vec<[T; N]>` and flattens it into a `Vec<T>`. |
2429 | /// |
2430 | /// # Panics |
2431 | /// |
2432 | /// Panics if the length of the resulting vector would overflow a `usize`. |
2433 | /// |
2434 | /// This is only possible when flattening a vector of arrays of zero-sized |
2435 | /// types, and thus tends to be irrelevant in practice. If |
2436 | /// `size_of::<T>() > 0`, this will never panic. |
2437 | /// |
2438 | /// # Examples |
2439 | /// |
2440 | /// ``` |
2441 | /// #![feature(slice_flatten)] |
2442 | /// |
2443 | /// let mut vec = vec![[1, 2, 3], [4, 5, 6], [7, 8, 9]]; |
2444 | /// assert_eq!(vec.pop(), Some([7, 8, 9])); |
2445 | /// |
2446 | /// let mut flattened = vec.into_flattened(); |
2447 | /// assert_eq!(flattened.pop(), Some(6)); |
2448 | /// ``` |
2449 | #[inline (always)] |
2450 | pub fn into_flattened(self) -> Vec<T, A> { |
2451 | let (ptr, len, cap, alloc) = self.into_raw_parts_with_alloc(); |
2452 | let (new_len, new_cap) = if size_of::<T>() == 0 { |
2453 | (len.checked_mul(N).expect("vec len overflow" ), usize::MAX) |
2454 | } else { |
2455 | // SAFETY: |
2456 | // - `cap * N` cannot overflow because the allocation is already in |
2457 | // the address space. |
2458 | // - Each `[T; N]` has `N` valid elements, so there are `len * N` |
2459 | // valid elements in the allocation. |
2460 | (len * N, cap * N) |
2461 | }; |
2462 | // SAFETY: |
2463 | // - `ptr` was allocated by `self` |
2464 | // - `ptr` is well-aligned because `[T; N]` has the same alignment as `T`. |
2465 | // - `new_cap` refers to the same sized allocation as `cap` because |
2466 | // `new_cap * size_of::<T>()` == `cap * size_of::<[T; N]>()` |
2467 | // - `len` <= `cap`, so `len * N` <= `cap * N`. |
2468 | unsafe { Vec::<T, A>::from_raw_parts_in(ptr.cast(), new_len, new_cap, alloc) } |
2469 | } |
2470 | } |
2471 | |
2472 | // This code generalizes `extend_with_{element,default}`. |
2473 | trait ExtendWith<T> { |
2474 | fn next(&mut self) -> T; |
2475 | fn last(self) -> T; |
2476 | } |
2477 | |
2478 | struct ExtendElement<T>(T); |
2479 | impl<T: Clone> ExtendWith<T> for ExtendElement<T> { |
2480 | #[inline (always)] |
2481 | fn next(&mut self) -> T { |
2482 | self.0.clone() |
2483 | } |
2484 | |
2485 | #[inline (always)] |
2486 | fn last(self) -> T { |
2487 | self.0 |
2488 | } |
2489 | } |
2490 | |
2491 | impl<T, A: Allocator> Vec<T, A> { |
2492 | #[cfg (not(no_global_oom_handling))] |
2493 | #[inline (always)] |
2494 | /// Extend the vector by `n` values, using the given generator. |
2495 | fn extend_with<E: ExtendWith<T>>(&mut self, n: usize, mut value: E) { |
2496 | self.reserve(n); |
2497 | |
2498 | unsafe { |
2499 | let mut ptr = self.as_mut_ptr().add(self.len()); |
2500 | // Use SetLenOnDrop to work around bug where compiler |
2501 | // might not realize the store through `ptr` through self.set_len() |
2502 | // don't alias. |
2503 | let mut local_len = SetLenOnDrop::new(&mut self.len); |
2504 | |
2505 | // Write all elements except the last one |
2506 | for _ in 1..n { |
2507 | ptr::write(ptr, value.next()); |
2508 | ptr = ptr.add(1); |
2509 | // Increment the length in every step in case next() panics |
2510 | local_len.increment_len(1); |
2511 | } |
2512 | |
2513 | if n > 0 { |
2514 | // We can write the last element directly without cloning needlessly |
2515 | ptr::write(ptr, value.last()); |
2516 | local_len.increment_len(1); |
2517 | } |
2518 | |
2519 | // len set by scope guard |
2520 | } |
2521 | } |
2522 | } |
2523 | |
2524 | impl<T: PartialEq, A: Allocator> Vec<T, A> { |
2525 | /// Removes consecutive repeated elements in the vector according to the |
2526 | /// [`PartialEq`] trait implementation. |
2527 | /// |
2528 | /// If the vector is sorted, this removes all duplicates. |
2529 | /// |
2530 | /// # Examples |
2531 | /// |
2532 | /// ``` |
2533 | /// let mut vec = vec![1, 2, 2, 3, 2]; |
2534 | /// |
2535 | /// vec.dedup(); |
2536 | /// |
2537 | /// assert_eq!(vec, [1, 2, 3, 2]); |
2538 | /// ``` |
2539 | #[inline (always)] |
2540 | pub fn dedup(&mut self) { |
2541 | self.dedup_by(|a: &mut T, b: &mut T| a == b) |
2542 | } |
2543 | } |
2544 | |
2545 | trait ExtendFromWithinSpec { |
2546 | /// # Safety |
2547 | /// |
2548 | /// - `src` needs to be valid index |
2549 | /// - `self.capacity() - self.len()` must be `>= src.len()` |
2550 | unsafe fn spec_extend_from_within(&mut self, src: Range<usize>); |
2551 | } |
2552 | |
2553 | // impl<T: Clone, A: Allocator> ExtendFromWithinSpec for Vec<T, A> { |
2554 | // default unsafe fn spec_extend_from_within(&mut self, src: Range<usize>) { |
2555 | // // SAFETY: |
2556 | // // - len is increased only after initializing elements |
2557 | // let (this, spare, len) = unsafe { self.split_at_spare_mut_with_len() }; |
2558 | |
2559 | // // SAFETY: |
2560 | // // - caller guarantees that src is a valid index |
2561 | // let to_clone = unsafe { this.get_unchecked(src) }; |
2562 | |
2563 | // iter::zip(to_clone, spare) |
2564 | // .map(|(src, dst)| dst.write(src.clone())) |
2565 | // // Note: |
2566 | // // - Element was just initialized with `MaybeUninit::write`, so it's ok to increase len |
2567 | // // - len is increased after each element to prevent leaks (see issue #82533) |
2568 | // .for_each(|_| *len += 1); |
2569 | // } |
2570 | // } |
2571 | |
2572 | impl<T: Copy, A: Allocator> ExtendFromWithinSpec for Vec<T, A> { |
2573 | #[inline (always)] |
2574 | unsafe fn spec_extend_from_within(&mut self, src: Range<usize>) { |
2575 | let count = src.len(); |
2576 | { |
2577 | let (init, spare) = self.split_at_spare_mut(); |
2578 | |
2579 | // SAFETY: |
2580 | // - caller guarantees that `src` is a valid index |
2581 | let source = unsafe { init.get_unchecked(src) }; |
2582 | |
2583 | // SAFETY: |
2584 | // - Both pointers are created from unique slice references (`&mut [_]`) |
2585 | // so they are valid and do not overlap. |
2586 | // - Elements are :Copy so it's OK to copy them, without doing |
2587 | // anything with the original values |
2588 | // - `count` is equal to the len of `source`, so source is valid for |
2589 | // `count` reads |
2590 | // - `.reserve(count)` guarantees that `spare.len() >= count` so spare |
2591 | // is valid for `count` writes |
2592 | unsafe { ptr::copy_nonoverlapping(source.as_ptr(), spare.as_mut_ptr() as _, count) }; |
2593 | } |
2594 | |
2595 | // SAFETY: |
2596 | // - The elements were just initialized by `copy_nonoverlapping` |
2597 | self.len += count; |
2598 | } |
2599 | } |
2600 | |
2601 | //////////////////////////////////////////////////////////////////////////////// |
2602 | // Common trait implementations for Vec |
2603 | //////////////////////////////////////////////////////////////////////////////// |
2604 | |
2605 | impl<T, A: Allocator> ops::Deref for Vec<T, A> { |
2606 | type Target = [T]; |
2607 | |
2608 | #[inline (always)] |
2609 | fn deref(&self) -> &[T] { |
2610 | unsafe { slice::from_raw_parts(self.as_ptr(), self.len) } |
2611 | } |
2612 | } |
2613 | |
2614 | impl<T, A: Allocator> ops::DerefMut for Vec<T, A> { |
2615 | #[inline (always)] |
2616 | fn deref_mut(&mut self) -> &mut [T] { |
2617 | unsafe { slice::from_raw_parts_mut(self.as_mut_ptr(), self.len) } |
2618 | } |
2619 | } |
2620 | |
2621 | #[cfg (not(no_global_oom_handling))] |
2622 | impl<T: Clone, A: Allocator + Clone> Clone for Vec<T, A> { |
2623 | #[inline (always)] |
2624 | fn clone(&self) -> Self { |
2625 | let alloc: A = self.allocator().clone(); |
2626 | let mut vec: Vec = Vec::with_capacity_in(self.len(), alloc); |
2627 | vec.extend_from_slice(self); |
2628 | vec |
2629 | } |
2630 | |
2631 | #[inline (always)] |
2632 | fn clone_from(&mut self, other: &Self) { |
2633 | // drop anything that will not be overwritten |
2634 | self.truncate(other.len()); |
2635 | |
2636 | // self.len <= other.len due to the truncate above, so the |
2637 | // slices here are always in-bounds. |
2638 | let (init: &[T], tail: &[T]) = other.split_at(self.len()); |
2639 | |
2640 | // reuse the contained values' allocations/resources. |
2641 | self.clone_from_slice(src:init); |
2642 | self.extend_from_slice(tail); |
2643 | } |
2644 | } |
2645 | |
2646 | /// The hash of a vector is the same as that of the corresponding slice, |
2647 | /// as required by the `core::borrow::Borrow` implementation. |
2648 | /// |
2649 | /// ``` |
2650 | /// #![feature(build_hasher_simple_hash_one)] |
2651 | /// use std::hash::BuildHasher; |
2652 | /// |
2653 | /// let b = std::collections::hash_map::RandomState::new(); |
2654 | /// let v: Vec<u8> = vec![0xa8, 0x3c, 0x09]; |
2655 | /// let s: &[u8] = &[0xa8, 0x3c, 0x09]; |
2656 | /// assert_eq!(b.hash_one(v), b.hash_one(s)); |
2657 | /// ``` |
2658 | impl<T: Hash, A: Allocator> Hash for Vec<T, A> { |
2659 | #[inline (always)] |
2660 | fn hash<H: Hasher>(&self, state: &mut H) { |
2661 | Hash::hash(&**self, state) |
2662 | } |
2663 | } |
2664 | |
2665 | impl<T, I: SliceIndex<[T]>, A: Allocator> Index<I> for Vec<T, A> { |
2666 | type Output = I::Output; |
2667 | |
2668 | #[inline (always)] |
2669 | fn index(&self, index: I) -> &Self::Output { |
2670 | Index::index(&**self, index) |
2671 | } |
2672 | } |
2673 | |
2674 | impl<T, I: SliceIndex<[T]>, A: Allocator> IndexMut<I> for Vec<T, A> { |
2675 | #[inline (always)] |
2676 | fn index_mut(&mut self, index: I) -> &mut Self::Output { |
2677 | IndexMut::index_mut(&mut **self, index) |
2678 | } |
2679 | } |
2680 | |
2681 | #[cfg (not(no_global_oom_handling))] |
2682 | impl<T> FromIterator<T> for Vec<T> { |
2683 | #[inline (always)] |
2684 | fn from_iter<I: IntoIterator<Item = T>>(iter: I) -> Vec<T> { |
2685 | let mut vec: Vec = Vec::new(); |
2686 | vec.extend(iter); |
2687 | vec |
2688 | } |
2689 | } |
2690 | |
2691 | impl<T, A: Allocator> IntoIterator for Vec<T, A> { |
2692 | type Item = T; |
2693 | type IntoIter = IntoIter<T, A>; |
2694 | |
2695 | /// Creates a consuming iterator, that is, one that moves each value out of |
2696 | /// the vector (from start to end). The vector cannot be used after calling |
2697 | /// this. |
2698 | /// |
2699 | /// # Examples |
2700 | /// |
2701 | /// ``` |
2702 | /// let v = vec!["a" .to_string(), "b" .to_string()]; |
2703 | /// let mut v_iter = v.into_iter(); |
2704 | /// |
2705 | /// let first_element: Option<String> = v_iter.next(); |
2706 | /// |
2707 | /// assert_eq!(first_element, Some("a" .to_string())); |
2708 | /// assert_eq!(v_iter.next(), Some("b" .to_string())); |
2709 | /// assert_eq!(v_iter.next(), None); |
2710 | /// ``` |
2711 | #[inline (always)] |
2712 | fn into_iter(self) -> Self::IntoIter { |
2713 | unsafe { |
2714 | let mut me = ManuallyDrop::new(self); |
2715 | let alloc = ManuallyDrop::new(ptr::read(me.allocator())); |
2716 | let begin = me.as_mut_ptr(); |
2717 | let end = if size_of::<T>() == 0 { |
2718 | begin.cast::<u8>().wrapping_add(me.len()).cast() |
2719 | } else { |
2720 | begin.add(me.len()) as *const T |
2721 | }; |
2722 | let cap = me.buf.capacity(); |
2723 | IntoIter { |
2724 | buf: NonNull::new_unchecked(begin), |
2725 | phantom: PhantomData, |
2726 | cap, |
2727 | alloc, |
2728 | ptr: begin, |
2729 | end, |
2730 | } |
2731 | } |
2732 | } |
2733 | } |
2734 | |
2735 | impl<'a, T, A: Allocator> IntoIterator for &'a Vec<T, A> { |
2736 | type Item = &'a T; |
2737 | type IntoIter = slice::Iter<'a, T>; |
2738 | |
2739 | #[inline (always)] |
2740 | fn into_iter(self) -> Self::IntoIter { |
2741 | self.iter() |
2742 | } |
2743 | } |
2744 | |
2745 | impl<'a, T, A: Allocator> IntoIterator for &'a mut Vec<T, A> { |
2746 | type Item = &'a mut T; |
2747 | type IntoIter = slice::IterMut<'a, T>; |
2748 | |
2749 | fn into_iter(self) -> Self::IntoIter { |
2750 | self.iter_mut() |
2751 | } |
2752 | } |
2753 | |
2754 | #[cfg (not(no_global_oom_handling))] |
2755 | impl<T, A: Allocator> Extend<T> for Vec<T, A> { |
2756 | #[inline (always)] |
2757 | fn extend<I: IntoIterator<Item = T>>(&mut self, iter: I) { |
2758 | // This is the case for a general iter. |
2759 | // |
2760 | // This function should be the moral equivalent of: |
2761 | // |
2762 | // for item in iter { |
2763 | // self.push(item); |
2764 | // } |
2765 | |
2766 | let mut iter = iter.into_iter(); |
2767 | while let Some(element) = iter.next() { |
2768 | let len = self.len(); |
2769 | if len == self.capacity() { |
2770 | let (lower, _) = iter.size_hint(); |
2771 | self.reserve(lower.saturating_add(1)); |
2772 | } |
2773 | unsafe { |
2774 | ptr::write(self.as_mut_ptr().add(len), element); |
2775 | // Since next() executes user code which can panic we have to bump the length |
2776 | // after each step. |
2777 | // NB can't overflow since we would have had to alloc the address space |
2778 | self.set_len(len + 1); |
2779 | } |
2780 | } |
2781 | } |
2782 | } |
2783 | |
2784 | impl<T, A: Allocator> Vec<T, A> { |
2785 | /// Creates a splicing iterator that replaces the specified range in the vector |
2786 | /// with the given `replace_with` iterator and yields the removed items. |
2787 | /// `replace_with` does not need to be the same length as `range`. |
2788 | /// |
2789 | /// `range` is removed even if the iterator is not consumed until the end. |
2790 | /// |
2791 | /// It is unspecified how many elements are removed from the vector |
2792 | /// if the `Splice` value is leaked. |
2793 | /// |
2794 | /// The input iterator `replace_with` is only consumed when the `Splice` value is dropped. |
2795 | /// |
2796 | /// This is optimal if: |
2797 | /// |
2798 | /// * The tail (elements in the vector after `range`) is empty, |
2799 | /// * or `replace_with` yields fewer or equal elements than `range`’s length |
2800 | /// * or the lower bound of its `size_hint()` is exact. |
2801 | /// |
2802 | /// Otherwise, a temporary vector is allocated and the tail is moved twice. |
2803 | /// |
2804 | /// # Panics |
2805 | /// |
2806 | /// Panics if the starting point is greater than the end point or if |
2807 | /// the end point is greater than the length of the vector. |
2808 | /// |
2809 | /// # Examples |
2810 | /// |
2811 | /// ``` |
2812 | /// let mut v = vec![1, 2, 3, 4]; |
2813 | /// let new = [7, 8, 9]; |
2814 | /// let u: Vec<_> = v.splice(1..3, new).collect(); |
2815 | /// assert_eq!(v, &[1, 7, 8, 9, 4]); |
2816 | /// assert_eq!(u, &[2, 3]); |
2817 | /// ``` |
2818 | #[cfg (not(no_global_oom_handling))] |
2819 | #[inline (always)] |
2820 | pub fn splice<R, I>(&mut self, range: R, replace_with: I) -> Splice<'_, I::IntoIter, A> |
2821 | where |
2822 | R: RangeBounds<usize>, |
2823 | I: IntoIterator<Item = T>, |
2824 | { |
2825 | Splice { |
2826 | drain: self.drain(range), |
2827 | replace_with: replace_with.into_iter(), |
2828 | } |
2829 | } |
2830 | } |
2831 | |
2832 | /// Extend implementation that copies elements out of references before pushing them onto the Vec. |
2833 | /// |
2834 | /// This implementation is specialized for slice iterators, where it uses [`copy_from_slice`] to |
2835 | /// append the entire slice at once. |
2836 | /// |
2837 | /// [`copy_from_slice`]: slice::copy_from_slice |
2838 | #[cfg (not(no_global_oom_handling))] |
2839 | impl<'a, T: Copy + 'a, A: Allocator + 'a> Extend<&'a T> for Vec<T, A> { |
2840 | #[inline (always)] |
2841 | fn extend<I: IntoIterator<Item = &'a T>>(&mut self, iter: I) { |
2842 | let mut iter: ::IntoIter = iter.into_iter(); |
2843 | while let Some(element: &T) = iter.next() { |
2844 | let len: usize = self.len(); |
2845 | if len == self.capacity() { |
2846 | let (lower: usize, _) = iter.size_hint(); |
2847 | self.reserve(additional:lower.saturating_add(1)); |
2848 | } |
2849 | unsafe { |
2850 | ptr::write(self.as_mut_ptr().add(len), *element); |
2851 | // Since next() executes user code which can panic we have to bump the length |
2852 | // after each step. |
2853 | // NB can't overflow since we would have had to alloc the address space |
2854 | self.set_len(new_len:len + 1); |
2855 | } |
2856 | } |
2857 | } |
2858 | } |
2859 | |
2860 | /// Implements comparison of vectors, [lexicographically](core::cmp::Ord#lexicographical-comparison). |
2861 | impl<T: PartialOrd, A: Allocator> PartialOrd for Vec<T, A> { |
2862 | #[inline (always)] |
2863 | fn partial_cmp(&self, other: &Self) -> Option<Ordering> { |
2864 | PartialOrd::partial_cmp(&**self, &**other) |
2865 | } |
2866 | } |
2867 | |
2868 | impl<T: Eq, A: Allocator> Eq for Vec<T, A> {} |
2869 | |
2870 | /// Implements ordering of vectors, [lexicographically](core::cmp::Ord#lexicographical-comparison). |
2871 | impl<T: Ord, A: Allocator> Ord for Vec<T, A> { |
2872 | #[inline (always)] |
2873 | fn cmp(&self, other: &Self) -> Ordering { |
2874 | Ord::cmp(&**self, &**other) |
2875 | } |
2876 | } |
2877 | |
2878 | impl<T, A: Allocator> Drop for Vec<T, A> { |
2879 | #[inline (always)] |
2880 | fn drop(&mut self) { |
2881 | unsafe { |
2882 | // use drop for [T] |
2883 | // use a raw slice to refer to the elements of the vector as weakest necessary type; |
2884 | // could avoid questions of validity in certain cases |
2885 | ptr::drop_in_place(to_drop:ptr::slice_from_raw_parts_mut(self.as_mut_ptr(), self.len)) |
2886 | } |
2887 | // RawVec handles deallocation |
2888 | } |
2889 | } |
2890 | |
2891 | impl<T> Default for Vec<T> { |
2892 | /// Creates an empty `Vec<T>`. |
2893 | /// |
2894 | /// The vector will not allocate until elements are pushed onto it. |
2895 | #[inline (always)] |
2896 | fn default() -> Vec<T> { |
2897 | Vec::new() |
2898 | } |
2899 | } |
2900 | |
2901 | impl<T: fmt::Debug, A: Allocator> fmt::Debug for Vec<T, A> { |
2902 | #[inline (always)] |
2903 | fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result { |
2904 | fmt::Debug::fmt(&**self, f) |
2905 | } |
2906 | } |
2907 | |
2908 | impl<T, A: Allocator> AsRef<Vec<T, A>> for Vec<T, A> { |
2909 | #[inline (always)] |
2910 | fn as_ref(&self) -> &Vec<T, A> { |
2911 | self |
2912 | } |
2913 | } |
2914 | |
2915 | impl<T, A: Allocator> AsMut<Vec<T, A>> for Vec<T, A> { |
2916 | #[inline (always)] |
2917 | fn as_mut(&mut self) -> &mut Vec<T, A> { |
2918 | self |
2919 | } |
2920 | } |
2921 | |
2922 | impl<T, A: Allocator> AsRef<[T]> for Vec<T, A> { |
2923 | #[inline (always)] |
2924 | fn as_ref(&self) -> &[T] { |
2925 | self |
2926 | } |
2927 | } |
2928 | |
2929 | impl<T, A: Allocator> AsMut<[T]> for Vec<T, A> { |
2930 | #[inline (always)] |
2931 | fn as_mut(&mut self) -> &mut [T] { |
2932 | self |
2933 | } |
2934 | } |
2935 | |
2936 | #[cfg (not(no_global_oom_handling))] |
2937 | impl<T: Clone> From<&[T]> for Vec<T> { |
2938 | /// Allocate a `Vec<T>` and fill it by cloning `s`'s items. |
2939 | /// |
2940 | /// # Examples |
2941 | /// |
2942 | /// ``` |
2943 | /// assert_eq!(Vec::from(&[1, 2, 3][..]), vec![1, 2, 3]); |
2944 | /// ``` |
2945 | #[inline (always)] |
2946 | fn from(s: &[T]) -> Vec<T> { |
2947 | let mut vec: Vec = Vec::with_capacity(s.len()); |
2948 | vec.extend_from_slice(s); |
2949 | vec |
2950 | } |
2951 | } |
2952 | |
2953 | #[cfg (not(no_global_oom_handling))] |
2954 | impl<T: Clone> From<&mut [T]> for Vec<T> { |
2955 | /// Allocate a `Vec<T>` and fill it by cloning `s`'s items. |
2956 | /// |
2957 | /// # Examples |
2958 | /// |
2959 | /// ``` |
2960 | /// assert_eq!(Vec::from(&mut [1, 2, 3][..]), vec![1, 2, 3]); |
2961 | /// ``` |
2962 | #[inline (always)] |
2963 | fn from(s: &mut [T]) -> Vec<T> { |
2964 | let mut vec: Vec = Vec::with_capacity(s.len()); |
2965 | vec.extend_from_slice(s); |
2966 | vec |
2967 | } |
2968 | } |
2969 | |
2970 | #[cfg (not(no_global_oom_handling))] |
2971 | impl<T, const N: usize> From<[T; N]> for Vec<T> { |
2972 | #[inline (always)] |
2973 | fn from(s: [T; N]) -> Vec<T> { |
2974 | Box::slice(Box::new(s)).into_vec() |
2975 | } |
2976 | } |
2977 | |
2978 | impl<T, A: Allocator> From<Box<[T], A>> for Vec<T, A> { |
2979 | /// Convert a boxed slice into a vector by transferring ownership of |
2980 | /// the existing heap allocation. |
2981 | /// |
2982 | /// # Examples |
2983 | /// |
2984 | /// ``` |
2985 | /// let b: Box<[i32]> = vec![1, 2, 3].into_boxed_slice(); |
2986 | /// assert_eq!(Vec::from(b), vec![1, 2, 3]); |
2987 | /// ``` |
2988 | #[inline (always)] |
2989 | fn from(s: Box<[T], A>) -> Self { |
2990 | s.into_vec() |
2991 | } |
2992 | } |
2993 | |
2994 | impl<T, A: Allocator, const N: usize> From<Box<[T; N], A>> for Vec<T, A> { |
2995 | /// Convert a boxed array into a vector by transferring ownership of |
2996 | /// the existing heap allocation. |
2997 | /// |
2998 | /// # Examples |
2999 | /// |
3000 | /// ``` |
3001 | /// let b: Box<[i32; 3]> = Box::new([1, 2, 3]); |
3002 | /// assert_eq!(Vec::from(b), vec![1, 2, 3]); |
3003 | /// ``` |
3004 | #[inline (always)] |
3005 | fn from(s: Box<[T; N], A>) -> Self { |
3006 | s.into_vec() |
3007 | } |
3008 | } |
3009 | |
3010 | // note: test pulls in libstd, which causes errors here |
3011 | #[cfg (not(no_global_oom_handling))] |
3012 | impl<T, A: Allocator> From<Vec<T, A>> for Box<[T], A> { |
3013 | /// Convert a vector into a boxed slice. |
3014 | /// |
3015 | /// If `v` has excess capacity, its items will be moved into a |
3016 | /// newly-allocated buffer with exactly the right capacity. |
3017 | /// |
3018 | /// # Examples |
3019 | /// |
3020 | /// ``` |
3021 | /// assert_eq!(Box::from(vec![1, 2, 3]), vec![1, 2, 3].into_boxed_slice()); |
3022 | /// ``` |
3023 | /// |
3024 | /// Any excess capacity is removed: |
3025 | /// ``` |
3026 | /// let mut vec = Vec::with_capacity(10); |
3027 | /// vec.extend([1, 2, 3]); |
3028 | /// |
3029 | /// assert_eq!(Box::from(vec), vec![1, 2, 3].into_boxed_slice()); |
3030 | /// ``` |
3031 | #[inline (always)] |
3032 | fn from(v: Vec<T, A>) -> Self { |
3033 | v.into_boxed_slice() |
3034 | } |
3035 | } |
3036 | |
3037 | #[cfg (not(no_global_oom_handling))] |
3038 | impl From<&str> for Vec<u8> { |
3039 | /// Allocate a `Vec<u8>` and fill it with a UTF-8 string. |
3040 | /// |
3041 | /// # Examples |
3042 | /// |
3043 | /// ``` |
3044 | /// assert_eq!(Vec::from("123" ), vec![b'1' , b'2' , b'3' ]); |
3045 | /// ``` |
3046 | #[inline (always)] |
3047 | fn from(s: &str) -> Vec<u8> { |
3048 | From::from(s.as_bytes()) |
3049 | } |
3050 | } |
3051 | |
3052 | impl<T, A: Allocator, const N: usize> TryFrom<Vec<T, A>> for [T; N] { |
3053 | type Error = Vec<T, A>; |
3054 | |
3055 | /// Gets the entire contents of the `Vec<T>` as an array, |
3056 | /// if its size exactly matches that of the requested array. |
3057 | /// |
3058 | /// # Examples |
3059 | /// |
3060 | /// ``` |
3061 | /// assert_eq!(vec![1, 2, 3].try_into(), Ok([1, 2, 3])); |
3062 | /// assert_eq!(<Vec<i32>>::new().try_into(), Ok([])); |
3063 | /// ``` |
3064 | /// |
3065 | /// If the length doesn't match, the input comes back in `Err`: |
3066 | /// ``` |
3067 | /// let r: Result<[i32; 4], _> = (0..10).collect::<Vec<_>>().try_into(); |
3068 | /// assert_eq!(r, Err(vec![0, 1, 2, 3, 4, 5, 6, 7, 8, 9])); |
3069 | /// ``` |
3070 | /// |
3071 | /// If you're fine with just getting a prefix of the `Vec<T>`, |
3072 | /// you can call [`.truncate(N)`](Vec::truncate) first. |
3073 | /// ``` |
3074 | /// let mut v = String::from("hello world" ).into_bytes(); |
3075 | /// v.sort(); |
3076 | /// v.truncate(2); |
3077 | /// let [a, b]: [_; 2] = v.try_into().unwrap(); |
3078 | /// assert_eq!(a, b' ' ); |
3079 | /// assert_eq!(b, b'd' ); |
3080 | /// ``` |
3081 | #[inline (always)] |
3082 | fn try_from(mut vec: Vec<T, A>) -> Result<[T; N], Vec<T, A>> { |
3083 | if vec.len() != N { |
3084 | return Err(vec); |
3085 | } |
3086 | |
3087 | // SAFETY: `.set_len(0)` is always sound. |
3088 | unsafe { vec.set_len(0) }; |
3089 | |
3090 | // SAFETY: A `Vec`'s pointer is always aligned properly, and |
3091 | // the alignment the array needs is the same as the items. |
3092 | // We checked earlier that we have sufficient items. |
3093 | // The items will not double-drop as the `set_len` |
3094 | // tells the `Vec` not to also drop them. |
3095 | let array = unsafe { ptr::read(vec.as_ptr() as *const [T; N]) }; |
3096 | Ok(array) |
3097 | } |
3098 | } |
3099 | |
3100 | #[inline (always)] |
3101 | #[cfg (not(no_global_oom_handling))] |
3102 | #[doc (hidden)] |
3103 | pub fn from_elem_in<T: Clone, A: Allocator>(elem: T, n: usize, alloc: A) -> Vec<T, A> { |
3104 | let mut v: Vec = Vec::with_capacity_in(capacity:n, alloc); |
3105 | v.extend_with(n, value:ExtendElement(elem)); |
3106 | v |
3107 | } |
3108 | |
3109 | #[inline (always)] |
3110 | #[cfg (not(no_global_oom_handling))] |
3111 | #[doc (hidden)] |
3112 | pub fn from_elem<T: Clone>(elem: T, n: usize) -> Vec<T> { |
3113 | let mut v: Vec = Vec::with_capacity(n); |
3114 | v.extend_with(n, value:ExtendElement(elem)); |
3115 | v |
3116 | } |
3117 | |
3118 | #[cfg (feature = "serde" )] |
3119 | impl<T, A> serde::Serialize for Vec<T, A> |
3120 | where |
3121 | T: serde::Serialize, |
3122 | A: Allocator, |
3123 | { |
3124 | #[inline (always)] |
3125 | fn serialize<S>(&self, serializer: S) -> Result<S::Ok, S::Error> |
3126 | where |
3127 | S: serde::ser::Serializer, |
3128 | { |
3129 | serializer.collect_seq(self) |
3130 | } |
3131 | } |
3132 | |
3133 | #[cfg (feature = "serde" )] |
3134 | impl<'de, T, A> serde::de::Deserialize<'de> for Vec<T, A> |
3135 | where |
3136 | T: serde::de::Deserialize<'de>, |
3137 | A: Allocator + Default, |
3138 | { |
3139 | #[inline (always)] |
3140 | fn deserialize<D>(deserializer: D) -> Result<Self, D::Error> |
3141 | where |
3142 | D: serde::de::Deserializer<'de>, |
3143 | { |
3144 | struct VecVisitor<T, A> { |
3145 | marker: PhantomData<(T, A)>, |
3146 | } |
3147 | |
3148 | impl<'de, T, A> serde::de::Visitor<'de> for VecVisitor<T, A> |
3149 | where |
3150 | T: serde::de::Deserialize<'de>, |
3151 | A: Allocator + Default, |
3152 | { |
3153 | type Value = Vec<T, A>; |
3154 | |
3155 | fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result { |
3156 | formatter.write_str("a sequence" ) |
3157 | } |
3158 | |
3159 | fn visit_seq<S>(self, mut seq: S) -> Result<Self::Value, S::Error> |
3160 | where |
3161 | S: serde::de::SeqAccess<'de>, |
3162 | { |
3163 | let mut values = Vec::with_capacity_in(cautious(seq.size_hint()), A::default()); |
3164 | |
3165 | while let Some(value) = seq.next_element()? { |
3166 | values.push(value); |
3167 | } |
3168 | |
3169 | Ok(values) |
3170 | } |
3171 | } |
3172 | |
3173 | let visitor = VecVisitor { |
3174 | marker: PhantomData, |
3175 | }; |
3176 | deserializer.deserialize_seq(visitor) |
3177 | } |
3178 | |
3179 | #[inline (always)] |
3180 | fn deserialize_in_place<D>(deserializer: D, place: &mut Self) -> Result<(), D::Error> |
3181 | where |
3182 | D: serde::de::Deserializer<'de>, |
3183 | { |
3184 | struct VecInPlaceVisitor<'a, T: 'a, A: Allocator + 'a>(&'a mut Vec<T, A>); |
3185 | |
3186 | impl<'a, 'de, T, A> serde::de::Visitor<'de> for VecInPlaceVisitor<'a, T, A> |
3187 | where |
3188 | T: serde::de::Deserialize<'de>, |
3189 | A: Allocator + Default, |
3190 | { |
3191 | type Value = (); |
3192 | |
3193 | fn expecting(&self, formatter: &mut fmt::Formatter) -> fmt::Result { |
3194 | formatter.write_str("a sequence" ) |
3195 | } |
3196 | |
3197 | fn visit_seq<S>(self, mut seq: S) -> Result<Self::Value, S::Error> |
3198 | where |
3199 | S: serde::de::SeqAccess<'de>, |
3200 | { |
3201 | let hint = cautious(seq.size_hint()); |
3202 | if let Some(additional) = hint.checked_sub(self.0.len()) { |
3203 | self.0.reserve(additional); |
3204 | } |
3205 | |
3206 | for i in 0..self.0.len() { |
3207 | let next = { |
3208 | let next_place = InPlaceSeed(&mut self.0[i]); |
3209 | seq.next_element_seed(next_place)? |
3210 | }; |
3211 | if next.is_none() { |
3212 | self.0.truncate(i); |
3213 | return Ok(()); |
3214 | } |
3215 | } |
3216 | |
3217 | while let Some(value) = seq.next_element()? { |
3218 | self.0.push(value); |
3219 | } |
3220 | |
3221 | Ok(()) |
3222 | } |
3223 | } |
3224 | |
3225 | deserializer.deserialize_seq(VecInPlaceVisitor(place)) |
3226 | } |
3227 | } |
3228 | |
3229 | #[cfg (feature = "serde" )] |
3230 | pub fn cautious(hint: Option<usize>) -> usize { |
3231 | cmp::min(hint.unwrap_or(0), 4096) |
3232 | } |
3233 | |
3234 | /// A DeserializeSeed helper for implementing deserialize_in_place Visitors. |
3235 | /// |
3236 | /// Wraps a mutable reference and calls deserialize_in_place on it. |
3237 | |
3238 | #[cfg (feature = "serde" )] |
3239 | pub struct InPlaceSeed<'a, T: 'a>(pub &'a mut T); |
3240 | |
3241 | #[cfg (feature = "serde" )] |
3242 | impl<'a, 'de, T> serde::de::DeserializeSeed<'de> for InPlaceSeed<'a, T> |
3243 | where |
3244 | T: serde::de::Deserialize<'de>, |
3245 | { |
3246 | type Value = (); |
3247 | fn deserialize<D>(self, deserializer: D) -> Result<Self::Value, D::Error> |
3248 | where |
3249 | D: serde::de::Deserializer<'de>, |
3250 | { |
3251 | T::deserialize_in_place(deserializer, self.0) |
3252 | } |
3253 | } |
3254 | |