core/slice/mod.rs
1//! Slice management and manipulation.
2//!
3//! For more details see [`std::slice`].
4//!
5//! [`std::slice`]: ../../std/slice/index.html
6
7#![stable(feature = "rust1", since = "1.0.0")]
8
9use crate::clone::TrivialClone;
10use crate::cmp::Ordering::{self, Equal, Greater, Less};
11use crate::intrinsics::{exact_div, unchecked_sub};
12use crate::marker::Destruct;
13use crate::mem::{self, MaybeUninit, SizedTypeProperties};
14use crate::num::NonZero;
15use crate::ops::{OneSidedRange, OneSidedRangeBound, Range, RangeBounds, RangeInclusive};
16use crate::panic::const_panic;
17use crate::simd::{self, Simd};
18use crate::ub_checks::assert_unsafe_precondition;
19use crate::{fmt, hint, ptr, range, slice};
20
21#[unstable(
22 feature = "slice_internals",
23 issue = "none",
24 reason = "exposed from core to be reused in std; use the memchr crate"
25)]
26#[doc(hidden)]
27/// Pure Rust memchr implementation, taken from rust-memchr
28pub mod memchr;
29
30#[unstable(
31 feature = "slice_internals",
32 issue = "none",
33 reason = "exposed from core to be reused in std;"
34)]
35#[doc(hidden)]
36pub mod sort;
37
38mod ascii;
39mod cmp;
40pub(crate) mod index;
41mod iter;
42mod raw;
43mod rotate;
44mod specialize;
45
46#[stable(feature = "inherent_ascii_escape", since = "1.60.0")]
47pub use ascii::EscapeAscii;
48#[unstable(feature = "str_internals", issue = "none")]
49#[doc(hidden)]
50pub use ascii::is_ascii_simple;
51#[stable(feature = "slice_get_slice", since = "1.28.0")]
52pub use index::SliceIndex;
53#[unstable(feature = "slice_range", issue = "76393")]
54pub use index::{range, try_range};
55#[stable(feature = "array_windows", since = "1.94.0")]
56pub use iter::ArrayWindows;
57#[stable(feature = "slice_group_by", since = "1.77.0")]
58pub use iter::{ChunkBy, ChunkByMut};
59#[stable(feature = "rust1", since = "1.0.0")]
60pub use iter::{Chunks, ChunksMut, Windows};
61#[stable(feature = "chunks_exact", since = "1.31.0")]
62pub use iter::{ChunksExact, ChunksExactMut};
63#[stable(feature = "rust1", since = "1.0.0")]
64pub use iter::{Iter, IterMut};
65#[stable(feature = "rchunks", since = "1.31.0")]
66pub use iter::{RChunks, RChunksExact, RChunksExactMut, RChunksMut};
67#[stable(feature = "slice_rsplit", since = "1.27.0")]
68pub use iter::{RSplit, RSplitMut};
69#[stable(feature = "rust1", since = "1.0.0")]
70pub use iter::{RSplitN, RSplitNMut, Split, SplitMut, SplitN, SplitNMut};
71#[stable(feature = "split_inclusive", since = "1.51.0")]
72pub use iter::{SplitInclusive, SplitInclusiveMut};
73#[stable(feature = "from_ref", since = "1.28.0")]
74pub use raw::{from_mut, from_ref};
75#[unstable(feature = "slice_from_ptr_range", issue = "89792")]
76pub use raw::{from_mut_ptr_range, from_ptr_range};
77#[stable(feature = "rust1", since = "1.0.0")]
78pub use raw::{from_raw_parts, from_raw_parts_mut};
79
80/// Calculates the direction and split point of a one-sided range.
81///
82/// This is a helper function for `split_off` and `split_off_mut` that returns
83/// the direction of the split (front or back) as well as the index at
84/// which to split. Returns `None` if the split index would overflow.
85#[inline]
86fn split_point_of(range: impl OneSidedRange<usize>) -> Option<(Direction, usize)> {
87 use OneSidedRangeBound::{End, EndInclusive, StartInclusive};
88
89 Some(match range.bound() {
90 (StartInclusive, i) => (Direction::Back, i),
91 (End, i) => (Direction::Front, i),
92 (EndInclusive, i) => (Direction::Front, i.checked_add(1)?),
93 })
94}
95
96enum Direction {
97 Front,
98 Back,
99}
100
101impl<T> [T] {
102 /// Returns the number of elements in the slice.
103 ///
104 /// # Examples
105 ///
106 /// ```
107 /// let a = [1, 2, 3];
108 /// assert_eq!(a.len(), 3);
109 /// ```
110 #[lang = "slice_len_fn"]
111 #[stable(feature = "rust1", since = "1.0.0")]
112 #[rustc_const_stable(feature = "const_slice_len", since = "1.39.0")]
113 #[rustc_no_implicit_autorefs]
114 #[inline]
115 #[must_use]
116 pub const fn len(&self) -> usize {
117 ptr::metadata(self)
118 }
119
120 /// Returns `true` if the slice has a length of 0.
121 ///
122 /// # Examples
123 ///
124 /// ```
125 /// let a = [1, 2, 3];
126 /// assert!(!a.is_empty());
127 ///
128 /// let b: &[i32] = &[];
129 /// assert!(b.is_empty());
130 /// ```
131 #[stable(feature = "rust1", since = "1.0.0")]
132 #[rustc_const_stable(feature = "const_slice_is_empty", since = "1.39.0")]
133 #[rustc_no_implicit_autorefs]
134 #[inline]
135 #[must_use]
136 pub const fn is_empty(&self) -> bool {
137 self.len() == 0
138 }
139
140 /// Returns the first element of the slice, or `None` if it is empty.
141 ///
142 /// # Examples
143 ///
144 /// ```
145 /// let v = [10, 40, 30];
146 /// assert_eq!(Some(&10), v.first());
147 ///
148 /// let w: &[i32] = &[];
149 /// assert_eq!(None, w.first());
150 /// ```
151 #[stable(feature = "rust1", since = "1.0.0")]
152 #[rustc_const_stable(feature = "const_slice_first_last_not_mut", since = "1.56.0")]
153 #[inline]
154 #[must_use]
155 pub const fn first(&self) -> Option<&T> {
156 if let [first, ..] = self { Some(first) } else { None }
157 }
158
159 /// Returns a mutable reference to the first element of the slice, or `None` if it is empty.
160 ///
161 /// # Examples
162 ///
163 /// ```
164 /// let x = &mut [0, 1, 2];
165 ///
166 /// if let Some(first) = x.first_mut() {
167 /// *first = 5;
168 /// }
169 /// assert_eq!(x, &[5, 1, 2]);
170 ///
171 /// let y: &mut [i32] = &mut [];
172 /// assert_eq!(None, y.first_mut());
173 /// ```
174 #[stable(feature = "rust1", since = "1.0.0")]
175 #[rustc_const_stable(feature = "const_slice_first_last", since = "1.83.0")]
176 #[inline]
177 #[must_use]
178 pub const fn first_mut(&mut self) -> Option<&mut T> {
179 if let [first, ..] = self { Some(first) } else { None }
180 }
181
182 /// Returns the first and all the rest of the elements of the slice, or `None` if it is empty.
183 ///
184 /// # Examples
185 ///
186 /// ```
187 /// let x = &[0, 1, 2];
188 ///
189 /// if let Some((first, elements)) = x.split_first() {
190 /// assert_eq!(first, &0);
191 /// assert_eq!(elements, &[1, 2]);
192 /// }
193 /// ```
194 #[stable(feature = "slice_splits", since = "1.5.0")]
195 #[rustc_const_stable(feature = "const_slice_first_last_not_mut", since = "1.56.0")]
196 #[inline]
197 #[must_use]
198 pub const fn split_first(&self) -> Option<(&T, &[T])> {
199 if let [first, tail @ ..] = self { Some((first, tail)) } else { None }
200 }
201
202 /// Returns the first and all the rest of the elements of the slice, or `None` if it is empty.
203 ///
204 /// # Examples
205 ///
206 /// ```
207 /// let x = &mut [0, 1, 2];
208 ///
209 /// if let Some((first, elements)) = x.split_first_mut() {
210 /// *first = 3;
211 /// elements[0] = 4;
212 /// elements[1] = 5;
213 /// }
214 /// assert_eq!(x, &[3, 4, 5]);
215 /// ```
216 #[stable(feature = "slice_splits", since = "1.5.0")]
217 #[rustc_const_stable(feature = "const_slice_first_last", since = "1.83.0")]
218 #[inline]
219 #[must_use]
220 pub const fn split_first_mut(&mut self) -> Option<(&mut T, &mut [T])> {
221 if let [first, tail @ ..] = self { Some((first, tail)) } else { None }
222 }
223
224 /// Returns the last and all the rest of the elements of the slice, or `None` if it is empty.
225 ///
226 /// # Examples
227 ///
228 /// ```
229 /// let x = &[0, 1, 2];
230 ///
231 /// if let Some((last, elements)) = x.split_last() {
232 /// assert_eq!(last, &2);
233 /// assert_eq!(elements, &[0, 1]);
234 /// }
235 /// ```
236 #[stable(feature = "slice_splits", since = "1.5.0")]
237 #[rustc_const_stable(feature = "const_slice_first_last_not_mut", since = "1.56.0")]
238 #[inline]
239 #[must_use]
240 pub const fn split_last(&self) -> Option<(&T, &[T])> {
241 if let [init @ .., last] = self { Some((last, init)) } else { None }
242 }
243
244 /// Returns the last and all the rest of the elements of the slice, or `None` if it is empty.
245 ///
246 /// # Examples
247 ///
248 /// ```
249 /// let x = &mut [0, 1, 2];
250 ///
251 /// if let Some((last, elements)) = x.split_last_mut() {
252 /// *last = 3;
253 /// elements[0] = 4;
254 /// elements[1] = 5;
255 /// }
256 /// assert_eq!(x, &[4, 5, 3]);
257 /// ```
258 #[stable(feature = "slice_splits", since = "1.5.0")]
259 #[rustc_const_stable(feature = "const_slice_first_last", since = "1.83.0")]
260 #[inline]
261 #[must_use]
262 pub const fn split_last_mut(&mut self) -> Option<(&mut T, &mut [T])> {
263 if let [init @ .., last] = self { Some((last, init)) } else { None }
264 }
265
266 /// Returns the last element of the slice, or `None` if it is empty.
267 ///
268 /// # Examples
269 ///
270 /// ```
271 /// let v = [10, 40, 30];
272 /// assert_eq!(Some(&30), v.last());
273 ///
274 /// let w: &[i32] = &[];
275 /// assert_eq!(None, w.last());
276 /// ```
277 #[stable(feature = "rust1", since = "1.0.0")]
278 #[rustc_const_stable(feature = "const_slice_first_last_not_mut", since = "1.56.0")]
279 #[inline]
280 #[must_use]
281 pub const fn last(&self) -> Option<&T> {
282 if let [.., last] = self { Some(last) } else { None }
283 }
284
285 /// Returns a mutable reference to the last item in the slice, or `None` if it is empty.
286 ///
287 /// # Examples
288 ///
289 /// ```
290 /// let x = &mut [0, 1, 2];
291 ///
292 /// if let Some(last) = x.last_mut() {
293 /// *last = 10;
294 /// }
295 /// assert_eq!(x, &[0, 1, 10]);
296 ///
297 /// let y: &mut [i32] = &mut [];
298 /// assert_eq!(None, y.last_mut());
299 /// ```
300 #[stable(feature = "rust1", since = "1.0.0")]
301 #[rustc_const_stable(feature = "const_slice_first_last", since = "1.83.0")]
302 #[inline]
303 #[must_use]
304 pub const fn last_mut(&mut self) -> Option<&mut T> {
305 if let [.., last] = self { Some(last) } else { None }
306 }
307
308 /// Returns an array reference to the first `N` items in the slice.
309 ///
310 /// If the slice is not at least `N` in length, this will return `None`.
311 ///
312 /// # Examples
313 ///
314 /// ```
315 /// let u = [10, 40, 30];
316 /// assert_eq!(Some(&[10, 40]), u.first_chunk::<2>());
317 ///
318 /// let v: &[i32] = &[10];
319 /// assert_eq!(None, v.first_chunk::<2>());
320 ///
321 /// let w: &[i32] = &[];
322 /// assert_eq!(Some(&[]), w.first_chunk::<0>());
323 /// ```
324 #[inline]
325 #[stable(feature = "slice_first_last_chunk", since = "1.77.0")]
326 #[rustc_const_stable(feature = "slice_first_last_chunk", since = "1.77.0")]
327 pub const fn first_chunk<const N: usize>(&self) -> Option<&[T; N]> {
328 if self.len() < N {
329 None
330 } else {
331 // SAFETY: We explicitly check for the correct number of elements,
332 // and do not let the reference outlive the slice.
333 Some(unsafe { &*(self.as_ptr().cast_array()) })
334 }
335 }
336
337 /// Returns a mutable array reference to the first `N` items in the slice.
338 ///
339 /// If the slice is not at least `N` in length, this will return `None`.
340 ///
341 /// # Examples
342 ///
343 /// ```
344 /// let x = &mut [0, 1, 2];
345 ///
346 /// if let Some(first) = x.first_chunk_mut::<2>() {
347 /// first[0] = 5;
348 /// first[1] = 4;
349 /// }
350 /// assert_eq!(x, &[5, 4, 2]);
351 ///
352 /// assert_eq!(None, x.first_chunk_mut::<4>());
353 /// ```
354 #[inline]
355 #[stable(feature = "slice_first_last_chunk", since = "1.77.0")]
356 #[rustc_const_stable(feature = "const_slice_first_last_chunk", since = "1.83.0")]
357 pub const fn first_chunk_mut<const N: usize>(&mut self) -> Option<&mut [T; N]> {
358 if self.len() < N {
359 None
360 } else {
361 // SAFETY: We explicitly check for the correct number of elements,
362 // do not let the reference outlive the slice,
363 // and require exclusive access to the entire slice to mutate the chunk.
364 Some(unsafe { &mut *(self.as_mut_ptr().cast_array()) })
365 }
366 }
367
368 /// Returns an array reference to the first `N` items in the slice and the remaining slice.
369 ///
370 /// If the slice is not at least `N` in length, this will return `None`.
371 ///
372 /// # Examples
373 ///
374 /// ```
375 /// let x = &[0, 1, 2];
376 ///
377 /// if let Some((first, elements)) = x.split_first_chunk::<2>() {
378 /// assert_eq!(first, &[0, 1]);
379 /// assert_eq!(elements, &[2]);
380 /// }
381 ///
382 /// assert_eq!(None, x.split_first_chunk::<4>());
383 /// ```
384 #[inline]
385 #[stable(feature = "slice_first_last_chunk", since = "1.77.0")]
386 #[rustc_const_stable(feature = "slice_first_last_chunk", since = "1.77.0")]
387 pub const fn split_first_chunk<const N: usize>(&self) -> Option<(&[T; N], &[T])> {
388 let Some((first, tail)) = self.split_at_checked(N) else { return None };
389
390 // SAFETY: We explicitly check for the correct number of elements,
391 // and do not let the references outlive the slice.
392 Some((unsafe { &*(first.as_ptr().cast_array()) }, tail))
393 }
394
395 /// Returns a mutable array reference to the first `N` items in the slice and the remaining
396 /// slice.
397 ///
398 /// If the slice is not at least `N` in length, this will return `None`.
399 ///
400 /// # Examples
401 ///
402 /// ```
403 /// let x = &mut [0, 1, 2];
404 ///
405 /// if let Some((first, elements)) = x.split_first_chunk_mut::<2>() {
406 /// first[0] = 3;
407 /// first[1] = 4;
408 /// elements[0] = 5;
409 /// }
410 /// assert_eq!(x, &[3, 4, 5]);
411 ///
412 /// assert_eq!(None, x.split_first_chunk_mut::<4>());
413 /// ```
414 #[inline]
415 #[stable(feature = "slice_first_last_chunk", since = "1.77.0")]
416 #[rustc_const_stable(feature = "const_slice_first_last_chunk", since = "1.83.0")]
417 pub const fn split_first_chunk_mut<const N: usize>(
418 &mut self,
419 ) -> Option<(&mut [T; N], &mut [T])> {
420 let Some((first, tail)) = self.split_at_mut_checked(N) else { return None };
421
422 // SAFETY: We explicitly check for the correct number of elements,
423 // do not let the reference outlive the slice,
424 // and enforce exclusive mutability of the chunk by the split.
425 Some((unsafe { &mut *(first.as_mut_ptr().cast_array()) }, tail))
426 }
427
428 /// Returns an array reference to the last `N` items in the slice and the remaining slice.
429 ///
430 /// If the slice is not at least `N` in length, this will return `None`.
431 ///
432 /// # Examples
433 ///
434 /// ```
435 /// let x = &[0, 1, 2];
436 ///
437 /// if let Some((elements, last)) = x.split_last_chunk::<2>() {
438 /// assert_eq!(elements, &[0]);
439 /// assert_eq!(last, &[1, 2]);
440 /// }
441 ///
442 /// assert_eq!(None, x.split_last_chunk::<4>());
443 /// ```
444 #[inline]
445 #[stable(feature = "slice_first_last_chunk", since = "1.77.0")]
446 #[rustc_const_stable(feature = "slice_first_last_chunk", since = "1.77.0")]
447 pub const fn split_last_chunk<const N: usize>(&self) -> Option<(&[T], &[T; N])> {
448 let Some(index) = self.len().checked_sub(N) else { return None };
449 let (init, last) = self.split_at(index);
450
451 // SAFETY: We explicitly check for the correct number of elements,
452 // and do not let the references outlive the slice.
453 Some((init, unsafe { &*(last.as_ptr().cast_array()) }))
454 }
455
456 /// Returns a mutable array reference to the last `N` items in the slice and the remaining
457 /// slice.
458 ///
459 /// If the slice is not at least `N` in length, this will return `None`.
460 ///
461 /// # Examples
462 ///
463 /// ```
464 /// let x = &mut [0, 1, 2];
465 ///
466 /// if let Some((elements, last)) = x.split_last_chunk_mut::<2>() {
467 /// last[0] = 3;
468 /// last[1] = 4;
469 /// elements[0] = 5;
470 /// }
471 /// assert_eq!(x, &[5, 3, 4]);
472 ///
473 /// assert_eq!(None, x.split_last_chunk_mut::<4>());
474 /// ```
475 #[inline]
476 #[stable(feature = "slice_first_last_chunk", since = "1.77.0")]
477 #[rustc_const_stable(feature = "const_slice_first_last_chunk", since = "1.83.0")]
478 pub const fn split_last_chunk_mut<const N: usize>(
479 &mut self,
480 ) -> Option<(&mut [T], &mut [T; N])> {
481 let Some(index) = self.len().checked_sub(N) else { return None };
482 let (init, last) = self.split_at_mut(index);
483
484 // SAFETY: We explicitly check for the correct number of elements,
485 // do not let the reference outlive the slice,
486 // and enforce exclusive mutability of the chunk by the split.
487 Some((init, unsafe { &mut *(last.as_mut_ptr().cast_array()) }))
488 }
489
490 /// Returns an array reference to the last `N` items in the slice.
491 ///
492 /// If the slice is not at least `N` in length, this will return `None`.
493 ///
494 /// # Examples
495 ///
496 /// ```
497 /// let u = [10, 40, 30];
498 /// assert_eq!(Some(&[40, 30]), u.last_chunk::<2>());
499 ///
500 /// let v: &[i32] = &[10];
501 /// assert_eq!(None, v.last_chunk::<2>());
502 ///
503 /// let w: &[i32] = &[];
504 /// assert_eq!(Some(&[]), w.last_chunk::<0>());
505 /// ```
506 #[inline]
507 #[stable(feature = "slice_first_last_chunk", since = "1.77.0")]
508 #[rustc_const_stable(feature = "const_slice_last_chunk", since = "1.80.0")]
509 pub const fn last_chunk<const N: usize>(&self) -> Option<&[T; N]> {
510 // FIXME(const-hack): Without const traits, we need this instead of `get`.
511 let Some(index) = self.len().checked_sub(N) else { return None };
512 let (_, last) = self.split_at(index);
513
514 // SAFETY: We explicitly check for the correct number of elements,
515 // and do not let the references outlive the slice.
516 Some(unsafe { &*(last.as_ptr().cast_array()) })
517 }
518
519 /// Returns a mutable array reference to the last `N` items in the slice.
520 ///
521 /// If the slice is not at least `N` in length, this will return `None`.
522 ///
523 /// # Examples
524 ///
525 /// ```
526 /// let x = &mut [0, 1, 2];
527 ///
528 /// if let Some(last) = x.last_chunk_mut::<2>() {
529 /// last[0] = 10;
530 /// last[1] = 20;
531 /// }
532 /// assert_eq!(x, &[0, 10, 20]);
533 ///
534 /// assert_eq!(None, x.last_chunk_mut::<4>());
535 /// ```
536 #[inline]
537 #[stable(feature = "slice_first_last_chunk", since = "1.77.0")]
538 #[rustc_const_stable(feature = "const_slice_first_last_chunk", since = "1.83.0")]
539 pub const fn last_chunk_mut<const N: usize>(&mut self) -> Option<&mut [T; N]> {
540 // FIXME(const-hack): Without const traits, we need this instead of `get`.
541 let Some(index) = self.len().checked_sub(N) else { return None };
542 let (_, last) = self.split_at_mut(index);
543
544 // SAFETY: We explicitly check for the correct number of elements,
545 // do not let the reference outlive the slice,
546 // and require exclusive access to the entire slice to mutate the chunk.
547 Some(unsafe { &mut *(last.as_mut_ptr().cast_array()) })
548 }
549
550 /// Returns a reference to an element or subslice depending on the type of
551 /// index.
552 ///
553 /// - If given a position, returns a reference to the element at that
554 /// position or `None` if out of bounds.
555 /// - If given a range, returns the subslice corresponding to that range,
556 /// or `None` if out of bounds.
557 ///
558 /// # Examples
559 ///
560 /// ```
561 /// let v = [10, 40, 30];
562 /// assert_eq!(Some(&40), v.get(1));
563 /// assert_eq!(Some(&[10, 40][..]), v.get(0..2));
564 /// assert_eq!(None, v.get(3));
565 /// assert_eq!(None, v.get(0..4));
566 /// ```
567 #[stable(feature = "rust1", since = "1.0.0")]
568 #[rustc_no_implicit_autorefs]
569 #[inline]
570 #[must_use]
571 #[rustc_const_unstable(feature = "const_index", issue = "143775")]
572 pub const fn get<I>(&self, index: I) -> Option<&I::Output>
573 where
574 I: [const] SliceIndex<Self>,
575 {
576 index.get(self)
577 }
578
579 /// Returns a mutable reference to an element or subslice depending on the
580 /// type of index (see [`get`]) or `None` if the index is out of bounds.
581 ///
582 /// [`get`]: slice::get
583 ///
584 /// # Examples
585 ///
586 /// ```
587 /// let x = &mut [0, 1, 2];
588 ///
589 /// if let Some(elem) = x.get_mut(1) {
590 /// *elem = 42;
591 /// }
592 /// assert_eq!(x, &[0, 42, 2]);
593 /// ```
594 #[stable(feature = "rust1", since = "1.0.0")]
595 #[rustc_no_implicit_autorefs]
596 #[inline]
597 #[must_use]
598 #[rustc_const_unstable(feature = "const_index", issue = "143775")]
599 #[rustc_no_writable]
600 pub const fn get_mut<I>(&mut self, index: I) -> Option<&mut I::Output>
601 where
602 I: [const] SliceIndex<Self>,
603 {
604 index.get_mut(self)
605 }
606
607 /// Returns a reference to an element or subslice, without doing bounds
608 /// checking.
609 ///
610 /// For a safe alternative see [`get`].
611 ///
612 /// # Safety
613 ///
614 /// Calling this method with an out-of-bounds index is *[undefined behavior]*
615 /// even if the resulting reference is not used.
616 ///
617 /// You can think of this like `.get(index).unwrap_unchecked()`. It's UB
618 /// to call `.get_unchecked(len)`, even if you immediately convert to a
619 /// pointer. And it's UB to call `.get_unchecked(..len + 1)`,
620 /// `.get_unchecked(..=len)`, or similar.
621 ///
622 /// [`get`]: slice::get
623 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
624 ///
625 /// # Examples
626 ///
627 /// ```
628 /// let x = &[1, 2, 4];
629 ///
630 /// unsafe {
631 /// assert_eq!(x.get_unchecked(1), &2);
632 /// }
633 /// ```
634 #[stable(feature = "rust1", since = "1.0.0")]
635 #[rustc_no_implicit_autorefs]
636 #[inline]
637 #[must_use]
638 #[track_caller]
639 #[rustc_const_unstable(feature = "const_index", issue = "143775")]
640 pub const unsafe fn get_unchecked<I>(&self, index: I) -> &I::Output
641 where
642 I: [const] SliceIndex<Self>,
643 {
644 // SAFETY: the caller must uphold most of the safety requirements for `get_unchecked`;
645 // the slice is dereferenceable because `self` is a safe reference.
646 // The returned pointer is safe because impls of `SliceIndex` have to guarantee that it is.
647 unsafe { &*index.get_unchecked(self) }
648 }
649
650 /// Returns a mutable reference to an element or subslice, without doing
651 /// bounds checking.
652 ///
653 /// For a safe alternative see [`get_mut`].
654 ///
655 /// # Safety
656 ///
657 /// Calling this method with an out-of-bounds index is *[undefined behavior]*
658 /// even if the resulting reference is not used.
659 ///
660 /// You can think of this like `.get_mut(index).unwrap_unchecked()`. It's
661 /// UB to call `.get_unchecked_mut(len)`, even if you immediately convert
662 /// to a pointer. And it's UB to call `.get_unchecked_mut(..len + 1)`,
663 /// `.get_unchecked_mut(..=len)`, or similar.
664 ///
665 /// [`get_mut`]: slice::get_mut
666 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
667 ///
668 /// # Examples
669 ///
670 /// ```
671 /// let x = &mut [1, 2, 4];
672 ///
673 /// unsafe {
674 /// let elem = x.get_unchecked_mut(1);
675 /// *elem = 13;
676 /// }
677 /// assert_eq!(x, &[1, 13, 4]);
678 /// ```
679 #[stable(feature = "rust1", since = "1.0.0")]
680 #[rustc_no_implicit_autorefs]
681 #[inline]
682 #[must_use]
683 #[track_caller]
684 #[rustc_const_unstable(feature = "const_index", issue = "143775")]
685 #[rustc_no_writable]
686 pub const unsafe fn get_unchecked_mut<I>(&mut self, index: I) -> &mut I::Output
687 where
688 I: [const] SliceIndex<Self>,
689 {
690 // SAFETY: the caller must uphold the safety requirements for `get_unchecked_mut`;
691 // the slice is dereferenceable because `self` is a safe reference.
692 // The returned pointer is safe because impls of `SliceIndex` have to guarantee that it is.
693 unsafe { &mut *index.get_unchecked_mut(self) }
694 }
695
696 /// Returns a raw pointer to the slice's buffer.
697 ///
698 /// The caller must ensure that the slice outlives the pointer this
699 /// function returns, or else it will end up dangling.
700 ///
701 /// The caller must also ensure that the memory the pointer (non-transitively) points to
702 /// is never written to (except inside an `UnsafeCell`) using this pointer or any pointer
703 /// derived from it. If you need to mutate the contents of the slice, use [`as_mut_ptr`].
704 ///
705 /// Modifying the container referenced by this slice may cause its buffer
706 /// to be reallocated, which would also make any pointers to it invalid.
707 ///
708 /// # Examples
709 ///
710 /// ```
711 /// let x = &[1, 2, 4];
712 /// let x_ptr = x.as_ptr();
713 ///
714 /// unsafe {
715 /// for i in 0..x.len() {
716 /// assert_eq!(x.get_unchecked(i), &*x_ptr.add(i));
717 /// }
718 /// }
719 /// ```
720 ///
721 /// [`as_mut_ptr`]: slice::as_mut_ptr
722 #[stable(feature = "rust1", since = "1.0.0")]
723 #[rustc_const_stable(feature = "const_slice_as_ptr", since = "1.32.0")]
724 #[rustc_never_returns_null_ptr]
725 #[rustc_as_ptr]
726 #[inline(always)]
727 #[must_use]
728 pub const fn as_ptr(&self) -> *const T {
729 self as *const [T] as *const T
730 }
731
732 /// Returns an unsafe mutable pointer to the slice's buffer.
733 ///
734 /// The caller must ensure that the slice outlives the pointer this
735 /// function returns, or else it will end up dangling.
736 ///
737 /// Modifying the container referenced by this slice may cause its buffer
738 /// to be reallocated, which would also make any pointers to it invalid.
739 ///
740 /// # Examples
741 ///
742 /// ```
743 /// let x = &mut [1, 2, 4];
744 /// let x_ptr = x.as_mut_ptr();
745 ///
746 /// unsafe {
747 /// for i in 0..x.len() {
748 /// *x_ptr.add(i) += 2;
749 /// }
750 /// }
751 /// assert_eq!(x, &[3, 4, 6]);
752 /// ```
753 #[stable(feature = "rust1", since = "1.0.0")]
754 #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")]
755 #[rustc_never_returns_null_ptr]
756 #[rustc_as_ptr]
757 #[inline(always)]
758 #[must_use]
759 #[rustc_no_writable]
760 pub const fn as_mut_ptr(&mut self) -> *mut T {
761 self as *mut [T] as *mut T
762 }
763
764 /// Returns the two raw pointers spanning the slice.
765 ///
766 /// The returned range is half-open, which means that the end pointer
767 /// points *one past* the last element of the slice. This way, an empty
768 /// slice is represented by two equal pointers, and the difference between
769 /// the two pointers represents the size of the slice.
770 ///
771 /// See [`as_ptr`] for warnings on using these pointers. The end pointer
772 /// requires extra caution, as it does not point to a valid element in the
773 /// slice.
774 ///
775 /// This function is useful for interacting with foreign interfaces which
776 /// use two pointers to refer to a range of elements in memory, as is
777 /// common in C++.
778 ///
779 /// It can also be useful to check if a pointer to an element refers to an
780 /// element of this slice:
781 ///
782 /// ```
783 /// let a = [1, 2, 3];
784 /// let x = &a[1] as *const _;
785 /// let y = &5 as *const _;
786 ///
787 /// assert!(a.as_ptr_range().contains(&x));
788 /// assert!(!a.as_ptr_range().contains(&y));
789 /// ```
790 ///
791 /// [`as_ptr`]: slice::as_ptr
792 #[stable(feature = "slice_ptr_range", since = "1.48.0")]
793 #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")]
794 #[inline]
795 #[must_use]
796 pub const fn as_ptr_range(&self) -> Range<*const T> {
797 let start = self.as_ptr();
798 // SAFETY: The `add` here is safe, because:
799 //
800 // - Both pointers are part of the same object, as pointing directly
801 // past the object also counts.
802 //
803 // - The size of the slice is never larger than `isize::MAX` bytes, as
804 // noted here:
805 // - https://github.com/rust-lang/unsafe-code-guidelines/issues/102#issuecomment-473340447
806 // - https://doc.rust-lang.org/reference/behavior-considered-undefined.html
807 // - https://doc.rust-lang.org/core/slice/fn.from_raw_parts.html#safety
808 // (This doesn't seem normative yet, but the very same assumption is
809 // made in many places, including the Index implementation of slices.)
810 //
811 // - There is no wrapping around involved, as slices do not wrap past
812 // the end of the address space.
813 //
814 // See the documentation of [`pointer::add`].
815 let end = unsafe { start.add(self.len()) };
816 start..end
817 }
818
819 /// Returns the two unsafe mutable pointers spanning the slice.
820 ///
821 /// The returned range is half-open, which means that the end pointer
822 /// points *one past* the last element of the slice. This way, an empty
823 /// slice is represented by two equal pointers, and the difference between
824 /// the two pointers represents the size of the slice.
825 ///
826 /// See [`as_mut_ptr`] for warnings on using these pointers. The end
827 /// pointer requires extra caution, as it does not point to a valid element
828 /// in the slice.
829 ///
830 /// This function is useful for interacting with foreign interfaces which
831 /// use two pointers to refer to a range of elements in memory, as is
832 /// common in C++.
833 ///
834 /// [`as_mut_ptr`]: slice::as_mut_ptr
835 #[stable(feature = "slice_ptr_range", since = "1.48.0")]
836 #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")]
837 #[inline]
838 #[must_use]
839 pub const fn as_mut_ptr_range(&mut self) -> Range<*mut T> {
840 let start = self.as_mut_ptr();
841 // SAFETY: See as_ptr_range() above for why `add` here is safe.
842 let end = unsafe { start.add(self.len()) };
843 start..end
844 }
845
846 /// Gets a reference to the underlying array.
847 ///
848 /// If `N` is not exactly equal to the length of `self`, then this method returns `None`.
849 #[stable(feature = "core_slice_as_array", since = "1.93.0")]
850 #[rustc_const_stable(feature = "core_slice_as_array", since = "1.93.0")]
851 #[inline]
852 #[must_use]
853 pub const fn as_array<const N: usize>(&self) -> Option<&[T; N]> {
854 if self.len() == N {
855 let ptr = self.as_ptr().cast_array();
856
857 // SAFETY: The underlying array of a slice can be reinterpreted as an actual array `[T; N]` if `N` is not greater than the slice's length.
858 let me = unsafe { &*ptr };
859 Some(me)
860 } else {
861 None
862 }
863 }
864
865 /// Gets a mutable reference to the slice's underlying array.
866 ///
867 /// If `N` is not exactly equal to the length of `self`, then this method returns `None`.
868 #[stable(feature = "core_slice_as_array", since = "1.93.0")]
869 #[rustc_const_stable(feature = "core_slice_as_array", since = "1.93.0")]
870 #[inline]
871 #[must_use]
872 pub const fn as_mut_array<const N: usize>(&mut self) -> Option<&mut [T; N]> {
873 if self.len() == N {
874 let ptr = self.as_mut_ptr().cast_array();
875
876 // SAFETY: The underlying array of a slice can be reinterpreted as an actual array `[T; N]` if `N` is not greater than the slice's length.
877 let me = unsafe { &mut *ptr };
878 Some(me)
879 } else {
880 None
881 }
882 }
883
884 /// Swaps two elements in the slice.
885 ///
886 /// If `a` equals to `b`, it's guaranteed that elements won't change value.
887 ///
888 /// # Arguments
889 ///
890 /// * a - The index of the first element
891 /// * b - The index of the second element
892 ///
893 /// # Panics
894 ///
895 /// Panics if `a` or `b` are out of bounds.
896 ///
897 /// # Examples
898 ///
899 /// ```
900 /// let mut v = ["a", "b", "c", "d", "e"];
901 /// v.swap(2, 4);
902 /// assert!(v == ["a", "b", "e", "d", "c"]);
903 /// ```
904 #[stable(feature = "rust1", since = "1.0.0")]
905 #[rustc_const_stable(feature = "const_swap", since = "1.85.0")]
906 #[inline]
907 #[track_caller]
908 pub const fn swap(&mut self, a: usize, b: usize) {
909 // FIXME: use swap_unchecked here (https://github.com/rust-lang/rust/pull/88540#issuecomment-944344343)
910 // Can't take two mutable loans from one vector, so instead use raw pointers.
911 let pa = &raw mut self[a];
912 let pb = &raw mut self[b];
913 // SAFETY: `pa` and `pb` have been created from safe mutable references and refer
914 // to elements in the slice and therefore are guaranteed to be valid and aligned.
915 // Note that accessing the elements behind `a` and `b` is checked and will
916 // panic when out of bounds.
917 unsafe {
918 ptr::swap(pa, pb);
919 }
920 }
921
922 /// Swaps two elements in the slice, without doing bounds checking.
923 ///
924 /// For a safe alternative see [`swap`].
925 ///
926 /// # Arguments
927 ///
928 /// * a - The index of the first element
929 /// * b - The index of the second element
930 ///
931 /// # Safety
932 ///
933 /// Calling this method with an out-of-bounds index is *[undefined behavior]*.
934 /// The caller has to ensure that `a < self.len()` and `b < self.len()`.
935 ///
936 /// # Examples
937 ///
938 /// ```
939 /// #![feature(slice_swap_unchecked)]
940 ///
941 /// let mut v = ["a", "b", "c", "d"];
942 /// // SAFETY: we know that 1 and 3 are both indices of the slice
943 /// unsafe { v.swap_unchecked(1, 3) };
944 /// assert!(v == ["a", "d", "c", "b"]);
945 /// ```
946 ///
947 /// [`swap`]: slice::swap
948 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
949 #[unstable(feature = "slice_swap_unchecked", issue = "88539")]
950 #[track_caller]
951 pub const unsafe fn swap_unchecked(&mut self, a: usize, b: usize) {
952 assert_unsafe_precondition!(
953 check_library_ub,
954 "slice::swap_unchecked requires that the indices are within the slice",
955 (
956 len: usize = self.len(),
957 a: usize = a,
958 b: usize = b,
959 ) => a < len && b < len,
960 );
961
962 let ptr = self.as_mut_ptr();
963 // SAFETY: caller has to guarantee that `a < self.len()` and `b < self.len()`
964 unsafe {
965 ptr::swap(ptr.add(a), ptr.add(b));
966 }
967 }
968
969 /// Reverses the order of elements in the slice, in place.
970 ///
971 /// # Examples
972 ///
973 /// ```
974 /// let mut v = [1, 2, 3];
975 /// v.reverse();
976 /// assert!(v == [3, 2, 1]);
977 /// ```
978 #[stable(feature = "rust1", since = "1.0.0")]
979 #[rustc_const_stable(feature = "const_slice_reverse", since = "1.90.0")]
980 #[inline]
981 pub const fn reverse(&mut self) {
982 let half_len = self.len() / 2;
983 let Range { start, end } = self.as_mut_ptr_range();
984
985 // These slices will skip the middle item for an odd length,
986 // since that one doesn't need to move.
987 let (front_half, back_half) =
988 // SAFETY: Both are subparts of the original slice, so the memory
989 // range is valid, and they don't overlap because they're each only
990 // half (or less) of the original slice.
991 unsafe {
992 (
993 slice::from_raw_parts_mut(start, half_len),
994 slice::from_raw_parts_mut(end.sub(half_len), half_len),
995 )
996 };
997
998 // Introducing a function boundary here means that the two halves
999 // get `noalias` markers, allowing better optimization as LLVM
1000 // knows that they're disjoint, unlike in the original slice.
1001 revswap(front_half, back_half, half_len);
1002
1003 #[inline]
1004 const fn revswap<T>(a: &mut [T], b: &mut [T], n: usize) {
1005 debug_assert!(a.len() == n);
1006 debug_assert!(b.len() == n);
1007
1008 // Because this function is first compiled in isolation,
1009 // this check tells LLVM that the indexing below is
1010 // in-bounds. Then after inlining -- once the actual
1011 // lengths of the slices are known -- it's removed.
1012 // FIXME(const_trait_impl) replace with let (a, b) = (&mut a[..n], &mut b[..n]);
1013 let (a, _) = a.split_at_mut(n);
1014 let (b, _) = b.split_at_mut(n);
1015
1016 let mut i = 0;
1017 while i < n {
1018 mem::swap(&mut a[i], &mut b[n - 1 - i]);
1019 i += 1;
1020 }
1021 }
1022 }
1023
1024 /// Returns an iterator over the slice.
1025 ///
1026 /// The iterator yields all items from start to end.
1027 ///
1028 /// # Examples
1029 ///
1030 /// ```
1031 /// let x = &[1, 2, 4];
1032 /// let mut iterator = x.iter();
1033 ///
1034 /// assert_eq!(iterator.next(), Some(&1));
1035 /// assert_eq!(iterator.next(), Some(&2));
1036 /// assert_eq!(iterator.next(), Some(&4));
1037 /// assert_eq!(iterator.next(), None);
1038 /// ```
1039 #[stable(feature = "rust1", since = "1.0.0")]
1040 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1041 #[inline]
1042 #[rustc_diagnostic_item = "slice_iter"]
1043 pub const fn iter(&self) -> Iter<'_, T> {
1044 Iter::new(self)
1045 }
1046
1047 /// Returns an iterator that allows modifying each value.
1048 ///
1049 /// The iterator yields all items from start to end.
1050 ///
1051 /// # Examples
1052 ///
1053 /// ```
1054 /// let x = &mut [1, 2, 4];
1055 /// for elem in x.iter_mut() {
1056 /// *elem += 2;
1057 /// }
1058 /// assert_eq!(x, &[3, 4, 6]);
1059 /// ```
1060 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1061 #[stable(feature = "rust1", since = "1.0.0")]
1062 #[inline]
1063 pub const fn iter_mut(&mut self) -> IterMut<'_, T> {
1064 IterMut::new(self)
1065 }
1066
1067 /// Returns an iterator over all contiguous windows of length
1068 /// `size`. The windows overlap. If the slice is shorter than
1069 /// `size`, the iterator returns no values.
1070 ///
1071 /// # Panics
1072 ///
1073 /// Panics if `size` is zero.
1074 ///
1075 /// # Examples
1076 ///
1077 /// ```
1078 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1079 /// let mut iter = slice.windows(3);
1080 /// assert_eq!(iter.next().unwrap(), &['l', 'o', 'r']);
1081 /// assert_eq!(iter.next().unwrap(), &['o', 'r', 'e']);
1082 /// assert_eq!(iter.next().unwrap(), &['r', 'e', 'm']);
1083 /// assert!(iter.next().is_none());
1084 /// ```
1085 ///
1086 /// If the slice is shorter than `size`:
1087 ///
1088 /// ```
1089 /// let slice = ['f', 'o', 'o'];
1090 /// let mut iter = slice.windows(4);
1091 /// assert!(iter.next().is_none());
1092 /// ```
1093 ///
1094 /// Because the [Iterator] trait cannot represent the required lifetimes,
1095 /// there is no `windows_mut` analog to `windows`;
1096 /// `[0,1,2].windows_mut(2).collect()` would violate [the rules of references]
1097 /// (though a [LendingIterator] analog is possible). You can sometimes use
1098 /// [`Cell::as_slice_of_cells`](crate::cell::Cell::as_slice_of_cells) in
1099 /// conjunction with `windows` instead:
1100 ///
1101 /// [the rules of references]: https://doc.rust-lang.org/book/ch04-02-references-and-borrowing.html#the-rules-of-references
1102 /// [LendingIterator]: https://blog.rust-lang.org/2022/10/28/gats-stabilization.html
1103 /// ```
1104 /// use std::cell::Cell;
1105 ///
1106 /// let mut array = ['R', 'u', 's', 't', ' ', '2', '0', '1', '5'];
1107 /// let slice = &mut array[..];
1108 /// let slice_of_cells: &[Cell<char>] = Cell::from_mut(slice).as_slice_of_cells();
1109 /// for w in slice_of_cells.windows(3) {
1110 /// Cell::swap(&w[0], &w[2]);
1111 /// }
1112 /// assert_eq!(array, ['s', 't', ' ', '2', '0', '1', '5', 'u', 'R']);
1113 /// ```
1114 #[stable(feature = "rust1", since = "1.0.0")]
1115 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1116 #[inline]
1117 #[track_caller]
1118 pub const fn windows(&self, size: usize) -> Windows<'_, T> {
1119 let size = NonZero::new(size).expect("window size must be non-zero");
1120 Windows::new(self, size)
1121 }
1122
1123 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
1124 /// beginning of the slice.
1125 ///
1126 /// The chunks are slices and do not overlap. If `chunk_size` does not divide the length of the
1127 /// slice, then the last chunk will not have length `chunk_size`.
1128 ///
1129 /// See [`chunks_exact`] for a variant of this iterator that returns chunks of always exactly
1130 /// `chunk_size` elements, and [`rchunks`] for the same iterator but starting at the end of the
1131 /// slice.
1132 ///
1133 /// If your `chunk_size` is a constant, consider using [`as_chunks`] instead, which will
1134 /// give references to arrays of exactly that length, rather than slices.
1135 ///
1136 /// # Panics
1137 ///
1138 /// Panics if `chunk_size` is zero.
1139 ///
1140 /// # Examples
1141 ///
1142 /// ```
1143 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1144 /// let mut iter = slice.chunks(2);
1145 /// assert_eq!(iter.next().unwrap(), &['l', 'o']);
1146 /// assert_eq!(iter.next().unwrap(), &['r', 'e']);
1147 /// assert_eq!(iter.next().unwrap(), &['m']);
1148 /// assert!(iter.next().is_none());
1149 /// ```
1150 ///
1151 /// [`chunks_exact`]: slice::chunks_exact
1152 /// [`rchunks`]: slice::rchunks
1153 /// [`as_chunks`]: slice::as_chunks
1154 #[stable(feature = "rust1", since = "1.0.0")]
1155 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1156 #[inline]
1157 #[track_caller]
1158 pub const fn chunks(&self, chunk_size: usize) -> Chunks<'_, T> {
1159 assert!(chunk_size != 0, "chunk size must be non-zero");
1160 Chunks::new(self, chunk_size)
1161 }
1162
1163 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
1164 /// beginning of the slice.
1165 ///
1166 /// The chunks are mutable slices, and do not overlap. If `chunk_size` does not divide the
1167 /// length of the slice, then the last chunk will not have length `chunk_size`.
1168 ///
1169 /// See [`chunks_exact_mut`] for a variant of this iterator that returns chunks of always
1170 /// exactly `chunk_size` elements, and [`rchunks_mut`] for the same iterator but starting at
1171 /// the end of the slice.
1172 ///
1173 /// If your `chunk_size` is a constant, consider using [`as_chunks_mut`] instead, which will
1174 /// give references to arrays of exactly that length, rather than slices.
1175 ///
1176 /// # Panics
1177 ///
1178 /// Panics if `chunk_size` is zero.
1179 ///
1180 /// # Examples
1181 ///
1182 /// ```
1183 /// let v = &mut [0, 0, 0, 0, 0];
1184 /// let mut count = 1;
1185 ///
1186 /// for chunk in v.chunks_mut(2) {
1187 /// for elem in chunk.iter_mut() {
1188 /// *elem += count;
1189 /// }
1190 /// count += 1;
1191 /// }
1192 /// assert_eq!(v, &[1, 1, 2, 2, 3]);
1193 /// ```
1194 ///
1195 /// [`chunks_exact_mut`]: slice::chunks_exact_mut
1196 /// [`rchunks_mut`]: slice::rchunks_mut
1197 /// [`as_chunks_mut`]: slice::as_chunks_mut
1198 #[stable(feature = "rust1", since = "1.0.0")]
1199 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1200 #[inline]
1201 #[track_caller]
1202 pub const fn chunks_mut(&mut self, chunk_size: usize) -> ChunksMut<'_, T> {
1203 assert!(chunk_size != 0, "chunk size must be non-zero");
1204 ChunksMut::new(self, chunk_size)
1205 }
1206
1207 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
1208 /// beginning of the slice.
1209 ///
1210 /// The chunks are slices and do not overlap. If `chunk_size` does not divide the length of the
1211 /// slice, then the last up to `chunk_size-1` elements will be omitted and can be retrieved
1212 /// from the `remainder` function of the iterator.
1213 ///
1214 /// Due to each chunk having exactly `chunk_size` elements, the compiler can often optimize the
1215 /// resulting code better than in the case of [`chunks`].
1216 ///
1217 /// See [`chunks`] for a variant of this iterator that also returns the remainder as a smaller
1218 /// chunk, and [`rchunks_exact`] for the same iterator but starting at the end of the slice.
1219 ///
1220 /// If your `chunk_size` is a constant, consider using [`as_chunks`] instead, which will
1221 /// give references to arrays of exactly that length, rather than slices.
1222 ///
1223 /// # Panics
1224 ///
1225 /// Panics if `chunk_size` is zero.
1226 ///
1227 /// # Examples
1228 ///
1229 /// ```
1230 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1231 /// let mut iter = slice.chunks_exact(2);
1232 /// assert_eq!(iter.next().unwrap(), &['l', 'o']);
1233 /// assert_eq!(iter.next().unwrap(), &['r', 'e']);
1234 /// assert!(iter.next().is_none());
1235 /// assert_eq!(iter.remainder(), &['m']);
1236 /// ```
1237 ///
1238 /// [`chunks`]: slice::chunks
1239 /// [`rchunks_exact`]: slice::rchunks_exact
1240 /// [`as_chunks`]: slice::as_chunks
1241 #[stable(feature = "chunks_exact", since = "1.31.0")]
1242 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1243 #[inline]
1244 #[track_caller]
1245 pub const fn chunks_exact(&self, chunk_size: usize) -> ChunksExact<'_, T> {
1246 assert!(chunk_size != 0, "chunk size must be non-zero");
1247 ChunksExact::new(self, chunk_size)
1248 }
1249
1250 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
1251 /// beginning of the slice.
1252 ///
1253 /// The chunks are mutable slices, and do not overlap. If `chunk_size` does not divide the
1254 /// length of the slice, then the last up to `chunk_size-1` elements will be omitted and can be
1255 /// retrieved from the `into_remainder` function of the iterator.
1256 ///
1257 /// Due to each chunk having exactly `chunk_size` elements, the compiler can often optimize the
1258 /// resulting code better than in the case of [`chunks_mut`].
1259 ///
1260 /// See [`chunks_mut`] for a variant of this iterator that also returns the remainder as a
1261 /// smaller chunk, and [`rchunks_exact_mut`] for the same iterator but starting at the end of
1262 /// the slice.
1263 ///
1264 /// If your `chunk_size` is a constant, consider using [`as_chunks_mut`] instead, which will
1265 /// give references to arrays of exactly that length, rather than slices.
1266 ///
1267 /// # Panics
1268 ///
1269 /// Panics if `chunk_size` is zero.
1270 ///
1271 /// # Examples
1272 ///
1273 /// ```
1274 /// let v = &mut [0, 0, 0, 0, 0];
1275 /// let mut count = 1;
1276 ///
1277 /// for chunk in v.chunks_exact_mut(2) {
1278 /// for elem in chunk.iter_mut() {
1279 /// *elem += count;
1280 /// }
1281 /// count += 1;
1282 /// }
1283 /// assert_eq!(v, &[1, 1, 2, 2, 0]);
1284 /// ```
1285 ///
1286 /// [`chunks_mut`]: slice::chunks_mut
1287 /// [`rchunks_exact_mut`]: slice::rchunks_exact_mut
1288 /// [`as_chunks_mut`]: slice::as_chunks_mut
1289 #[stable(feature = "chunks_exact", since = "1.31.0")]
1290 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1291 #[inline]
1292 #[track_caller]
1293 pub const fn chunks_exact_mut(&mut self, chunk_size: usize) -> ChunksExactMut<'_, T> {
1294 assert!(chunk_size != 0, "chunk size must be non-zero");
1295 ChunksExactMut::new(self, chunk_size)
1296 }
1297
1298 /// Splits the slice into a slice of `N`-element arrays,
1299 /// assuming that there's no remainder.
1300 ///
1301 /// This is the inverse operation to [`as_flattened`].
1302 ///
1303 /// [`as_flattened`]: slice::as_flattened
1304 ///
1305 /// As this is `unsafe`, consider whether you could use [`as_chunks`] or
1306 /// [`as_rchunks`] instead, perhaps via something like
1307 /// `if let (chunks, []) = slice.as_chunks()` or
1308 /// `let (chunks, []) = slice.as_chunks() else { unreachable!() };`.
1309 ///
1310 /// [`as_chunks`]: slice::as_chunks
1311 /// [`as_rchunks`]: slice::as_rchunks
1312 ///
1313 /// # Safety
1314 ///
1315 /// This may only be called when
1316 /// - The slice splits exactly into `N`-element chunks (aka `self.len() % N == 0`).
1317 /// - `N != 0`.
1318 ///
1319 /// # Examples
1320 ///
1321 /// ```
1322 /// let slice: &[char] = &['l', 'o', 'r', 'e', 'm', '!'];
1323 /// let chunks: &[[char; 1]] =
1324 /// // SAFETY: 1-element chunks never have remainder
1325 /// unsafe { slice.as_chunks_unchecked() };
1326 /// assert_eq!(chunks, &[['l'], ['o'], ['r'], ['e'], ['m'], ['!']]);
1327 /// let chunks: &[[char; 3]] =
1328 /// // SAFETY: The slice length (6) is a multiple of 3
1329 /// unsafe { slice.as_chunks_unchecked() };
1330 /// assert_eq!(chunks, &[['l', 'o', 'r'], ['e', 'm', '!']]);
1331 ///
1332 /// // These would be unsound:
1333 /// // let chunks: &[[_; 5]] = slice.as_chunks_unchecked() // The slice length is not a multiple of 5
1334 /// // let chunks: &[[_; 0]] = slice.as_chunks_unchecked() // Zero-length chunks are never allowed
1335 /// ```
1336 #[stable(feature = "slice_as_chunks", since = "1.88.0")]
1337 #[rustc_const_stable(feature = "slice_as_chunks", since = "1.88.0")]
1338 #[inline]
1339 #[must_use]
1340 #[track_caller]
1341 pub const unsafe fn as_chunks_unchecked<const N: usize>(&self) -> &[[T; N]] {
1342 assert_unsafe_precondition!(
1343 check_language_ub,
1344 "slice::as_chunks_unchecked requires `N != 0` and the slice to split exactly into `N`-element chunks",
1345 (n: usize = N, len: usize = self.len()) => n != 0 && len.is_multiple_of(n),
1346 );
1347 // SAFETY: Caller must guarantee that `N` is nonzero and exactly divides the slice length
1348 let new_len = unsafe { exact_div(self.len(), N) };
1349 // SAFETY: We cast a slice of `new_len * N` elements into
1350 // a slice of `new_len` many `N` elements chunks.
1351 unsafe { from_raw_parts(self.as_ptr().cast(), new_len) }
1352 }
1353
1354 /// Splits the slice into a slice of `N`-element arrays,
1355 /// starting at the beginning of the slice,
1356 /// and a remainder slice with length strictly less than `N`.
1357 ///
1358 /// The remainder is meaningful in the division sense. Given
1359 /// `let (chunks, remainder) = slice.as_chunks()`, then:
1360 /// - `chunks.len()` equals `slice.len() / N`,
1361 /// - `remainder.len()` equals `slice.len() % N`, and
1362 /// - `slice.len()` equals `chunks.len() * N + remainder.len()`.
1363 ///
1364 /// You can flatten the chunks back into a slice-of-`T` with [`as_flattened`].
1365 ///
1366 /// [`as_flattened`]: slice::as_flattened
1367 ///
1368 /// # Panics
1369 ///
1370 /// Panics if `N` is zero.
1371 ///
1372 /// Note that this check is against a const generic parameter, not a runtime
1373 /// value, and thus a particular monomorphization will either always panic
1374 /// or it will never panic.
1375 ///
1376 /// # Examples
1377 ///
1378 /// ```
1379 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1380 /// let (chunks, remainder) = slice.as_chunks();
1381 /// assert_eq!(chunks, &[['l', 'o'], ['r', 'e']]);
1382 /// assert_eq!(remainder, &['m']);
1383 /// ```
1384 ///
1385 /// If you expect the slice to be an exact multiple, you can combine
1386 /// `let`-`else` with an empty slice pattern:
1387 /// ```
1388 /// let slice = ['R', 'u', 's', 't'];
1389 /// let (chunks, []) = slice.as_chunks::<2>() else {
1390 /// panic!("slice didn't have even length")
1391 /// };
1392 /// assert_eq!(chunks, &[['R', 'u'], ['s', 't']]);
1393 /// ```
1394 #[stable(feature = "slice_as_chunks", since = "1.88.0")]
1395 #[rustc_const_stable(feature = "slice_as_chunks", since = "1.88.0")]
1396 #[inline]
1397 #[track_caller]
1398 #[must_use]
1399 pub const fn as_chunks<const N: usize>(&self) -> (&[[T; N]], &[T]) {
1400 assert!(N != 0, "chunk size must be non-zero");
1401 let len_rounded_down = self.len() / N * N;
1402 // SAFETY: The rounded-down value is always the same or smaller than the
1403 // original length, and thus must be in-bounds of the slice.
1404 let (multiple_of_n, remainder) = unsafe { self.split_at_unchecked(len_rounded_down) };
1405 // SAFETY: We already panicked for zero, and ensured by construction
1406 // that the length of the subslice is a multiple of N.
1407 let array_slice = unsafe { multiple_of_n.as_chunks_unchecked() };
1408 (array_slice, remainder)
1409 }
1410
1411 /// Splits the slice into a slice of `N`-element arrays,
1412 /// starting at the end of the slice,
1413 /// and a remainder slice with length strictly less than `N`.
1414 ///
1415 /// The remainder is meaningful in the division sense. Given
1416 /// `let (remainder, chunks) = slice.as_rchunks()`, then:
1417 /// - `remainder.len()` equals `slice.len() % N`,
1418 /// - `chunks.len()` equals `slice.len() / N`, and
1419 /// - `slice.len()` equals `chunks.len() * N + remainder.len()`.
1420 ///
1421 /// You can flatten the chunks back into a slice-of-`T` with [`as_flattened`].
1422 ///
1423 /// [`as_flattened`]: slice::as_flattened
1424 ///
1425 /// # Panics
1426 ///
1427 /// Panics if `N` is zero.
1428 ///
1429 /// Note that this check is against a const generic parameter, not a runtime
1430 /// value, and thus a particular monomorphization will either always panic
1431 /// or it will never panic.
1432 ///
1433 /// # Examples
1434 ///
1435 /// ```
1436 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1437 /// let (remainder, chunks) = slice.as_rchunks();
1438 /// assert_eq!(remainder, &['l']);
1439 /// assert_eq!(chunks, &[['o', 'r'], ['e', 'm']]);
1440 /// ```
1441 #[stable(feature = "slice_as_chunks", since = "1.88.0")]
1442 #[rustc_const_stable(feature = "slice_as_chunks", since = "1.88.0")]
1443 #[inline]
1444 #[track_caller]
1445 #[must_use]
1446 pub const fn as_rchunks<const N: usize>(&self) -> (&[T], &[[T; N]]) {
1447 assert!(N != 0, "chunk size must be non-zero");
1448 let len = self.len() / N;
1449 let (remainder, multiple_of_n) = self.split_at(self.len() - len * N);
1450 // SAFETY: We already panicked for zero, and ensured by construction
1451 // that the length of the subslice is a multiple of N.
1452 let array_slice = unsafe { multiple_of_n.as_chunks_unchecked() };
1453 (remainder, array_slice)
1454 }
1455
1456 /// Splits the slice into a slice of `N`-element arrays,
1457 /// assuming that there's no remainder.
1458 ///
1459 /// This is the inverse operation to [`as_flattened_mut`].
1460 ///
1461 /// [`as_flattened_mut`]: slice::as_flattened_mut
1462 ///
1463 /// As this is `unsafe`, consider whether you could use [`as_chunks_mut`] or
1464 /// [`as_rchunks_mut`] instead, perhaps via something like
1465 /// `if let (chunks, []) = slice.as_chunks_mut()` or
1466 /// `let (chunks, []) = slice.as_chunks_mut() else { unreachable!() };`.
1467 ///
1468 /// [`as_chunks_mut`]: slice::as_chunks_mut
1469 /// [`as_rchunks_mut`]: slice::as_rchunks_mut
1470 ///
1471 /// # Safety
1472 ///
1473 /// This may only be called when
1474 /// - The slice splits exactly into `N`-element chunks (aka `self.len() % N == 0`).
1475 /// - `N != 0`.
1476 ///
1477 /// # Examples
1478 ///
1479 /// ```
1480 /// let slice: &mut [char] = &mut ['l', 'o', 'r', 'e', 'm', '!'];
1481 /// let chunks: &mut [[char; 1]] =
1482 /// // SAFETY: 1-element chunks never have remainder
1483 /// unsafe { slice.as_chunks_unchecked_mut() };
1484 /// chunks[0] = ['L'];
1485 /// assert_eq!(chunks, &[['L'], ['o'], ['r'], ['e'], ['m'], ['!']]);
1486 /// let chunks: &mut [[char; 3]] =
1487 /// // SAFETY: The slice length (6) is a multiple of 3
1488 /// unsafe { slice.as_chunks_unchecked_mut() };
1489 /// chunks[1] = ['a', 'x', '?'];
1490 /// assert_eq!(slice, &['L', 'o', 'r', 'a', 'x', '?']);
1491 ///
1492 /// // These would be unsound:
1493 /// // let chunks: &[[_; 5]] = slice.as_chunks_unchecked_mut() // The slice length is not a multiple of 5
1494 /// // let chunks: &[[_; 0]] = slice.as_chunks_unchecked_mut() // Zero-length chunks are never allowed
1495 /// ```
1496 #[stable(feature = "slice_as_chunks", since = "1.88.0")]
1497 #[rustc_const_stable(feature = "slice_as_chunks", since = "1.88.0")]
1498 #[inline]
1499 #[must_use]
1500 #[track_caller]
1501 pub const unsafe fn as_chunks_unchecked_mut<const N: usize>(&mut self) -> &mut [[T; N]] {
1502 assert_unsafe_precondition!(
1503 check_language_ub,
1504 "slice::as_chunks_unchecked requires `N != 0` and the slice to split exactly into `N`-element chunks",
1505 (n: usize = N, len: usize = self.len()) => n != 0 && len.is_multiple_of(n)
1506 );
1507 // SAFETY: Caller must guarantee that `N` is nonzero and exactly divides the slice length
1508 let new_len = unsafe { exact_div(self.len(), N) };
1509 // SAFETY: We cast a slice of `new_len * N` elements into
1510 // a slice of `new_len` many `N` elements chunks.
1511 unsafe { from_raw_parts_mut(self.as_mut_ptr().cast(), new_len) }
1512 }
1513
1514 /// Splits the slice into a slice of `N`-element arrays,
1515 /// starting at the beginning of the slice,
1516 /// and a remainder slice with length strictly less than `N`.
1517 ///
1518 /// The remainder is meaningful in the division sense. Given
1519 /// `let (chunks, remainder) = slice.as_chunks_mut()`, then:
1520 /// - `chunks.len()` equals `slice.len() / N`,
1521 /// - `remainder.len()` equals `slice.len() % N`, and
1522 /// - `slice.len()` equals `chunks.len() * N + remainder.len()`.
1523 ///
1524 /// You can flatten the chunks back into a slice-of-`T` with [`as_flattened_mut`].
1525 ///
1526 /// [`as_flattened_mut`]: slice::as_flattened_mut
1527 ///
1528 /// # Panics
1529 ///
1530 /// Panics if `N` is zero.
1531 ///
1532 /// Note that this check is against a const generic parameter, not a runtime
1533 /// value, and thus a particular monomorphization will either always panic
1534 /// or it will never panic.
1535 ///
1536 /// # Examples
1537 ///
1538 /// ```
1539 /// let v = &mut [0, 0, 0, 0, 0];
1540 /// let mut count = 1;
1541 ///
1542 /// let (chunks, remainder) = v.as_chunks_mut();
1543 /// remainder[0] = 9;
1544 /// for chunk in chunks {
1545 /// *chunk = [count; 2];
1546 /// count += 1;
1547 /// }
1548 /// assert_eq!(v, &[1, 1, 2, 2, 9]);
1549 /// ```
1550 #[stable(feature = "slice_as_chunks", since = "1.88.0")]
1551 #[rustc_const_stable(feature = "slice_as_chunks", since = "1.88.0")]
1552 #[inline]
1553 #[track_caller]
1554 #[must_use]
1555 pub const fn as_chunks_mut<const N: usize>(&mut self) -> (&mut [[T; N]], &mut [T]) {
1556 assert!(N != 0, "chunk size must be non-zero");
1557 let len_rounded_down = self.len() / N * N;
1558 // SAFETY: The rounded-down value is always the same or smaller than the
1559 // original length, and thus must be in-bounds of the slice.
1560 let (multiple_of_n, remainder) = unsafe { self.split_at_mut_unchecked(len_rounded_down) };
1561 // SAFETY: We already panicked for zero, and ensured by construction
1562 // that the length of the subslice is a multiple of N.
1563 let array_slice = unsafe { multiple_of_n.as_chunks_unchecked_mut() };
1564 (array_slice, remainder)
1565 }
1566
1567 /// Splits the slice into a slice of `N`-element arrays,
1568 /// starting at the end of the slice,
1569 /// and a remainder slice with length strictly less than `N`.
1570 ///
1571 /// The remainder is meaningful in the division sense. Given
1572 /// `let (remainder, chunks) = slice.as_rchunks_mut()`, then:
1573 /// - `remainder.len()` equals `slice.len() % N`,
1574 /// - `chunks.len()` equals `slice.len() / N`, and
1575 /// - `slice.len()` equals `chunks.len() * N + remainder.len()`.
1576 ///
1577 /// You can flatten the chunks back into a slice-of-`T` with [`as_flattened_mut`].
1578 ///
1579 /// [`as_flattened_mut`]: slice::as_flattened_mut
1580 ///
1581 /// # Panics
1582 ///
1583 /// Panics if `N` is zero.
1584 ///
1585 /// Note that this check is against a const generic parameter, not a runtime
1586 /// value, and thus a particular monomorphization will either always panic
1587 /// or it will never panic.
1588 ///
1589 /// # Examples
1590 ///
1591 /// ```
1592 /// let v = &mut [0, 0, 0, 0, 0];
1593 /// let mut count = 1;
1594 ///
1595 /// let (remainder, chunks) = v.as_rchunks_mut();
1596 /// remainder[0] = 9;
1597 /// for chunk in chunks {
1598 /// *chunk = [count; 2];
1599 /// count += 1;
1600 /// }
1601 /// assert_eq!(v, &[9, 1, 1, 2, 2]);
1602 /// ```
1603 #[stable(feature = "slice_as_chunks", since = "1.88.0")]
1604 #[rustc_const_stable(feature = "slice_as_chunks", since = "1.88.0")]
1605 #[inline]
1606 #[track_caller]
1607 #[must_use]
1608 pub const fn as_rchunks_mut<const N: usize>(&mut self) -> (&mut [T], &mut [[T; N]]) {
1609 assert!(N != 0, "chunk size must be non-zero");
1610 let len = self.len() / N;
1611 let (remainder, multiple_of_n) = self.split_at_mut(self.len() - len * N);
1612 // SAFETY: We already panicked for zero, and ensured by construction
1613 // that the length of the subslice is a multiple of N.
1614 let array_slice = unsafe { multiple_of_n.as_chunks_unchecked_mut() };
1615 (remainder, array_slice)
1616 }
1617
1618 /// Returns an iterator over overlapping windows of `N` elements of a slice,
1619 /// starting at the beginning of the slice.
1620 ///
1621 /// This is the const generic equivalent of [`windows`].
1622 ///
1623 /// If `N` is greater than the size of the slice, it will return no windows.
1624 ///
1625 /// # Panics
1626 ///
1627 /// Panics if `N` is zero.
1628 ///
1629 /// Note that this check is against a const generic parameter, not a runtime
1630 /// value, and thus a particular monomorphization will either always panic
1631 /// or it will never panic.
1632 ///
1633 /// # Examples
1634 ///
1635 /// ```
1636 /// let slice = [0, 1, 2, 3];
1637 /// let mut iter = slice.array_windows();
1638 /// assert_eq!(iter.next().unwrap(), &[0, 1]);
1639 /// assert_eq!(iter.next().unwrap(), &[1, 2]);
1640 /// assert_eq!(iter.next().unwrap(), &[2, 3]);
1641 /// assert!(iter.next().is_none());
1642 /// ```
1643 ///
1644 /// [`windows`]: slice::windows
1645 #[stable(feature = "array_windows", since = "1.94.0")]
1646 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1647 #[inline]
1648 #[track_caller]
1649 pub const fn array_windows<const N: usize>(&self) -> ArrayWindows<'_, T, N> {
1650 assert!(N != 0, "window size must be non-zero");
1651 ArrayWindows::new(self)
1652 }
1653
1654 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the end
1655 /// of the slice.
1656 ///
1657 /// The chunks are slices and do not overlap. If `chunk_size` does not divide the length of the
1658 /// slice, then the last chunk will not have length `chunk_size`.
1659 ///
1660 /// See [`rchunks_exact`] for a variant of this iterator that returns chunks of always exactly
1661 /// `chunk_size` elements, and [`chunks`] for the same iterator but starting at the beginning
1662 /// of the slice.
1663 ///
1664 /// If your `chunk_size` is a constant, consider using [`as_rchunks`] instead, which will
1665 /// give references to arrays of exactly that length, rather than slices.
1666 ///
1667 /// # Panics
1668 ///
1669 /// Panics if `chunk_size` is zero.
1670 ///
1671 /// # Examples
1672 ///
1673 /// ```
1674 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1675 /// let mut iter = slice.rchunks(2);
1676 /// assert_eq!(iter.next().unwrap(), &['e', 'm']);
1677 /// assert_eq!(iter.next().unwrap(), &['o', 'r']);
1678 /// assert_eq!(iter.next().unwrap(), &['l']);
1679 /// assert!(iter.next().is_none());
1680 /// ```
1681 ///
1682 /// [`rchunks_exact`]: slice::rchunks_exact
1683 /// [`chunks`]: slice::chunks
1684 /// [`as_rchunks`]: slice::as_rchunks
1685 #[stable(feature = "rchunks", since = "1.31.0")]
1686 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1687 #[inline]
1688 #[track_caller]
1689 pub const fn rchunks(&self, chunk_size: usize) -> RChunks<'_, T> {
1690 assert!(chunk_size != 0, "chunk size must be non-zero");
1691 RChunks::new(self, chunk_size)
1692 }
1693
1694 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the end
1695 /// of the slice.
1696 ///
1697 /// The chunks are mutable slices, and do not overlap. If `chunk_size` does not divide the
1698 /// length of the slice, then the last chunk will not have length `chunk_size`.
1699 ///
1700 /// See [`rchunks_exact_mut`] for a variant of this iterator that returns chunks of always
1701 /// exactly `chunk_size` elements, and [`chunks_mut`] for the same iterator but starting at the
1702 /// beginning of the slice.
1703 ///
1704 /// If your `chunk_size` is a constant, consider using [`as_rchunks_mut`] instead, which will
1705 /// give references to arrays of exactly that length, rather than slices.
1706 ///
1707 /// # Panics
1708 ///
1709 /// Panics if `chunk_size` is zero.
1710 ///
1711 /// # Examples
1712 ///
1713 /// ```
1714 /// let v = &mut [0, 0, 0, 0, 0];
1715 /// let mut count = 1;
1716 ///
1717 /// for chunk in v.rchunks_mut(2) {
1718 /// for elem in chunk.iter_mut() {
1719 /// *elem += count;
1720 /// }
1721 /// count += 1;
1722 /// }
1723 /// assert_eq!(v, &[3, 2, 2, 1, 1]);
1724 /// ```
1725 ///
1726 /// [`rchunks_exact_mut`]: slice::rchunks_exact_mut
1727 /// [`chunks_mut`]: slice::chunks_mut
1728 /// [`as_rchunks_mut`]: slice::as_rchunks_mut
1729 #[stable(feature = "rchunks", since = "1.31.0")]
1730 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1731 #[inline]
1732 #[track_caller]
1733 pub const fn rchunks_mut(&mut self, chunk_size: usize) -> RChunksMut<'_, T> {
1734 assert!(chunk_size != 0, "chunk size must be non-zero");
1735 RChunksMut::new(self, chunk_size)
1736 }
1737
1738 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
1739 /// end of the slice.
1740 ///
1741 /// The chunks are slices and do not overlap. If `chunk_size` does not divide the length of the
1742 /// slice, then the last up to `chunk_size-1` elements will be omitted and can be retrieved
1743 /// from the `remainder` function of the iterator.
1744 ///
1745 /// Due to each chunk having exactly `chunk_size` elements, the compiler can often optimize the
1746 /// resulting code better than in the case of [`rchunks`].
1747 ///
1748 /// See [`rchunks`] for a variant of this iterator that also returns the remainder as a smaller
1749 /// chunk, and [`chunks_exact`] for the same iterator but starting at the beginning of the
1750 /// slice.
1751 ///
1752 /// If your `chunk_size` is a constant, consider using [`as_rchunks`] instead, which will
1753 /// give references to arrays of exactly that length, rather than slices.
1754 ///
1755 /// # Panics
1756 ///
1757 /// Panics if `chunk_size` is zero.
1758 ///
1759 /// # Examples
1760 ///
1761 /// ```
1762 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1763 /// let mut iter = slice.rchunks_exact(2);
1764 /// assert_eq!(iter.next().unwrap(), &['e', 'm']);
1765 /// assert_eq!(iter.next().unwrap(), &['o', 'r']);
1766 /// assert!(iter.next().is_none());
1767 /// assert_eq!(iter.remainder(), &['l']);
1768 /// ```
1769 ///
1770 /// [`chunks`]: slice::chunks
1771 /// [`rchunks`]: slice::rchunks
1772 /// [`chunks_exact`]: slice::chunks_exact
1773 /// [`as_rchunks`]: slice::as_rchunks
1774 #[stable(feature = "rchunks", since = "1.31.0")]
1775 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1776 #[inline]
1777 #[track_caller]
1778 pub const fn rchunks_exact(&self, chunk_size: usize) -> RChunksExact<'_, T> {
1779 assert!(chunk_size != 0, "chunk size must be non-zero");
1780 RChunksExact::new(self, chunk_size)
1781 }
1782
1783 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the end
1784 /// of the slice.
1785 ///
1786 /// The chunks are mutable slices, and do not overlap. If `chunk_size` does not divide the
1787 /// length of the slice, then the last up to `chunk_size-1` elements will be omitted and can be
1788 /// retrieved from the `into_remainder` function of the iterator.
1789 ///
1790 /// Due to each chunk having exactly `chunk_size` elements, the compiler can often optimize the
1791 /// resulting code better than in the case of [`chunks_mut`].
1792 ///
1793 /// See [`rchunks_mut`] for a variant of this iterator that also returns the remainder as a
1794 /// smaller chunk, and [`chunks_exact_mut`] for the same iterator but starting at the beginning
1795 /// of the slice.
1796 ///
1797 /// If your `chunk_size` is a constant, consider using [`as_rchunks_mut`] instead, which will
1798 /// give references to arrays of exactly that length, rather than slices.
1799 ///
1800 /// # Panics
1801 ///
1802 /// Panics if `chunk_size` is zero.
1803 ///
1804 /// # Examples
1805 ///
1806 /// ```
1807 /// let v = &mut [0, 0, 0, 0, 0];
1808 /// let mut count = 1;
1809 ///
1810 /// for chunk in v.rchunks_exact_mut(2) {
1811 /// for elem in chunk.iter_mut() {
1812 /// *elem += count;
1813 /// }
1814 /// count += 1;
1815 /// }
1816 /// assert_eq!(v, &[0, 2, 2, 1, 1]);
1817 /// ```
1818 ///
1819 /// [`chunks_mut`]: slice::chunks_mut
1820 /// [`rchunks_mut`]: slice::rchunks_mut
1821 /// [`chunks_exact_mut`]: slice::chunks_exact_mut
1822 /// [`as_rchunks_mut`]: slice::as_rchunks_mut
1823 #[stable(feature = "rchunks", since = "1.31.0")]
1824 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1825 #[inline]
1826 #[track_caller]
1827 pub const fn rchunks_exact_mut(&mut self, chunk_size: usize) -> RChunksExactMut<'_, T> {
1828 assert!(chunk_size != 0, "chunk size must be non-zero");
1829 RChunksExactMut::new(self, chunk_size)
1830 }
1831
1832 /// Returns an iterator over the slice producing non-overlapping runs
1833 /// of elements using the predicate to separate them.
1834 ///
1835 /// The predicate is called for every pair of consecutive elements,
1836 /// meaning that it is called on `slice[0]` and `slice[1]`,
1837 /// followed by `slice[1]` and `slice[2]`, and so on.
1838 ///
1839 /// # Examples
1840 ///
1841 /// ```
1842 /// let slice = &[1, 1, 1, 3, 3, 2, 2, 2];
1843 ///
1844 /// let mut iter = slice.chunk_by(|a, b| a == b);
1845 ///
1846 /// assert_eq!(iter.next(), Some(&[1, 1, 1][..]));
1847 /// assert_eq!(iter.next(), Some(&[3, 3][..]));
1848 /// assert_eq!(iter.next(), Some(&[2, 2, 2][..]));
1849 /// assert_eq!(iter.next(), None);
1850 /// ```
1851 ///
1852 /// This method can be used to extract the sorted subslices:
1853 ///
1854 /// ```
1855 /// let slice = &[1, 1, 2, 3, 2, 3, 2, 3, 4];
1856 ///
1857 /// let mut iter = slice.chunk_by(|a, b| a <= b);
1858 ///
1859 /// assert_eq!(iter.next(), Some(&[1, 1, 2, 3][..]));
1860 /// assert_eq!(iter.next(), Some(&[2, 3][..]));
1861 /// assert_eq!(iter.next(), Some(&[2, 3, 4][..]));
1862 /// assert_eq!(iter.next(), None);
1863 /// ```
1864 #[stable(feature = "slice_group_by", since = "1.77.0")]
1865 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1866 #[inline]
1867 pub const fn chunk_by<F>(&self, pred: F) -> ChunkBy<'_, T, F>
1868 where
1869 F: FnMut(&T, &T) -> bool,
1870 {
1871 ChunkBy::new(self, pred)
1872 }
1873
1874 /// Returns an iterator over the slice producing non-overlapping mutable
1875 /// runs of elements using the predicate to separate them.
1876 ///
1877 /// The predicate is called for every pair of consecutive elements,
1878 /// meaning that it is called on `slice[0]` and `slice[1]`,
1879 /// followed by `slice[1]` and `slice[2]`, and so on.
1880 ///
1881 /// # Examples
1882 ///
1883 /// ```
1884 /// let slice = &mut [1, 1, 1, 3, 3, 2, 2, 2];
1885 ///
1886 /// let mut iter = slice.chunk_by_mut(|a, b| a == b);
1887 ///
1888 /// assert_eq!(iter.next(), Some(&mut [1, 1, 1][..]));
1889 /// assert_eq!(iter.next(), Some(&mut [3, 3][..]));
1890 /// assert_eq!(iter.next(), Some(&mut [2, 2, 2][..]));
1891 /// assert_eq!(iter.next(), None);
1892 /// ```
1893 ///
1894 /// This method can be used to extract the sorted subslices:
1895 ///
1896 /// ```
1897 /// let slice = &mut [1, 1, 2, 3, 2, 3, 2, 3, 4];
1898 ///
1899 /// let mut iter = slice.chunk_by_mut(|a, b| a <= b);
1900 ///
1901 /// assert_eq!(iter.next(), Some(&mut [1, 1, 2, 3][..]));
1902 /// assert_eq!(iter.next(), Some(&mut [2, 3][..]));
1903 /// assert_eq!(iter.next(), Some(&mut [2, 3, 4][..]));
1904 /// assert_eq!(iter.next(), None);
1905 /// ```
1906 #[stable(feature = "slice_group_by", since = "1.77.0")]
1907 #[rustc_const_unstable(feature = "const_slice_make_iter", issue = "137737")]
1908 #[inline]
1909 pub const fn chunk_by_mut<F>(&mut self, pred: F) -> ChunkByMut<'_, T, F>
1910 where
1911 F: FnMut(&T, &T) -> bool,
1912 {
1913 ChunkByMut::new(self, pred)
1914 }
1915
1916 /// Divides one slice into two at an index.
1917 ///
1918 /// The first will contain all indices from `[0, mid)` (excluding
1919 /// the index `mid` itself) and the second will contain all
1920 /// indices from `[mid, len)` (excluding the index `len` itself).
1921 ///
1922 /// # Panics
1923 ///
1924 /// Panics if `mid > len`. For a non-panicking alternative see
1925 /// [`split_at_checked`](slice::split_at_checked).
1926 ///
1927 /// # Examples
1928 ///
1929 /// ```
1930 /// let v = ['a', 'b', 'c'];
1931 ///
1932 /// {
1933 /// let (left, right) = v.split_at(0);
1934 /// assert_eq!(left, []);
1935 /// assert_eq!(right, ['a', 'b', 'c']);
1936 /// }
1937 ///
1938 /// {
1939 /// let (left, right) = v.split_at(2);
1940 /// assert_eq!(left, ['a', 'b']);
1941 /// assert_eq!(right, ['c']);
1942 /// }
1943 ///
1944 /// {
1945 /// let (left, right) = v.split_at(3);
1946 /// assert_eq!(left, ['a', 'b', 'c']);
1947 /// assert_eq!(right, []);
1948 /// }
1949 /// ```
1950 #[stable(feature = "rust1", since = "1.0.0")]
1951 #[rustc_const_stable(feature = "const_slice_split_at_not_mut", since = "1.71.0")]
1952 #[inline]
1953 #[track_caller]
1954 #[must_use]
1955 pub const fn split_at(&self, mid: usize) -> (&[T], &[T]) {
1956 match self.split_at_checked(mid) {
1957 Some(pair) => pair,
1958 None => panic!("mid > len"),
1959 }
1960 }
1961
1962 /// Divides one mutable slice into two at an index.
1963 ///
1964 /// The first will contain all indices from `[0, mid)` (excluding
1965 /// the index `mid` itself) and the second will contain all
1966 /// indices from `[mid, len)` (excluding the index `len` itself).
1967 ///
1968 /// # Panics
1969 ///
1970 /// Panics if `mid > len`. For a non-panicking alternative see
1971 /// [`split_at_mut_checked`](slice::split_at_mut_checked).
1972 ///
1973 /// # Examples
1974 ///
1975 /// ```
1976 /// let mut v = [1, 0, 3, 0, 5, 6];
1977 /// let (left, right) = v.split_at_mut(2);
1978 /// assert_eq!(left, [1, 0]);
1979 /// assert_eq!(right, [3, 0, 5, 6]);
1980 /// left[1] = 2;
1981 /// right[1] = 4;
1982 /// assert_eq!(v, [1, 2, 3, 4, 5, 6]);
1983 /// ```
1984 #[stable(feature = "rust1", since = "1.0.0")]
1985 #[inline]
1986 #[track_caller]
1987 #[must_use]
1988 #[rustc_const_stable(feature = "const_slice_split_at_mut", since = "1.83.0")]
1989 pub const fn split_at_mut(&mut self, mid: usize) -> (&mut [T], &mut [T]) {
1990 match self.split_at_mut_checked(mid) {
1991 Some(pair) => pair,
1992 None => panic!("mid > len"),
1993 }
1994 }
1995
1996 /// Divides one slice into two at an index, without doing bounds checking.
1997 ///
1998 /// The first will contain all indices from `[0, mid)` (excluding
1999 /// the index `mid` itself) and the second will contain all
2000 /// indices from `[mid, len)` (excluding the index `len` itself).
2001 ///
2002 /// For a safe alternative see [`split_at`].
2003 ///
2004 /// # Safety
2005 ///
2006 /// Calling this method with an out-of-bounds index is *[undefined behavior]*
2007 /// even if the resulting reference is not used. The caller has to ensure that
2008 /// `0 <= mid <= self.len()`.
2009 ///
2010 /// [`split_at`]: slice::split_at
2011 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
2012 ///
2013 /// # Examples
2014 ///
2015 /// ```
2016 /// let v = ['a', 'b', 'c'];
2017 ///
2018 /// unsafe {
2019 /// let (left, right) = v.split_at_unchecked(0);
2020 /// assert_eq!(left, []);
2021 /// assert_eq!(right, ['a', 'b', 'c']);
2022 /// }
2023 ///
2024 /// unsafe {
2025 /// let (left, right) = v.split_at_unchecked(2);
2026 /// assert_eq!(left, ['a', 'b']);
2027 /// assert_eq!(right, ['c']);
2028 /// }
2029 ///
2030 /// unsafe {
2031 /// let (left, right) = v.split_at_unchecked(3);
2032 /// assert_eq!(left, ['a', 'b', 'c']);
2033 /// assert_eq!(right, []);
2034 /// }
2035 /// ```
2036 #[stable(feature = "slice_split_at_unchecked", since = "1.79.0")]
2037 #[rustc_const_stable(feature = "const_slice_split_at_unchecked", since = "1.77.0")]
2038 #[inline]
2039 #[must_use]
2040 #[track_caller]
2041 pub const unsafe fn split_at_unchecked(&self, mid: usize) -> (&[T], &[T]) {
2042 // FIXME(const-hack): the const function `from_raw_parts` is used to make this
2043 // function const; previously the implementation used
2044 // `(self.get_unchecked(..mid), self.get_unchecked(mid..))`
2045
2046 let len = self.len();
2047 let ptr = self.as_ptr();
2048
2049 assert_unsafe_precondition!(
2050 check_library_ub,
2051 "slice::split_at_unchecked requires the index to be within the slice",
2052 (mid: usize = mid, len: usize = len) => mid <= len,
2053 );
2054
2055 // SAFETY: Caller has to check that `0 <= mid <= self.len()`
2056 unsafe { (from_raw_parts(ptr, mid), from_raw_parts(ptr.add(mid), unchecked_sub(len, mid))) }
2057 }
2058
2059 /// Divides one mutable slice into two at an index, without doing bounds checking.
2060 ///
2061 /// The first will contain all indices from `[0, mid)` (excluding
2062 /// the index `mid` itself) and the second will contain all
2063 /// indices from `[mid, len)` (excluding the index `len` itself).
2064 ///
2065 /// For a safe alternative see [`split_at_mut`].
2066 ///
2067 /// # Safety
2068 ///
2069 /// Calling this method with an out-of-bounds index is *[undefined behavior]*
2070 /// even if the resulting reference is not used. The caller has to ensure that
2071 /// `0 <= mid <= self.len()`.
2072 ///
2073 /// [`split_at_mut`]: slice::split_at_mut
2074 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
2075 ///
2076 /// # Examples
2077 ///
2078 /// ```
2079 /// let mut v = [1, 0, 3, 0, 5, 6];
2080 /// // scoped to restrict the lifetime of the borrows
2081 /// unsafe {
2082 /// let (left, right) = v.split_at_mut_unchecked(2);
2083 /// assert_eq!(left, [1, 0]);
2084 /// assert_eq!(right, [3, 0, 5, 6]);
2085 /// left[1] = 2;
2086 /// right[1] = 4;
2087 /// }
2088 /// assert_eq!(v, [1, 2, 3, 4, 5, 6]);
2089 /// ```
2090 #[stable(feature = "slice_split_at_unchecked", since = "1.79.0")]
2091 #[rustc_const_stable(feature = "const_slice_split_at_mut", since = "1.83.0")]
2092 #[inline]
2093 #[must_use]
2094 #[track_caller]
2095 pub const unsafe fn split_at_mut_unchecked(&mut self, mid: usize) -> (&mut [T], &mut [T]) {
2096 let len = self.len();
2097 let ptr = self.as_mut_ptr();
2098
2099 assert_unsafe_precondition!(
2100 check_library_ub,
2101 "slice::split_at_mut_unchecked requires the index to be within the slice",
2102 (mid: usize = mid, len: usize = len) => mid <= len,
2103 );
2104
2105 // SAFETY: Caller has to check that `0 <= mid <= self.len()`.
2106 //
2107 // `[ptr; mid]` and `[mid; len]` are not overlapping, so returning a mutable reference
2108 // is fine.
2109 unsafe {
2110 (
2111 from_raw_parts_mut(ptr, mid),
2112 from_raw_parts_mut(ptr.add(mid), unchecked_sub(len, mid)),
2113 )
2114 }
2115 }
2116
2117 /// Divides one slice into two at an index, returning `None` if the slice is
2118 /// too short.
2119 ///
2120 /// If `mid ≤ len` returns a pair of slices where the first will contain all
2121 /// indices from `[0, mid)` (excluding the index `mid` itself) and the
2122 /// second will contain all indices from `[mid, len)` (excluding the index
2123 /// `len` itself).
2124 ///
2125 /// Otherwise, if `mid > len`, returns `None`.
2126 ///
2127 /// # Examples
2128 ///
2129 /// ```
2130 /// let v = [1, -2, 3, -4, 5, -6];
2131 ///
2132 /// {
2133 /// let (left, right) = v.split_at_checked(0).unwrap();
2134 /// assert_eq!(left, []);
2135 /// assert_eq!(right, [1, -2, 3, -4, 5, -6]);
2136 /// }
2137 ///
2138 /// {
2139 /// let (left, right) = v.split_at_checked(2).unwrap();
2140 /// assert_eq!(left, [1, -2]);
2141 /// assert_eq!(right, [3, -4, 5, -6]);
2142 /// }
2143 ///
2144 /// {
2145 /// let (left, right) = v.split_at_checked(6).unwrap();
2146 /// assert_eq!(left, [1, -2, 3, -4, 5, -6]);
2147 /// assert_eq!(right, []);
2148 /// }
2149 ///
2150 /// assert_eq!(None, v.split_at_checked(7));
2151 /// ```
2152 #[stable(feature = "split_at_checked", since = "1.80.0")]
2153 #[rustc_const_stable(feature = "split_at_checked", since = "1.80.0")]
2154 #[inline]
2155 #[must_use]
2156 pub const fn split_at_checked(&self, mid: usize) -> Option<(&[T], &[T])> {
2157 if mid <= self.len() {
2158 // SAFETY: `[ptr; mid]` and `[mid; len]` are inside `self`, which
2159 // fulfills the requirements of `split_at_unchecked`.
2160 Some(unsafe { self.split_at_unchecked(mid) })
2161 } else {
2162 None
2163 }
2164 }
2165
2166 /// Divides one mutable slice into two at an index, returning `None` if the
2167 /// slice is too short.
2168 ///
2169 /// If `mid ≤ len` returns a pair of slices where the first will contain all
2170 /// indices from `[0, mid)` (excluding the index `mid` itself) and the
2171 /// second will contain all indices from `[mid, len)` (excluding the index
2172 /// `len` itself).
2173 ///
2174 /// Otherwise, if `mid > len`, returns `None`.
2175 ///
2176 /// # Examples
2177 ///
2178 /// ```
2179 /// let mut v = [1, 0, 3, 0, 5, 6];
2180 ///
2181 /// if let Some((left, right)) = v.split_at_mut_checked(2) {
2182 /// assert_eq!(left, [1, 0]);
2183 /// assert_eq!(right, [3, 0, 5, 6]);
2184 /// left[1] = 2;
2185 /// right[1] = 4;
2186 /// }
2187 /// assert_eq!(v, [1, 2, 3, 4, 5, 6]);
2188 ///
2189 /// assert_eq!(None, v.split_at_mut_checked(7));
2190 /// ```
2191 #[stable(feature = "split_at_checked", since = "1.80.0")]
2192 #[rustc_const_stable(feature = "const_slice_split_at_mut", since = "1.83.0")]
2193 #[inline]
2194 #[must_use]
2195 pub const fn split_at_mut_checked(&mut self, mid: usize) -> Option<(&mut [T], &mut [T])> {
2196 if mid <= self.len() {
2197 // SAFETY: `[ptr; mid]` and `[mid; len]` are inside `self`, which
2198 // fulfills the requirements of `split_at_unchecked`.
2199 Some(unsafe { self.split_at_mut_unchecked(mid) })
2200 } else {
2201 None
2202 }
2203 }
2204
2205 /// Returns an iterator over subslices separated by elements that match
2206 /// `pred`. The matched element is not contained in the subslices.
2207 ///
2208 /// # Examples
2209 ///
2210 /// ```
2211 /// let slice = [10, 40, 33, 20];
2212 /// let mut iter = slice.split(|num| num % 3 == 0);
2213 ///
2214 /// assert_eq!(iter.next().unwrap(), &[10, 40]);
2215 /// assert_eq!(iter.next().unwrap(), &[20]);
2216 /// assert!(iter.next().is_none());
2217 /// ```
2218 ///
2219 /// If the first element is matched, an empty slice will be the first item
2220 /// returned by the iterator. Similarly, if the last element in the slice
2221 /// is matched, an empty slice will be the last item returned by the
2222 /// iterator:
2223 ///
2224 /// ```
2225 /// let slice = [10, 40, 33];
2226 /// let mut iter = slice.split(|num| num % 3 == 0);
2227 ///
2228 /// assert_eq!(iter.next().unwrap(), &[10, 40]);
2229 /// assert_eq!(iter.next().unwrap(), &[]);
2230 /// assert!(iter.next().is_none());
2231 /// ```
2232 ///
2233 /// If two matched elements are directly adjacent, an empty slice will be
2234 /// present between them:
2235 ///
2236 /// ```
2237 /// let slice = [10, 6, 33, 20];
2238 /// let mut iter = slice.split(|num| num % 3 == 0);
2239 ///
2240 /// assert_eq!(iter.next().unwrap(), &[10]);
2241 /// assert_eq!(iter.next().unwrap(), &[]);
2242 /// assert_eq!(iter.next().unwrap(), &[20]);
2243 /// assert!(iter.next().is_none());
2244 /// ```
2245 #[stable(feature = "rust1", since = "1.0.0")]
2246 #[inline]
2247 pub fn split<F>(&self, pred: F) -> Split<'_, T, F>
2248 where
2249 F: FnMut(&T) -> bool,
2250 {
2251 Split::new(self, pred)
2252 }
2253
2254 /// Returns an iterator over mutable subslices separated by elements that
2255 /// match `pred`. The matched element is not contained in the subslices.
2256 ///
2257 /// # Examples
2258 ///
2259 /// ```
2260 /// let mut v = [10, 40, 30, 20, 60, 50];
2261 ///
2262 /// for group in v.split_mut(|num| *num % 3 == 0) {
2263 /// group[0] = 1;
2264 /// }
2265 /// assert_eq!(v, [1, 40, 30, 1, 60, 1]);
2266 /// ```
2267 #[stable(feature = "rust1", since = "1.0.0")]
2268 #[inline]
2269 pub fn split_mut<F>(&mut self, pred: F) -> SplitMut<'_, T, F>
2270 where
2271 F: FnMut(&T) -> bool,
2272 {
2273 SplitMut::new(self, pred)
2274 }
2275
2276 /// Returns an iterator over subslices separated by elements that match
2277 /// `pred`. The matched element is contained in the end of the previous
2278 /// subslice as a terminator.
2279 ///
2280 /// # Examples
2281 ///
2282 /// ```
2283 /// let slice = [10, 40, 33, 20];
2284 /// let mut iter = slice.split_inclusive(|num| num % 3 == 0);
2285 ///
2286 /// assert_eq!(iter.next().unwrap(), &[10, 40, 33]);
2287 /// assert_eq!(iter.next().unwrap(), &[20]);
2288 /// assert!(iter.next().is_none());
2289 /// ```
2290 ///
2291 /// If the last element of the slice is matched,
2292 /// that element will be considered the terminator of the preceding slice.
2293 /// That slice will be the last item returned by the iterator.
2294 ///
2295 /// ```
2296 /// let slice = [3, 10, 40, 33];
2297 /// let mut iter = slice.split_inclusive(|num| num % 3 == 0);
2298 ///
2299 /// assert_eq!(iter.next().unwrap(), &[3]);
2300 /// assert_eq!(iter.next().unwrap(), &[10, 40, 33]);
2301 /// assert!(iter.next().is_none());
2302 /// ```
2303 #[stable(feature = "split_inclusive", since = "1.51.0")]
2304 #[inline]
2305 pub fn split_inclusive<F>(&self, pred: F) -> SplitInclusive<'_, T, F>
2306 where
2307 F: FnMut(&T) -> bool,
2308 {
2309 SplitInclusive::new(self, pred)
2310 }
2311
2312 /// Returns an iterator over mutable subslices separated by elements that
2313 /// match `pred`. The matched element is contained in the previous
2314 /// subslice as a terminator.
2315 ///
2316 /// # Examples
2317 ///
2318 /// ```
2319 /// let mut v = [10, 40, 30, 20, 60, 50];
2320 ///
2321 /// for group in v.split_inclusive_mut(|num| *num % 3 == 0) {
2322 /// let terminator_idx = group.len()-1;
2323 /// group[terminator_idx] = 1;
2324 /// }
2325 /// assert_eq!(v, [10, 40, 1, 20, 1, 1]);
2326 /// ```
2327 #[stable(feature = "split_inclusive", since = "1.51.0")]
2328 #[inline]
2329 pub fn split_inclusive_mut<F>(&mut self, pred: F) -> SplitInclusiveMut<'_, T, F>
2330 where
2331 F: FnMut(&T) -> bool,
2332 {
2333 SplitInclusiveMut::new(self, pred)
2334 }
2335
2336 /// Returns an iterator over subslices separated by elements that match
2337 /// `pred`, starting at the end of the slice and working backwards.
2338 /// The matched element is not contained in the subslices.
2339 ///
2340 /// # Examples
2341 ///
2342 /// ```
2343 /// let slice = [11, 22, 33, 0, 44, 55];
2344 /// let mut iter = slice.rsplit(|num| *num == 0);
2345 ///
2346 /// assert_eq!(iter.next().unwrap(), &[44, 55]);
2347 /// assert_eq!(iter.next().unwrap(), &[11, 22, 33]);
2348 /// assert_eq!(iter.next(), None);
2349 /// ```
2350 ///
2351 /// As with `split()`, if the first or last element is matched, an empty
2352 /// slice will be the first (or last) item returned by the iterator.
2353 ///
2354 /// ```
2355 /// let v = &[0, 1, 1, 2, 3, 5, 8];
2356 /// let mut it = v.rsplit(|n| *n % 2 == 0);
2357 /// assert_eq!(it.next().unwrap(), &[]);
2358 /// assert_eq!(it.next().unwrap(), &[3, 5]);
2359 /// assert_eq!(it.next().unwrap(), &[1, 1]);
2360 /// assert_eq!(it.next().unwrap(), &[]);
2361 /// assert_eq!(it.next(), None);
2362 /// ```
2363 #[stable(feature = "slice_rsplit", since = "1.27.0")]
2364 #[inline]
2365 pub fn rsplit<F>(&self, pred: F) -> RSplit<'_, T, F>
2366 where
2367 F: FnMut(&T) -> bool,
2368 {
2369 RSplit::new(self, pred)
2370 }
2371
2372 /// Returns an iterator over mutable subslices separated by elements that
2373 /// match `pred`, starting at the end of the slice and working
2374 /// backwards. The matched element is not contained in the subslices.
2375 ///
2376 /// # Examples
2377 ///
2378 /// ```
2379 /// let mut v = [100, 400, 300, 200, 600, 500];
2380 ///
2381 /// let mut count = 0;
2382 /// for group in v.rsplit_mut(|num| *num % 3 == 0) {
2383 /// count += 1;
2384 /// group[0] = count;
2385 /// }
2386 /// assert_eq!(v, [3, 400, 300, 2, 600, 1]);
2387 /// ```
2388 ///
2389 #[stable(feature = "slice_rsplit", since = "1.27.0")]
2390 #[inline]
2391 pub fn rsplit_mut<F>(&mut self, pred: F) -> RSplitMut<'_, T, F>
2392 where
2393 F: FnMut(&T) -> bool,
2394 {
2395 RSplitMut::new(self, pred)
2396 }
2397
2398 /// Returns an iterator over subslices separated by elements that match
2399 /// `pred`, limited to returning at most `n` items. The matched element is
2400 /// not contained in the subslices.
2401 ///
2402 /// The last element returned, if any, will contain the remainder of the
2403 /// slice.
2404 ///
2405 /// # Examples
2406 ///
2407 /// Print the slice split once by numbers divisible by 3 (i.e., `[10, 40]`,
2408 /// `[20, 60, 50]`):
2409 ///
2410 /// ```
2411 /// let v = [10, 40, 30, 20, 60, 50];
2412 ///
2413 /// for group in v.splitn(2, |num| *num % 3 == 0) {
2414 /// println!("{group:?}");
2415 /// }
2416 /// ```
2417 #[stable(feature = "rust1", since = "1.0.0")]
2418 #[inline]
2419 pub fn splitn<F>(&self, n: usize, pred: F) -> SplitN<'_, T, F>
2420 where
2421 F: FnMut(&T) -> bool,
2422 {
2423 SplitN::new(self.split(pred), n)
2424 }
2425
2426 /// Returns an iterator over mutable subslices separated by elements that match
2427 /// `pred`, limited to returning at most `n` items. The matched element is
2428 /// not contained in the subslices.
2429 ///
2430 /// The last element returned, if any, will contain the remainder of the
2431 /// slice.
2432 ///
2433 /// # Examples
2434 ///
2435 /// ```
2436 /// let mut v = [10, 40, 30, 20, 60, 50];
2437 ///
2438 /// for group in v.splitn_mut(2, |num| *num % 3 == 0) {
2439 /// group[0] = 1;
2440 /// }
2441 /// assert_eq!(v, [1, 40, 30, 1, 60, 50]);
2442 /// ```
2443 #[stable(feature = "rust1", since = "1.0.0")]
2444 #[inline]
2445 pub fn splitn_mut<F>(&mut self, n: usize, pred: F) -> SplitNMut<'_, T, F>
2446 where
2447 F: FnMut(&T) -> bool,
2448 {
2449 SplitNMut::new(self.split_mut(pred), n)
2450 }
2451
2452 /// Returns an iterator over subslices separated by elements that match
2453 /// `pred` limited to returning at most `n` items. This starts at the end of
2454 /// the slice and works backwards. The matched element is not contained in
2455 /// the subslices.
2456 ///
2457 /// The last element returned, if any, will contain the remainder of the
2458 /// slice.
2459 ///
2460 /// # Examples
2461 ///
2462 /// Print the slice split once, starting from the end, by numbers divisible
2463 /// by 3 (i.e., `[50]`, `[10, 40, 30, 20]`):
2464 ///
2465 /// ```
2466 /// let v = [10, 40, 30, 20, 60, 50];
2467 ///
2468 /// for group in v.rsplitn(2, |num| *num % 3 == 0) {
2469 /// println!("{group:?}");
2470 /// }
2471 /// ```
2472 #[stable(feature = "rust1", since = "1.0.0")]
2473 #[inline]
2474 pub fn rsplitn<F>(&self, n: usize, pred: F) -> RSplitN<'_, T, F>
2475 where
2476 F: FnMut(&T) -> bool,
2477 {
2478 RSplitN::new(self.rsplit(pred), n)
2479 }
2480
2481 /// Returns an iterator over subslices separated by elements that match
2482 /// `pred` limited to returning at most `n` items. This starts at the end of
2483 /// the slice and works backwards. The matched element is not contained in
2484 /// the subslices.
2485 ///
2486 /// The last element returned, if any, will contain the remainder of the
2487 /// slice.
2488 ///
2489 /// # Examples
2490 ///
2491 /// ```
2492 /// let mut s = [10, 40, 30, 20, 60, 50];
2493 ///
2494 /// for group in s.rsplitn_mut(2, |num| *num % 3 == 0) {
2495 /// group[0] = 1;
2496 /// }
2497 /// assert_eq!(s, [1, 40, 30, 20, 60, 1]);
2498 /// ```
2499 #[stable(feature = "rust1", since = "1.0.0")]
2500 #[inline]
2501 pub fn rsplitn_mut<F>(&mut self, n: usize, pred: F) -> RSplitNMut<'_, T, F>
2502 where
2503 F: FnMut(&T) -> bool,
2504 {
2505 RSplitNMut::new(self.rsplit_mut(pred), n)
2506 }
2507
2508 /// Splits the slice on the first element that matches the specified
2509 /// predicate.
2510 ///
2511 /// If any matching elements are present in the slice, returns the prefix
2512 /// before the match and suffix after. The matching element itself is not
2513 /// included. If no elements match, returns `None`.
2514 ///
2515 /// # Examples
2516 ///
2517 /// ```
2518 /// #![feature(slice_split_once)]
2519 /// let s = [1, 2, 3, 2, 4];
2520 /// assert_eq!(s.split_once(|&x| x == 2), Some((
2521 /// &[1][..],
2522 /// &[3, 2, 4][..]
2523 /// )));
2524 /// assert_eq!(s.split_once(|&x| x == 0), None);
2525 /// ```
2526 #[unstable(feature = "slice_split_once", issue = "112811")]
2527 #[inline]
2528 pub fn split_once<F>(&self, pred: F) -> Option<(&[T], &[T])>
2529 where
2530 F: FnMut(&T) -> bool,
2531 {
2532 let index = self.iter().position(pred)?;
2533 Some((&self[..index], &self[index + 1..]))
2534 }
2535
2536 /// Splits the slice on the last element that matches the specified
2537 /// predicate.
2538 ///
2539 /// If any matching elements are present in the slice, returns the prefix
2540 /// before the match and suffix after. The matching element itself is not
2541 /// included. If no elements match, returns `None`.
2542 ///
2543 /// # Examples
2544 ///
2545 /// ```
2546 /// #![feature(slice_split_once)]
2547 /// let s = [1, 2, 3, 2, 4];
2548 /// assert_eq!(s.rsplit_once(|&x| x == 2), Some((
2549 /// &[1, 2, 3][..],
2550 /// &[4][..]
2551 /// )));
2552 /// assert_eq!(s.rsplit_once(|&x| x == 0), None);
2553 /// ```
2554 #[unstable(feature = "slice_split_once", issue = "112811")]
2555 #[inline]
2556 pub fn rsplit_once<F>(&self, pred: F) -> Option<(&[T], &[T])>
2557 where
2558 F: FnMut(&T) -> bool,
2559 {
2560 let index = self.iter().rposition(pred)?;
2561 Some((&self[..index], &self[index + 1..]))
2562 }
2563
2564 /// Returns `true` if the slice contains an element with the given value.
2565 ///
2566 /// This operation is *O*(*n*).
2567 ///
2568 /// Note that if you have a sorted slice, [`binary_search`] may be faster.
2569 ///
2570 /// [`binary_search`]: slice::binary_search
2571 ///
2572 /// # Examples
2573 ///
2574 /// ```
2575 /// let v = [10, 40, 30];
2576 /// assert!(v.contains(&30));
2577 /// assert!(!v.contains(&50));
2578 /// ```
2579 ///
2580 /// If you do not have a `&T`, but some other value that you can compare
2581 /// with one (for example, `String` implements `PartialEq<str>`), you can
2582 /// use `iter().any`:
2583 ///
2584 /// ```
2585 /// let v = [String::from("hello"), String::from("world")]; // slice of `String`
2586 /// assert!(v.iter().any(|e| e == "hello")); // search with `&str`
2587 /// assert!(!v.iter().any(|e| e == "hi"));
2588 /// ```
2589 #[stable(feature = "rust1", since = "1.0.0")]
2590 #[inline]
2591 #[must_use]
2592 pub fn contains(&self, x: &T) -> bool
2593 where
2594 T: PartialEq,
2595 {
2596 cmp::SliceContains::slice_contains(x, self)
2597 }
2598
2599 /// Returns `true` if `needle` is a prefix of the slice or equal to the slice.
2600 ///
2601 /// # Examples
2602 ///
2603 /// ```
2604 /// let v = [10, 40, 30];
2605 /// assert!(v.starts_with(&[10]));
2606 /// assert!(v.starts_with(&[10, 40]));
2607 /// assert!(v.starts_with(&v));
2608 /// assert!(!v.starts_with(&[50]));
2609 /// assert!(!v.starts_with(&[10, 50]));
2610 /// ```
2611 ///
2612 /// Always returns `true` if `needle` is an empty slice:
2613 ///
2614 /// ```
2615 /// let v = &[10, 40, 30];
2616 /// assert!(v.starts_with(&[]));
2617 /// let v: &[u8] = &[];
2618 /// assert!(v.starts_with(&[]));
2619 /// ```
2620 #[stable(feature = "rust1", since = "1.0.0")]
2621 #[must_use]
2622 pub fn starts_with(&self, needle: &[T]) -> bool
2623 where
2624 T: PartialEq,
2625 {
2626 let n = needle.len();
2627 self.len() >= n && needle == &self[..n]
2628 }
2629
2630 /// Returns `true` if `needle` is a suffix of the slice or equal to the slice.
2631 ///
2632 /// # Examples
2633 ///
2634 /// ```
2635 /// let v = [10, 40, 30];
2636 /// assert!(v.ends_with(&[30]));
2637 /// assert!(v.ends_with(&[40, 30]));
2638 /// assert!(v.ends_with(&v));
2639 /// assert!(!v.ends_with(&[50]));
2640 /// assert!(!v.ends_with(&[50, 30]));
2641 /// ```
2642 ///
2643 /// Always returns `true` if `needle` is an empty slice:
2644 ///
2645 /// ```
2646 /// let v = &[10, 40, 30];
2647 /// assert!(v.ends_with(&[]));
2648 /// let v: &[u8] = &[];
2649 /// assert!(v.ends_with(&[]));
2650 /// ```
2651 #[stable(feature = "rust1", since = "1.0.0")]
2652 #[must_use]
2653 pub fn ends_with(&self, needle: &[T]) -> bool
2654 where
2655 T: PartialEq,
2656 {
2657 let (m, n) = (self.len(), needle.len());
2658 m >= n && needle == &self[m - n..]
2659 }
2660
2661 /// Returns a subslice with the prefix removed.
2662 ///
2663 /// If the slice starts with `prefix`, returns the subslice after the prefix, wrapped in `Some`.
2664 /// If `prefix` is empty, simply returns the original slice. If `prefix` is equal to the
2665 /// original slice, returns an empty slice.
2666 ///
2667 /// If the slice does not start with `prefix`, returns `None`.
2668 ///
2669 /// # Examples
2670 ///
2671 /// ```
2672 /// let v = &[10, 40, 30];
2673 /// assert_eq!(v.strip_prefix(&[10]), Some(&[40, 30][..]));
2674 /// assert_eq!(v.strip_prefix(&[10, 40]), Some(&[30][..]));
2675 /// assert_eq!(v.strip_prefix(&[10, 40, 30]), Some(&[][..]));
2676 /// assert_eq!(v.strip_prefix(&[50]), None);
2677 /// assert_eq!(v.strip_prefix(&[10, 50]), None);
2678 ///
2679 /// let prefix : &str = "he";
2680 /// assert_eq!(b"hello".strip_prefix(prefix.as_bytes()),
2681 /// Some(b"llo".as_ref()));
2682 /// ```
2683 #[must_use = "returns the subslice without modifying the original"]
2684 #[stable(feature = "slice_strip", since = "1.51.0")]
2685 pub fn strip_prefix<P: SlicePattern<Item = T> + ?Sized>(&self, prefix: &P) -> Option<&[T]>
2686 where
2687 T: PartialEq,
2688 {
2689 // This function will need rewriting if and when SlicePattern becomes more sophisticated.
2690 let prefix = prefix.as_slice();
2691 let n = prefix.len();
2692 if n <= self.len() {
2693 let (head, tail) = self.split_at(n);
2694 if head == prefix {
2695 return Some(tail);
2696 }
2697 }
2698 None
2699 }
2700
2701 /// Returns a subslice with the suffix removed.
2702 ///
2703 /// If the slice ends with `suffix`, returns the subslice before the suffix, wrapped in `Some`.
2704 /// If `suffix` is empty, simply returns the original slice. If `suffix` is equal to the
2705 /// original slice, returns an empty slice.
2706 ///
2707 /// If the slice does not end with `suffix`, returns `None`.
2708 ///
2709 /// # Examples
2710 ///
2711 /// ```
2712 /// let v = &[10, 40, 30];
2713 /// assert_eq!(v.strip_suffix(&[30]), Some(&[10, 40][..]));
2714 /// assert_eq!(v.strip_suffix(&[40, 30]), Some(&[10][..]));
2715 /// assert_eq!(v.strip_suffix(&[10, 40, 30]), Some(&[][..]));
2716 /// assert_eq!(v.strip_suffix(&[50]), None);
2717 /// assert_eq!(v.strip_suffix(&[50, 30]), None);
2718 /// ```
2719 #[must_use = "returns the subslice without modifying the original"]
2720 #[stable(feature = "slice_strip", since = "1.51.0")]
2721 pub fn strip_suffix<P: SlicePattern<Item = T> + ?Sized>(&self, suffix: &P) -> Option<&[T]>
2722 where
2723 T: PartialEq,
2724 {
2725 // This function will need rewriting if and when SlicePattern becomes more sophisticated.
2726 let suffix = suffix.as_slice();
2727 let (len, n) = (self.len(), suffix.len());
2728 if n <= len {
2729 let (head, tail) = self.split_at(len - n);
2730 if tail == suffix {
2731 return Some(head);
2732 }
2733 }
2734 None
2735 }
2736
2737 /// Returns a subslice with the prefix and suffix removed.
2738 ///
2739 /// If the slice starts with `prefix` and ends with `suffix`, returns the subslice after the
2740 /// prefix and before the suffix, wrapped in `Some`.
2741 ///
2742 /// If the slice does not start with `prefix` or does not end with `suffix`, returns `None`.
2743 ///
2744 /// # Examples
2745 ///
2746 /// ```
2747 /// let v = &[10, 50, 40, 30];
2748 /// assert_eq!(v.strip_circumfix(&[10], &[30]), Some(&[50, 40][..]));
2749 /// assert_eq!(v.strip_circumfix(&[10], &[40, 30]), Some(&[50][..]));
2750 /// assert_eq!(v.strip_circumfix(&[10, 50], &[40, 30]), Some(&[][..]));
2751 /// assert_eq!(v.strip_circumfix(&[50], &[30]), None);
2752 /// assert_eq!(v.strip_circumfix(&[10], &[40]), None);
2753 /// assert_eq!(v.strip_circumfix(&[], &[40, 30]), Some(&[10, 50][..]));
2754 /// assert_eq!(v.strip_circumfix(&[10, 50], &[]), Some(&[40, 30][..]));
2755 /// ```
2756 #[must_use = "returns the subslice without modifying the original"]
2757 #[stable(feature = "strip_circumfix", since = "CURRENT_RUSTC_VERSION")]
2758 pub fn strip_circumfix<S, P>(&self, prefix: &P, suffix: &S) -> Option<&[T]>
2759 where
2760 T: PartialEq,
2761 S: SlicePattern<Item = T> + ?Sized,
2762 P: SlicePattern<Item = T> + ?Sized,
2763 {
2764 self.strip_prefix(prefix)?.strip_suffix(suffix)
2765 }
2766
2767 /// Returns a subslice with the optional prefix removed.
2768 ///
2769 /// If the slice starts with `prefix`, returns the subslice after the prefix. If `prefix`
2770 /// is empty or the slice does not start with `prefix`, simply returns the original slice.
2771 /// If `prefix` is equal to the original slice, returns an empty slice.
2772 ///
2773 /// # Examples
2774 ///
2775 /// ```
2776 /// #![feature(trim_prefix_suffix)]
2777 ///
2778 /// let v = &[10, 40, 30];
2779 ///
2780 /// // Prefix present - removes it
2781 /// assert_eq!(v.trim_prefix(&[10]), &[40, 30][..]);
2782 /// assert_eq!(v.trim_prefix(&[10, 40]), &[30][..]);
2783 /// assert_eq!(v.trim_prefix(&[10, 40, 30]), &[][..]);
2784 ///
2785 /// // Prefix absent - returns original slice
2786 /// assert_eq!(v.trim_prefix(&[50]), &[10, 40, 30][..]);
2787 /// assert_eq!(v.trim_prefix(&[10, 50]), &[10, 40, 30][..]);
2788 ///
2789 /// let prefix : &str = "he";
2790 /// assert_eq!(b"hello".trim_prefix(prefix.as_bytes()), b"llo".as_ref());
2791 /// ```
2792 #[must_use = "returns the subslice without modifying the original"]
2793 #[unstable(feature = "trim_prefix_suffix", issue = "142312")]
2794 pub fn trim_prefix<P: SlicePattern<Item = T> + ?Sized>(&self, prefix: &P) -> &[T]
2795 where
2796 T: PartialEq,
2797 {
2798 // This function will need rewriting if and when SlicePattern becomes more sophisticated.
2799 let prefix = prefix.as_slice();
2800 let n = prefix.len();
2801 if n <= self.len() {
2802 let (head, tail) = self.split_at(n);
2803 if head == prefix {
2804 return tail;
2805 }
2806 }
2807 self
2808 }
2809
2810 /// Returns a subslice with the optional suffix removed.
2811 ///
2812 /// If the slice ends with `suffix`, returns the subslice before the suffix. If `suffix`
2813 /// is empty or the slice does not end with `suffix`, simply returns the original slice.
2814 /// If `suffix` is equal to the original slice, returns an empty slice.
2815 ///
2816 /// # Examples
2817 ///
2818 /// ```
2819 /// #![feature(trim_prefix_suffix)]
2820 ///
2821 /// let v = &[10, 40, 30];
2822 ///
2823 /// // Suffix present - removes it
2824 /// assert_eq!(v.trim_suffix(&[30]), &[10, 40][..]);
2825 /// assert_eq!(v.trim_suffix(&[40, 30]), &[10][..]);
2826 /// assert_eq!(v.trim_suffix(&[10, 40, 30]), &[][..]);
2827 ///
2828 /// // Suffix absent - returns original slice
2829 /// assert_eq!(v.trim_suffix(&[50]), &[10, 40, 30][..]);
2830 /// assert_eq!(v.trim_suffix(&[50, 30]), &[10, 40, 30][..]);
2831 /// ```
2832 #[must_use = "returns the subslice without modifying the original"]
2833 #[unstable(feature = "trim_prefix_suffix", issue = "142312")]
2834 pub fn trim_suffix<P: SlicePattern<Item = T> + ?Sized>(&self, suffix: &P) -> &[T]
2835 where
2836 T: PartialEq,
2837 {
2838 // This function will need rewriting if and when SlicePattern becomes more sophisticated.
2839 let suffix = suffix.as_slice();
2840 let (len, n) = (self.len(), suffix.len());
2841 if n <= len {
2842 let (head, tail) = self.split_at(len - n);
2843 if tail == suffix {
2844 return head;
2845 }
2846 }
2847 self
2848 }
2849
2850 /// Binary searches this slice for a given element.
2851 /// If the slice is not sorted, the returned result is unspecified and
2852 /// meaningless.
2853 ///
2854 /// If the value is found then [`Result::Ok`] is returned, containing the
2855 /// index of the matching element. If there are multiple matches, then any
2856 /// one of the matches could be returned. The index is chosen
2857 /// deterministically, but is subject to change in future versions of Rust.
2858 /// If the value is not found then [`Result::Err`] is returned, containing
2859 /// the index where a matching element could be inserted while maintaining
2860 /// sorted order.
2861 ///
2862 /// See also [`binary_search_by`], [`binary_search_by_key`], and [`partition_point`].
2863 ///
2864 /// [`binary_search_by`]: slice::binary_search_by
2865 /// [`binary_search_by_key`]: slice::binary_search_by_key
2866 /// [`partition_point`]: slice::partition_point
2867 ///
2868 /// # Examples
2869 ///
2870 /// Looks up a series of four elements. The first is found, with a
2871 /// uniquely determined position; the second and third are not
2872 /// found; the fourth could match any position in `[1, 4]`.
2873 ///
2874 /// ```
2875 /// let s = [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
2876 ///
2877 /// assert_eq!(s.binary_search(&13), Ok(9));
2878 /// assert_eq!(s.binary_search(&4), Err(7));
2879 /// assert_eq!(s.binary_search(&100), Err(13));
2880 /// let r = s.binary_search(&1);
2881 /// assert!(match r { Ok(1..=4) => true, _ => false, });
2882 /// ```
2883 ///
2884 /// If you want to find that whole *range* of matching items, rather than
2885 /// an arbitrary matching one, that can be done using [`partition_point`]:
2886 /// ```
2887 /// let s = [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
2888 ///
2889 /// let low = s.partition_point(|x| x < &1);
2890 /// assert_eq!(low, 1);
2891 /// let high = s.partition_point(|x| x <= &1);
2892 /// assert_eq!(high, 5);
2893 /// let r = s.binary_search(&1);
2894 /// assert!((low..high).contains(&r.unwrap()));
2895 ///
2896 /// assert!(s[..low].iter().all(|&x| x < 1));
2897 /// assert!(s[low..high].iter().all(|&x| x == 1));
2898 /// assert!(s[high..].iter().all(|&x| x > 1));
2899 ///
2900 /// // For something not found, the "range" of equal items is empty
2901 /// assert_eq!(s.partition_point(|x| x < &11), 9);
2902 /// assert_eq!(s.partition_point(|x| x <= &11), 9);
2903 /// assert_eq!(s.binary_search(&11), Err(9));
2904 /// ```
2905 ///
2906 /// If you want to insert an item to a sorted vector, while maintaining
2907 /// sort order, consider using [`partition_point`]:
2908 ///
2909 /// ```
2910 /// let mut s = vec![0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
2911 /// let num = 42;
2912 /// let idx = s.partition_point(|&x| x <= num);
2913 /// // If `num` is unique, `s.partition_point(|&x| x < num)` (with `<`) is equivalent to
2914 /// // `s.binary_search(&num).unwrap_or_else(|x| x)`, but using `<=` will allow `insert`
2915 /// // to shift less elements.
2916 /// s.insert(idx, num);
2917 /// assert_eq!(s, [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 42, 55]);
2918 /// ```
2919 #[stable(feature = "rust1", since = "1.0.0")]
2920 pub fn binary_search(&self, x: &T) -> Result<usize, usize>
2921 where
2922 T: Ord,
2923 {
2924 self.binary_search_by(|p| p.cmp(x))
2925 }
2926
2927 /// Binary searches this slice with a comparator function.
2928 ///
2929 /// The comparator function should return an order code that indicates
2930 /// whether its argument is `Less`, `Equal` or `Greater` the desired
2931 /// target.
2932 /// If the slice is not sorted or if the comparator function does not
2933 /// implement an order consistent with the sort order of the underlying
2934 /// slice, the returned result is unspecified and meaningless.
2935 ///
2936 /// If the value is found then [`Result::Ok`] is returned, containing the
2937 /// index of the matching element. If there are multiple matches, then any
2938 /// one of the matches could be returned. The index is chosen
2939 /// deterministically, but is subject to change in future versions of Rust.
2940 /// If the value is not found then [`Result::Err`] is returned, containing
2941 /// the index where a matching element could be inserted while maintaining
2942 /// sorted order.
2943 ///
2944 /// See also [`binary_search`], [`binary_search_by_key`], and [`partition_point`].
2945 ///
2946 /// [`binary_search`]: slice::binary_search
2947 /// [`binary_search_by_key`]: slice::binary_search_by_key
2948 /// [`partition_point`]: slice::partition_point
2949 ///
2950 /// # Examples
2951 ///
2952 /// Looks up a series of four elements. The first is found, with a
2953 /// uniquely determined position; the second and third are not
2954 /// found; the fourth could match any position in `[1, 4]`.
2955 ///
2956 /// ```
2957 /// let s = [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
2958 ///
2959 /// let seek = 13;
2960 /// assert_eq!(s.binary_search_by(|probe| probe.cmp(&seek)), Ok(9));
2961 /// let seek = 4;
2962 /// assert_eq!(s.binary_search_by(|probe| probe.cmp(&seek)), Err(7));
2963 /// let seek = 100;
2964 /// assert_eq!(s.binary_search_by(|probe| probe.cmp(&seek)), Err(13));
2965 /// let seek = 1;
2966 /// let r = s.binary_search_by(|probe| probe.cmp(&seek));
2967 /// assert!(match r { Ok(1..=4) => true, _ => false, });
2968 /// ```
2969 #[stable(feature = "rust1", since = "1.0.0")]
2970 #[inline]
2971 pub fn binary_search_by<'a, F>(&'a self, mut f: F) -> Result<usize, usize>
2972 where
2973 F: FnMut(&'a T) -> Ordering,
2974 {
2975 let mut size = self.len();
2976 if size == 0 {
2977 return Err(0);
2978 }
2979 let mut base = 0usize;
2980
2981 // This loop intentionally doesn't have an early exit if the comparison
2982 // returns Equal. We want the number of loop iterations to depend *only*
2983 // on the size of the input slice so that the CPU can reliably predict
2984 // the loop count.
2985 while size > 1 {
2986 let half = size / 2;
2987 let mid = base + half;
2988
2989 // SAFETY: the call is made safe by the following invariants:
2990 // - `mid >= 0`: by definition
2991 // - `mid < size`: `mid = size / 2 + size / 4 + size / 8 ...`
2992 let cmp = f(unsafe { self.get_unchecked(mid) });
2993
2994 // Binary search interacts poorly with branch prediction, so force
2995 // the compiler to use conditional moves if supported by the target
2996 // architecture.
2997 base = hint::select_unpredictable(cmp == Greater, base, mid);
2998
2999 // This is imprecise in the case where `size` is odd and the
3000 // comparison returns Greater: the mid element still gets included
3001 // by `size` even though it's known to be larger than the element
3002 // being searched for.
3003 //
3004 // This is fine though: we gain more performance by keeping the
3005 // loop iteration count invariant (and thus predictable) than we
3006 // lose from considering one additional element.
3007 size -= half;
3008 }
3009
3010 // SAFETY: base is always in [0, size) because base <= mid.
3011 let cmp = f(unsafe { self.get_unchecked(base) });
3012 if cmp == Equal {
3013 // SAFETY: same as the `get_unchecked` above.
3014 unsafe { hint::assert_unchecked(base < self.len()) };
3015 Ok(base)
3016 } else {
3017 let result = base + (cmp == Less) as usize;
3018 // SAFETY: same as the `get_unchecked` above.
3019 // Note that this is `<=`, unlike the assume in the `Ok` path.
3020 unsafe { hint::assert_unchecked(result <= self.len()) };
3021 Err(result)
3022 }
3023 }
3024
3025 /// Binary searches this slice with a key extraction function.
3026 ///
3027 /// Assumes that the slice is sorted by the key, for instance with
3028 /// [`sort_by_key`] using the same key extraction function.
3029 /// If the slice is not sorted by the key, the returned result is
3030 /// unspecified and meaningless.
3031 ///
3032 /// If the value is found then [`Result::Ok`] is returned, containing the
3033 /// index of the matching element. If there are multiple matches, then any
3034 /// one of the matches could be returned. The index is chosen
3035 /// deterministically, but is subject to change in future versions of Rust.
3036 /// If the value is not found then [`Result::Err`] is returned, containing
3037 /// the index where a matching element could be inserted while maintaining
3038 /// sorted order.
3039 ///
3040 /// See also [`binary_search`], [`binary_search_by`], and [`partition_point`].
3041 ///
3042 /// [`sort_by_key`]: slice::sort_by_key
3043 /// [`binary_search`]: slice::binary_search
3044 /// [`binary_search_by`]: slice::binary_search_by
3045 /// [`partition_point`]: slice::partition_point
3046 ///
3047 /// # Examples
3048 ///
3049 /// Looks up a series of four elements in a slice of pairs sorted by
3050 /// their second elements. The first is found, with a uniquely
3051 /// determined position; the second and third are not found; the
3052 /// fourth could match any position in `[1, 4]`.
3053 ///
3054 /// ```
3055 /// let s = [(0, 0), (2, 1), (4, 1), (5, 1), (3, 1),
3056 /// (1, 2), (2, 3), (4, 5), (5, 8), (3, 13),
3057 /// (1, 21), (2, 34), (4, 55)];
3058 ///
3059 /// assert_eq!(s.binary_search_by_key(&13, |&(a, b)| b), Ok(9));
3060 /// assert_eq!(s.binary_search_by_key(&4, |&(a, b)| b), Err(7));
3061 /// assert_eq!(s.binary_search_by_key(&100, |&(a, b)| b), Err(13));
3062 /// let r = s.binary_search_by_key(&1, |&(a, b)| b);
3063 /// assert!(match r { Ok(1..=4) => true, _ => false, });
3064 /// ```
3065 // Lint rustdoc::broken_intra_doc_links is allowed as `slice::sort_by_key` is
3066 // in crate `alloc`, and as such doesn't exists yet when building `core`: #74481.
3067 // This breaks links when slice is displayed in core, but changing it to use relative links
3068 // would break when the item is re-exported. So allow the core links to be broken for now.
3069 #[allow(rustdoc::broken_intra_doc_links)]
3070 #[stable(feature = "slice_binary_search_by_key", since = "1.10.0")]
3071 #[inline]
3072 pub fn binary_search_by_key<'a, B, F>(&'a self, b: &B, mut f: F) -> Result<usize, usize>
3073 where
3074 F: FnMut(&'a T) -> B,
3075 B: Ord,
3076 {
3077 self.binary_search_by(|k| f(k).cmp(b))
3078 }
3079
3080 /// Sorts the slice in ascending order **without** preserving the initial order of equal elements.
3081 ///
3082 /// This sort is unstable (i.e., may reorder equal elements), in-place (i.e., does not
3083 /// allocate), and *O*(*n* \* log(*n*)) worst-case.
3084 ///
3085 /// If the implementation of [`Ord`] for `T` does not implement a [total order], the function
3086 /// may panic; even if the function exits normally, the resulting order of elements in the slice
3087 /// is unspecified. See also the note on panicking below.
3088 ///
3089 /// For example `|a, b| (a - b).cmp(a)` is a comparison function that is neither transitive nor
3090 /// reflexive nor total, `a < b < c < a` with `a = 1, b = 2, c = 3`. For more information and
3091 /// examples see the [`Ord`] documentation.
3092 ///
3093 ///
3094 /// All original elements will remain in the slice and any possible modifications via interior
3095 /// mutability are observed in the input. Same is true if the implementation of [`Ord`] for `T` panics.
3096 ///
3097 /// Sorting types that only implement [`PartialOrd`] such as [`f32`] and [`f64`] require
3098 /// additional precautions. For example, `f32::NAN != f32::NAN`, which doesn't fulfill the
3099 /// reflexivity requirement of [`Ord`]. By using an alternative comparison function with
3100 /// `slice::sort_unstable_by` such as [`f32::total_cmp`] or [`f64::total_cmp`] that defines a
3101 /// [total order] users can sort slices containing floating-point values. Alternatively, if all
3102 /// values in the slice are guaranteed to be in a subset for which [`PartialOrd::partial_cmp`]
3103 /// forms a [total order], it's possible to sort the slice with `sort_unstable_by(|a, b|
3104 /// a.partial_cmp(b).unwrap())`.
3105 ///
3106 /// # Current implementation
3107 ///
3108 /// The current implementation is based on [ipnsort] by Lukas Bergdoll and Orson Peters, which
3109 /// combines the fast average case of quicksort with the fast worst case of heapsort, achieving
3110 /// linear time on fully sorted and reversed inputs. On inputs with k distinct elements, the
3111 /// expected time to sort the data is *O*(*n* \* log(*k*)).
3112 ///
3113 /// It is typically faster than stable sorting, except in a few special cases, e.g., when the
3114 /// slice is partially sorted.
3115 ///
3116 /// # Panics
3117 ///
3118 /// May panic if the implementation of [`Ord`] for `T` does not implement a [total order], or if
3119 /// the [`Ord`] implementation panics.
3120 ///
3121 /// # Examples
3122 ///
3123 /// ```
3124 /// let mut v = [4, -5, 1, -3, 2];
3125 ///
3126 /// v.sort_unstable();
3127 /// assert_eq!(v, [-5, -3, 1, 2, 4]);
3128 /// ```
3129 ///
3130 /// [ipnsort]: https://github.com/Voultapher/sort-research-rs/tree/main/ipnsort
3131 /// [total order]: https://en.wikipedia.org/wiki/Total_order
3132 #[stable(feature = "sort_unstable", since = "1.20.0")]
3133 #[inline]
3134 pub fn sort_unstable(&mut self)
3135 where
3136 T: Ord,
3137 {
3138 sort::unstable::sort(self, &mut T::lt);
3139 }
3140
3141 /// Sorts the slice in ascending order with a comparison function, **without** preserving the
3142 /// initial order of equal elements.
3143 ///
3144 /// This sort is unstable (i.e., may reorder equal elements), in-place (i.e., does not
3145 /// allocate), and *O*(*n* \* log(*n*)) worst-case.
3146 ///
3147 /// If the comparison function `compare` does not implement a [total order], the function
3148 /// may panic; even if the function exits normally, the resulting order of elements in the slice
3149 /// is unspecified. See also the note on panicking below.
3150 ///
3151 /// For example `|a, b| (a - b).cmp(a)` is a comparison function that is neither transitive nor
3152 /// reflexive nor total, `a < b < c < a` with `a = 1, b = 2, c = 3`. For more information and
3153 /// examples see the [`Ord`] documentation.
3154 ///
3155 /// All original elements will remain in the slice and any possible modifications via interior
3156 /// mutability are observed in the input. Same is true if `compare` panics.
3157 ///
3158 /// # Current implementation
3159 ///
3160 /// The current implementation is based on [ipnsort] by Lukas Bergdoll and Orson Peters, which
3161 /// combines the fast average case of quicksort with the fast worst case of heapsort, achieving
3162 /// linear time on fully sorted and reversed inputs. On inputs with k distinct elements, the
3163 /// expected time to sort the data is *O*(*n* \* log(*k*)).
3164 ///
3165 /// It is typically faster than stable sorting, except in a few special cases, e.g., when the
3166 /// slice is partially sorted.
3167 ///
3168 /// # Panics
3169 ///
3170 /// May panic if the `compare` does not implement a [total order], or if
3171 /// the `compare` itself panics.
3172 ///
3173 /// # Examples
3174 ///
3175 /// ```
3176 /// let mut v = [4, -5, 1, -3, 2];
3177 /// v.sort_unstable_by(|a, b| a.cmp(b));
3178 /// assert_eq!(v, [-5, -3, 1, 2, 4]);
3179 ///
3180 /// // reverse sorting
3181 /// v.sort_unstable_by(|a, b| b.cmp(a));
3182 /// assert_eq!(v, [4, 2, 1, -3, -5]);
3183 /// ```
3184 ///
3185 /// [ipnsort]: https://github.com/Voultapher/sort-research-rs/tree/main/ipnsort
3186 /// [total order]: https://en.wikipedia.org/wiki/Total_order
3187 #[stable(feature = "sort_unstable", since = "1.20.0")]
3188 #[inline]
3189 pub fn sort_unstable_by<F>(&mut self, mut compare: F)
3190 where
3191 F: FnMut(&T, &T) -> Ordering,
3192 {
3193 sort::unstable::sort(self, &mut |a, b| compare(a, b) == Ordering::Less);
3194 }
3195
3196 /// Sorts the slice in ascending order with a key extraction function, **without** preserving
3197 /// the initial order of equal elements.
3198 ///
3199 /// This sort is unstable (i.e., may reorder equal elements), in-place (i.e., does not
3200 /// allocate), and *O*(*n* \* log(*n*)) worst-case.
3201 ///
3202 /// If the implementation of [`Ord`] for `K` does not implement a [total order], the function
3203 /// may panic; even if the function exits normally, the resulting order of elements in the slice
3204 /// is unspecified. See also the note on panicking below.
3205 ///
3206 /// For example `|a, b| (a - b).cmp(a)` is a comparison function that is neither transitive nor
3207 /// reflexive nor total, `a < b < c < a` with `a = 1, b = 2, c = 3`. For more information and
3208 /// examples see the [`Ord`] documentation.
3209 ///
3210 /// All original elements will remain in the slice and any possible modifications via interior
3211 /// mutability are observed in the input. Same is true if the implementation of [`Ord`] for `K` panics.
3212 ///
3213 /// # Current implementation
3214 ///
3215 /// The current implementation is based on [ipnsort] by Lukas Bergdoll and Orson Peters, which
3216 /// combines the fast average case of quicksort with the fast worst case of heapsort, achieving
3217 /// linear time on fully sorted and reversed inputs. On inputs with k distinct elements, the
3218 /// expected time to sort the data is *O*(*n* \* log(*k*)).
3219 ///
3220 /// It is typically faster than stable sorting, except in a few special cases, e.g., when the
3221 /// slice is partially sorted.
3222 ///
3223 /// # Panics
3224 ///
3225 /// May panic if the implementation of [`Ord`] for `K` does not implement a [total order], or if
3226 /// the [`Ord`] implementation panics.
3227 ///
3228 /// # Examples
3229 ///
3230 /// ```
3231 /// let mut v = [4i32, -5, 1, -3, 2];
3232 ///
3233 /// v.sort_unstable_by_key(|k| k.abs());
3234 /// assert_eq!(v, [1, 2, -3, 4, -5]);
3235 /// ```
3236 ///
3237 /// [ipnsort]: https://github.com/Voultapher/sort-research-rs/tree/main/ipnsort
3238 /// [total order]: https://en.wikipedia.org/wiki/Total_order
3239 #[stable(feature = "sort_unstable", since = "1.20.0")]
3240 #[inline]
3241 pub fn sort_unstable_by_key<K, F>(&mut self, mut f: F)
3242 where
3243 F: FnMut(&T) -> K,
3244 K: Ord,
3245 {
3246 sort::unstable::sort(self, &mut |a, b| f(a).lt(&f(b)));
3247 }
3248
3249 /// Partially sorts the slice in ascending order **without** preserving the initial order of equal elements.
3250 ///
3251 /// Upon completion, for the specified range `start..end`, it's guaranteed that:
3252 ///
3253 /// 1. Every element in `self[..start]` is smaller than or equal to
3254 /// 2. Every element in `self[start..end]`, which is sorted, and smaller than or equal to
3255 /// 3. Every element in `self[end..]`.
3256 ///
3257 /// This partial sort is unstable, meaning it may reorder equal elements in the specified range.
3258 /// It may reorder elements outside the specified range as well, but the guarantees above still hold.
3259 ///
3260 /// This partial sort is in-place (i.e., does not allocate), and *O*(*n* + *k* \* log(*k*)) worst-case,
3261 /// where *n* is the length of the slice and *k* is the length of the specified range.
3262 ///
3263 /// See the documentation of [`sort_unstable`] for implementation notes.
3264 ///
3265 /// # Panics
3266 ///
3267 /// May panic if the implementation of [`Ord`] for `T` does not implement a total order, or if
3268 /// the [`Ord`] implementation panics, or if the specified range is out of bounds.
3269 ///
3270 /// # Examples
3271 ///
3272 /// ```
3273 /// #![feature(slice_partial_sort_unstable)]
3274 ///
3275 /// let mut v = [4, -5, 1, -3, 2];
3276 ///
3277 /// // empty range at the beginning, nothing changed
3278 /// v.partial_sort_unstable(0..0);
3279 /// assert_eq!(v, [4, -5, 1, -3, 2]);
3280 ///
3281 /// // empty range in the middle, partitioning the slice
3282 /// v.partial_sort_unstable(2..2);
3283 /// for i in 0..2 {
3284 /// assert!(v[i] <= v[2]);
3285 /// }
3286 /// for i in 3..v.len() {
3287 /// assert!(v[2] <= v[i]);
3288 /// }
3289 ///
3290 /// // single element range, same as select_nth_unstable
3291 /// v.partial_sort_unstable(2..3);
3292 /// for i in 0..2 {
3293 /// assert!(v[i] <= v[2]);
3294 /// }
3295 /// for i in 3..v.len() {
3296 /// assert!(v[2] <= v[i]);
3297 /// }
3298 ///
3299 /// // partial sort a subrange
3300 /// v.partial_sort_unstable(1..4);
3301 /// assert_eq!(&v[1..4], [-3, 1, 2]);
3302 ///
3303 /// // partial sort the whole range, same as sort_unstable
3304 /// v.partial_sort_unstable(..);
3305 /// assert_eq!(v, [-5, -3, 1, 2, 4]);
3306 /// ```
3307 ///
3308 /// [`sort_unstable`]: slice::sort_unstable
3309 #[unstable(feature = "slice_partial_sort_unstable", issue = "149046")]
3310 #[inline]
3311 pub fn partial_sort_unstable<R>(&mut self, range: R)
3312 where
3313 T: Ord,
3314 R: RangeBounds<usize>,
3315 {
3316 sort::unstable::partial_sort(self, range, T::lt);
3317 }
3318
3319 /// Partially sorts the slice in ascending order with a comparison function, **without**
3320 /// preserving the initial order of equal elements.
3321 ///
3322 /// Upon completion, for the specified range `start..end`, it's guaranteed that:
3323 ///
3324 /// 1. Every element in `self[..start]` is smaller than or equal to
3325 /// 2. Every element in `self[start..end]`, which is sorted, and smaller than or equal to
3326 /// 3. Every element in `self[end..]`.
3327 ///
3328 /// This partial sort is unstable, meaning it may reorder equal elements in the specified range.
3329 /// It may reorder elements outside the specified range as well, but the guarantees above still hold.
3330 ///
3331 /// This partial sort is in-place (i.e., does not allocate), and *O*(*n* + *k* \* log(*k*)) worst-case,
3332 /// where *n* is the length of the slice and *k* is the length of the specified range.
3333 ///
3334 /// See the documentation of [`sort_unstable_by`] for implementation notes.
3335 ///
3336 /// # Panics
3337 ///
3338 /// May panic if the `compare` does not implement a total order, or if
3339 /// the `compare` itself panics, or if the specified range is out of bounds.
3340 ///
3341 /// # Examples
3342 ///
3343 /// ```
3344 /// #![feature(slice_partial_sort_unstable)]
3345 ///
3346 /// let mut v = [4, -5, 1, -3, 2];
3347 ///
3348 /// // empty range at the beginning, nothing changed
3349 /// v.partial_sort_unstable_by(0..0, |a, b| b.cmp(a));
3350 /// assert_eq!(v, [4, -5, 1, -3, 2]);
3351 ///
3352 /// // empty range in the middle, partitioning the slice
3353 /// v.partial_sort_unstable_by(2..2, |a, b| b.cmp(a));
3354 /// for i in 0..2 {
3355 /// assert!(v[i] >= v[2]);
3356 /// }
3357 /// for i in 3..v.len() {
3358 /// assert!(v[2] >= v[i]);
3359 /// }
3360 ///
3361 /// // single element range, same as select_nth_unstable
3362 /// v.partial_sort_unstable_by(2..3, |a, b| b.cmp(a));
3363 /// for i in 0..2 {
3364 /// assert!(v[i] >= v[2]);
3365 /// }
3366 /// for i in 3..v.len() {
3367 /// assert!(v[2] >= v[i]);
3368 /// }
3369 ///
3370 /// // partial sort a subrange
3371 /// v.partial_sort_unstable_by(1..4, |a, b| b.cmp(a));
3372 /// assert_eq!(&v[1..4], [2, 1, -3]);
3373 ///
3374 /// // partial sort the whole range, same as sort_unstable
3375 /// v.partial_sort_unstable_by(.., |a, b| b.cmp(a));
3376 /// assert_eq!(v, [4, 2, 1, -3, -5]);
3377 /// ```
3378 ///
3379 /// [`sort_unstable_by`]: slice::sort_unstable_by
3380 #[unstable(feature = "slice_partial_sort_unstable", issue = "149046")]
3381 #[inline]
3382 pub fn partial_sort_unstable_by<F, R>(&mut self, range: R, mut compare: F)
3383 where
3384 F: FnMut(&T, &T) -> Ordering,
3385 R: RangeBounds<usize>,
3386 {
3387 sort::unstable::partial_sort(self, range, |a, b| compare(a, b) == Less);
3388 }
3389
3390 /// Partially sorts the slice in ascending order with a key extraction function, **without**
3391 /// preserving the initial order of equal elements.
3392 ///
3393 /// Upon completion, for the specified range `start..end`, it's guaranteed that:
3394 ///
3395 /// 1. Every element in `self[..start]` is smaller than or equal to
3396 /// 2. Every element in `self[start..end]`, which is sorted, and smaller than or equal to
3397 /// 3. Every element in `self[end..]`.
3398 ///
3399 /// This partial sort is unstable, meaning it may reorder equal elements in the specified range.
3400 /// It may reorder elements outside the specified range as well, but the guarantees above still hold.
3401 ///
3402 /// This partial sort is in-place (i.e., does not allocate), and *O*(*n* + *k* \* log(*k*)) worst-case,
3403 /// where *n* is the length of the slice and *k* is the length of the specified range.
3404 ///
3405 /// See the documentation of [`sort_unstable_by_key`] for implementation notes.
3406 ///
3407 /// # Panics
3408 ///
3409 /// May panic if the implementation of [`Ord`] for `K` does not implement a total order, or if
3410 /// the [`Ord`] implementation panics, or if the specified range is out of bounds.
3411 ///
3412 /// # Examples
3413 ///
3414 /// ```
3415 /// #![feature(slice_partial_sort_unstable)]
3416 ///
3417 /// let mut v = [4i32, -5, 1, -3, 2];
3418 ///
3419 /// // empty range at the beginning, nothing changed
3420 /// v.partial_sort_unstable_by_key(0..0, |k| k.abs());
3421 /// assert_eq!(v, [4, -5, 1, -3, 2]);
3422 ///
3423 /// // empty range in the middle, partitioning the slice
3424 /// v.partial_sort_unstable_by_key(2..2, |k| k.abs());
3425 /// for i in 0..2 {
3426 /// assert!(v[i].abs() <= v[2].abs());
3427 /// }
3428 /// for i in 3..v.len() {
3429 /// assert!(v[2].abs() <= v[i].abs());
3430 /// }
3431 ///
3432 /// // single element range, same as select_nth_unstable
3433 /// v.partial_sort_unstable_by_key(2..3, |k| k.abs());
3434 /// for i in 0..2 {
3435 /// assert!(v[i].abs() <= v[2].abs());
3436 /// }
3437 /// for i in 3..v.len() {
3438 /// assert!(v[2].abs() <= v[i].abs());
3439 /// }
3440 ///
3441 /// // partial sort a subrange
3442 /// v.partial_sort_unstable_by_key(1..4, |k| k.abs());
3443 /// assert_eq!(&v[1..4], [2, -3, 4]);
3444 ///
3445 /// // partial sort the whole range, same as sort_unstable
3446 /// v.partial_sort_unstable_by_key(.., |k| k.abs());
3447 /// assert_eq!(v, [1, 2, -3, 4, -5]);
3448 /// ```
3449 ///
3450 /// [`sort_unstable_by_key`]: slice::sort_unstable_by_key
3451 #[unstable(feature = "slice_partial_sort_unstable", issue = "149046")]
3452 #[inline]
3453 pub fn partial_sort_unstable_by_key<K, F, R>(&mut self, range: R, mut f: F)
3454 where
3455 F: FnMut(&T) -> K,
3456 K: Ord,
3457 R: RangeBounds<usize>,
3458 {
3459 sort::unstable::partial_sort(self, range, |a, b| f(a).lt(&f(b)));
3460 }
3461
3462 /// Reorders the slice such that the element at `index` is at a sort-order position. All
3463 /// elements before `index` will be `<=` to this value, and all elements after will be `>=` to
3464 /// it.
3465 ///
3466 /// This reordering is unstable (i.e. any element that compares equal to the nth element may end
3467 /// up at that position), in-place (i.e. does not allocate), and runs in *O*(*n*) time. This
3468 /// function is also known as "kth element" in other libraries.
3469 ///
3470 /// Returns a triple that partitions the reordered slice:
3471 ///
3472 /// * The unsorted subslice before `index`, whose elements all satisfy `x <= self[index]`.
3473 ///
3474 /// * The element at `index`.
3475 ///
3476 /// * The unsorted subslice after `index`, whose elements all satisfy `x >= self[index]`.
3477 ///
3478 /// # Current implementation
3479 ///
3480 /// The current algorithm is an introselect implementation based on [ipnsort] by Lukas Bergdoll
3481 /// and Orson Peters, which is also the basis for [`sort_unstable`]. The fallback algorithm is
3482 /// Median of Medians using Tukey's Ninther for pivot selection, which guarantees linear runtime
3483 /// for all inputs.
3484 ///
3485 /// [`sort_unstable`]: slice::sort_unstable
3486 ///
3487 /// # Panics
3488 ///
3489 /// Panics when `index >= len()`, and so always panics on empty slices.
3490 ///
3491 /// May panic if the implementation of [`Ord`] for `T` does not implement a [total order].
3492 ///
3493 /// # Examples
3494 ///
3495 /// ```
3496 /// let mut v = [-5i32, 4, 2, -3, 1];
3497 ///
3498 /// // Find the items `<=` to the median, the median itself, and the items `>=` to it.
3499 /// let (lesser, median, greater) = v.select_nth_unstable(2);
3500 ///
3501 /// assert!(lesser == [-3, -5] || lesser == [-5, -3]);
3502 /// assert_eq!(median, &mut 1);
3503 /// assert!(greater == [4, 2] || greater == [2, 4]);
3504 ///
3505 /// // We are only guaranteed the slice will be one of the following, based on the way we sort
3506 /// // about the specified index.
3507 /// assert!(v == [-3, -5, 1, 2, 4] ||
3508 /// v == [-5, -3, 1, 2, 4] ||
3509 /// v == [-3, -5, 1, 4, 2] ||
3510 /// v == [-5, -3, 1, 4, 2]);
3511 /// ```
3512 ///
3513 /// [ipnsort]: https://github.com/Voultapher/sort-research-rs/tree/main/ipnsort
3514 /// [total order]: https://en.wikipedia.org/wiki/Total_order
3515 #[stable(feature = "slice_select_nth_unstable", since = "1.49.0")]
3516 #[inline]
3517 pub fn select_nth_unstable(&mut self, index: usize) -> (&mut [T], &mut T, &mut [T])
3518 where
3519 T: Ord,
3520 {
3521 sort::select::partition_at_index(self, index, T::lt)
3522 }
3523
3524 /// Reorders the slice with a comparator function such that the element at `index` is at a
3525 /// sort-order position. All elements before `index` will be `<=` to this value, and all
3526 /// elements after will be `>=` to it, according to the comparator function.
3527 ///
3528 /// This reordering is unstable (i.e. any element that compares equal to the nth element may end
3529 /// up at that position), in-place (i.e. does not allocate), and runs in *O*(*n*) time. This
3530 /// function is also known as "kth element" in other libraries.
3531 ///
3532 /// Returns a triple partitioning the reordered slice:
3533 ///
3534 /// * The unsorted subslice before `index`, whose elements all satisfy
3535 /// `compare(x, self[index]).is_le()`.
3536 ///
3537 /// * The element at `index`.
3538 ///
3539 /// * The unsorted subslice after `index`, whose elements all satisfy
3540 /// `compare(x, self[index]).is_ge()`.
3541 ///
3542 /// # Current implementation
3543 ///
3544 /// The current algorithm is an introselect implementation based on [ipnsort] by Lukas Bergdoll
3545 /// and Orson Peters, which is also the basis for [`sort_unstable`]. The fallback algorithm is
3546 /// Median of Medians using Tukey's Ninther for pivot selection, which guarantees linear runtime
3547 /// for all inputs.
3548 ///
3549 /// [`sort_unstable`]: slice::sort_unstable
3550 ///
3551 /// # Panics
3552 ///
3553 /// Panics when `index >= len()`, and so always panics on empty slices.
3554 ///
3555 /// May panic if `compare` does not implement a [total order].
3556 ///
3557 /// # Examples
3558 ///
3559 /// ```
3560 /// let mut v = [-5i32, 4, 2, -3, 1];
3561 ///
3562 /// // Find the items `>=` to the median, the median itself, and the items `<=` to it, by using
3563 /// // a reversed comparator.
3564 /// let (before, median, after) = v.select_nth_unstable_by(2, |a, b| b.cmp(a));
3565 ///
3566 /// assert!(before == [4, 2] || before == [2, 4]);
3567 /// assert_eq!(median, &mut 1);
3568 /// assert!(after == [-3, -5] || after == [-5, -3]);
3569 ///
3570 /// // We are only guaranteed the slice will be one of the following, based on the way we sort
3571 /// // about the specified index.
3572 /// assert!(v == [2, 4, 1, -5, -3] ||
3573 /// v == [2, 4, 1, -3, -5] ||
3574 /// v == [4, 2, 1, -5, -3] ||
3575 /// v == [4, 2, 1, -3, -5]);
3576 /// ```
3577 ///
3578 /// [ipnsort]: https://github.com/Voultapher/sort-research-rs/tree/main/ipnsort
3579 /// [total order]: https://en.wikipedia.org/wiki/Total_order
3580 #[stable(feature = "slice_select_nth_unstable", since = "1.49.0")]
3581 #[inline]
3582 pub fn select_nth_unstable_by<F>(
3583 &mut self,
3584 index: usize,
3585 mut compare: F,
3586 ) -> (&mut [T], &mut T, &mut [T])
3587 where
3588 F: FnMut(&T, &T) -> Ordering,
3589 {
3590 sort::select::partition_at_index(self, index, |a: &T, b: &T| compare(a, b) == Less)
3591 }
3592
3593 /// Reorders the slice with a key extraction function such that the element at `index` is at a
3594 /// sort-order position. All elements before `index` will have keys `<=` to the key at `index`,
3595 /// and all elements after will have keys `>=` to it.
3596 ///
3597 /// This reordering is unstable (i.e. any element that compares equal to the nth element may end
3598 /// up at that position), in-place (i.e. does not allocate), and runs in *O*(*n*) time. This
3599 /// function is also known as "kth element" in other libraries.
3600 ///
3601 /// Returns a triple partitioning the reordered slice:
3602 ///
3603 /// * The unsorted subslice before `index`, whose elements all satisfy `f(x) <= f(self[index])`.
3604 ///
3605 /// * The element at `index`.
3606 ///
3607 /// * The unsorted subslice after `index`, whose elements all satisfy `f(x) >= f(self[index])`.
3608 ///
3609 /// # Current implementation
3610 ///
3611 /// The current algorithm is an introselect implementation based on [ipnsort] by Lukas Bergdoll
3612 /// and Orson Peters, which is also the basis for [`sort_unstable`]. The fallback algorithm is
3613 /// Median of Medians using Tukey's Ninther for pivot selection, which guarantees linear runtime
3614 /// for all inputs.
3615 ///
3616 /// [`sort_unstable`]: slice::sort_unstable
3617 ///
3618 /// # Panics
3619 ///
3620 /// Panics when `index >= len()`, meaning it always panics on empty slices.
3621 ///
3622 /// May panic if `K: Ord` does not implement a total order.
3623 ///
3624 /// # Examples
3625 ///
3626 /// ```
3627 /// let mut v = [-5i32, 4, 1, -3, 2];
3628 ///
3629 /// // Find the items `<=` to the absolute median, the absolute median itself, and the items
3630 /// // `>=` to it.
3631 /// let (lesser, median, greater) = v.select_nth_unstable_by_key(2, |a| a.abs());
3632 ///
3633 /// assert!(lesser == [1, 2] || lesser == [2, 1]);
3634 /// assert_eq!(median, &mut -3);
3635 /// assert!(greater == [4, -5] || greater == [-5, 4]);
3636 ///
3637 /// // We are only guaranteed the slice will be one of the following, based on the way we sort
3638 /// // about the specified index.
3639 /// assert!(v == [1, 2, -3, 4, -5] ||
3640 /// v == [1, 2, -3, -5, 4] ||
3641 /// v == [2, 1, -3, 4, -5] ||
3642 /// v == [2, 1, -3, -5, 4]);
3643 /// ```
3644 ///
3645 /// [ipnsort]: https://github.com/Voultapher/sort-research-rs/tree/main/ipnsort
3646 /// [total order]: https://en.wikipedia.org/wiki/Total_order
3647 #[stable(feature = "slice_select_nth_unstable", since = "1.49.0")]
3648 #[inline]
3649 pub fn select_nth_unstable_by_key<K, F>(
3650 &mut self,
3651 index: usize,
3652 mut f: F,
3653 ) -> (&mut [T], &mut T, &mut [T])
3654 where
3655 F: FnMut(&T) -> K,
3656 K: Ord,
3657 {
3658 sort::select::partition_at_index(self, index, |a: &T, b: &T| f(a).lt(&f(b)))
3659 }
3660
3661 /// Moves all consecutive repeated elements to the end of the slice according to the
3662 /// [`PartialEq`] trait implementation.
3663 ///
3664 /// Returns two slices. The first contains no consecutive repeated elements.
3665 /// The second contains all the duplicates in no specified order.
3666 ///
3667 /// If the slice is sorted, the first returned slice contains no duplicates.
3668 ///
3669 /// # Examples
3670 ///
3671 /// ```
3672 /// #![feature(slice_partition_dedup)]
3673 ///
3674 /// let mut slice = [1, 2, 2, 3, 3, 2, 1, 1];
3675 ///
3676 /// let (dedup, duplicates) = slice.partition_dedup();
3677 ///
3678 /// assert_eq!(dedup, [1, 2, 3, 2, 1]);
3679 /// assert_eq!(duplicates, [2, 3, 1]);
3680 /// ```
3681 #[unstable(feature = "slice_partition_dedup", issue = "54279")]
3682 #[inline]
3683 pub fn partition_dedup(&mut self) -> (&mut [T], &mut [T])
3684 where
3685 T: PartialEq,
3686 {
3687 self.partition_dedup_by(|a, b| a == b)
3688 }
3689
3690 /// Moves all but the first of consecutive elements to the end of the slice satisfying
3691 /// a given equality relation.
3692 ///
3693 /// Returns two slices. The first contains no consecutive repeated elements.
3694 /// The second contains all the duplicates in no specified order.
3695 ///
3696 /// The `same_bucket` function is passed references to two elements from the slice and
3697 /// must determine if the elements compare equal. The elements are passed in opposite order
3698 /// from their order in the slice, so if `same_bucket(a, b)` returns `true`, `a` is moved
3699 /// at the end of the slice.
3700 ///
3701 /// If the slice is sorted, the first returned slice contains no duplicates.
3702 ///
3703 /// # Examples
3704 ///
3705 /// ```
3706 /// #![feature(slice_partition_dedup)]
3707 ///
3708 /// let mut slice = ["foo", "Foo", "BAZ", "Bar", "bar", "baz", "BAZ"];
3709 ///
3710 /// let (dedup, duplicates) = slice.partition_dedup_by(|a, b| a.eq_ignore_ascii_case(b));
3711 ///
3712 /// assert_eq!(dedup, ["foo", "BAZ", "Bar", "baz"]);
3713 /// assert_eq!(duplicates, ["bar", "Foo", "BAZ"]);
3714 /// ```
3715 #[unstable(feature = "slice_partition_dedup", issue = "54279")]
3716 #[inline]
3717 pub fn partition_dedup_by<F>(&mut self, mut same_bucket: F) -> (&mut [T], &mut [T])
3718 where
3719 F: FnMut(&mut T, &mut T) -> bool,
3720 {
3721 // Although we have a mutable reference to `self`, we cannot make
3722 // *arbitrary* changes. The `same_bucket` calls could panic, so we
3723 // must ensure that the slice is in a valid state at all times.
3724 //
3725 // The way that we handle this is by using swaps; we iterate
3726 // over all the elements, swapping as we go so that at the end
3727 // the elements we wish to keep are in the front, and those we
3728 // wish to reject are at the back. We can then split the slice.
3729 // This operation is still `O(n)`.
3730 //
3731 // Example: We start in this state, where `r` represents "next
3732 // read" and `w` represents "next_write".
3733 //
3734 // r
3735 // +---+---+---+---+---+---+
3736 // | 0 | 1 | 1 | 2 | 3 | 3 |
3737 // +---+---+---+---+---+---+
3738 // w
3739 //
3740 // Comparing self[r] against self[w-1], this is not a duplicate, so
3741 // we swap self[r] and self[w] (no effect as r==w) and then increment both
3742 // r and w, leaving us with:
3743 //
3744 // r
3745 // +---+---+---+---+---+---+
3746 // | 0 | 1 | 1 | 2 | 3 | 3 |
3747 // +---+---+---+---+---+---+
3748 // w
3749 //
3750 // Comparing self[r] against self[w-1], this value is a duplicate,
3751 // so we increment `r` but leave everything else unchanged:
3752 //
3753 // r
3754 // +---+---+---+---+---+---+
3755 // | 0 | 1 | 1 | 2 | 3 | 3 |
3756 // +---+---+---+---+---+---+
3757 // w
3758 //
3759 // Comparing self[r] against self[w-1], this is not a duplicate,
3760 // so swap self[r] and self[w] and advance r and w:
3761 //
3762 // r
3763 // +---+---+---+---+---+---+
3764 // | 0 | 1 | 2 | 1 | 3 | 3 |
3765 // +---+---+---+---+---+---+
3766 // w
3767 //
3768 // Not a duplicate, repeat:
3769 //
3770 // r
3771 // +---+---+---+---+---+---+
3772 // | 0 | 1 | 2 | 3 | 1 | 3 |
3773 // +---+---+---+---+---+---+
3774 // w
3775 //
3776 // Duplicate, advance r. End of slice. Split at w.
3777
3778 let len = self.len();
3779 if len <= 1 {
3780 return (self, &mut []);
3781 }
3782
3783 let ptr = self.as_mut_ptr();
3784 let mut next_read: usize = 1;
3785 let mut next_write: usize = 1;
3786
3787 // SAFETY: the `while` condition guarantees `next_read` and `next_write`
3788 // are less than `len`, thus are inside `self`. `prev_ptr_write` points to
3789 // one element before `ptr_write`, but `next_write` starts at 1, so
3790 // `prev_ptr_write` is never less than 0 and is inside the slice.
3791 // This fulfils the requirements for dereferencing `ptr_read`, `prev_ptr_write`
3792 // and `ptr_write`, and for using `ptr.add(next_read)`, `ptr.add(next_write - 1)`
3793 // and `prev_ptr_write.offset(1)`.
3794 //
3795 // `next_write` is also incremented at most once per loop at most meaning
3796 // no element is skipped when it may need to be swapped.
3797 //
3798 // `ptr_read` and `prev_ptr_write` never point to the same element. This
3799 // is required for `&mut *ptr_read`, `&mut *prev_ptr_write` to be safe.
3800 // The explanation is simply that `next_read >= next_write` is always true,
3801 // thus `next_read > next_write - 1` is too.
3802 unsafe {
3803 // Avoid bounds checks by using raw pointers.
3804 while next_read < len {
3805 let ptr_read = ptr.add(next_read);
3806 let prev_ptr_write = ptr.add(next_write - 1);
3807 if !same_bucket(&mut *ptr_read, &mut *prev_ptr_write) {
3808 if next_read != next_write {
3809 let ptr_write = prev_ptr_write.add(1);
3810 mem::swap(&mut *ptr_read, &mut *ptr_write);
3811 }
3812 next_write += 1;
3813 }
3814 next_read += 1;
3815 }
3816 }
3817
3818 self.split_at_mut(next_write)
3819 }
3820
3821 /// Moves all but the first of consecutive elements to the end of the slice that resolve
3822 /// to the same key.
3823 ///
3824 /// Returns two slices. The first contains no consecutive repeated elements.
3825 /// The second contains all the duplicates in no specified order.
3826 ///
3827 /// If the slice is sorted, the first returned slice contains no duplicates.
3828 ///
3829 /// # Examples
3830 ///
3831 /// ```
3832 /// #![feature(slice_partition_dedup)]
3833 ///
3834 /// let mut slice = [10, 20, 21, 30, 30, 20, 11, 13];
3835 ///
3836 /// let (dedup, duplicates) = slice.partition_dedup_by_key(|i| *i / 10);
3837 ///
3838 /// assert_eq!(dedup, [10, 20, 30, 20, 11]);
3839 /// assert_eq!(duplicates, [21, 30, 13]);
3840 /// ```
3841 #[unstable(feature = "slice_partition_dedup", issue = "54279")]
3842 #[inline]
3843 pub fn partition_dedup_by_key<K, F>(&mut self, mut key: F) -> (&mut [T], &mut [T])
3844 where
3845 F: FnMut(&mut T) -> K,
3846 K: PartialEq,
3847 {
3848 self.partition_dedup_by(|a, b| key(a) == key(b))
3849 }
3850
3851 /// Rotates the slice in-place such that the first `mid` elements of the
3852 /// slice move to the end while the last `self.len() - mid` elements move to
3853 /// the front.
3854 ///
3855 /// After calling `rotate_left`, the element previously at index `mid` will
3856 /// become the first element in the slice.
3857 ///
3858 /// # Panics
3859 ///
3860 /// This function will panic if `mid` is greater than the length of the
3861 /// slice. Note that `mid == self.len()` does _not_ panic and is a no-op
3862 /// rotation.
3863 ///
3864 /// # Complexity
3865 ///
3866 /// Takes linear (in `self.len()`) time.
3867 ///
3868 /// # Examples
3869 ///
3870 /// ```
3871 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3872 /// a.rotate_left(2);
3873 /// assert_eq!(a, ['c', 'd', 'e', 'f', 'a', 'b']);
3874 /// ```
3875 ///
3876 /// Rotating a subslice:
3877 ///
3878 /// ```
3879 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3880 /// a[1..5].rotate_left(1);
3881 /// assert_eq!(a, ['a', 'c', 'd', 'e', 'b', 'f']);
3882 /// ```
3883 #[stable(feature = "slice_rotate", since = "1.26.0")]
3884 #[rustc_const_stable(feature = "const_slice_rotate", since = "1.92.0")]
3885 pub const fn rotate_left(&mut self, mid: usize) {
3886 assert!(mid <= self.len());
3887 let k = self.len() - mid;
3888 let p = self.as_mut_ptr();
3889
3890 // SAFETY: The range `[p.add(mid) - mid, p.add(mid) + k)` is trivially
3891 // valid for reading and writing, as required by `ptr_rotate`.
3892 unsafe {
3893 rotate::ptr_rotate(mid, p.add(mid), k);
3894 }
3895 }
3896
3897 /// Rotates the slice in-place such that the first `self.len() - k`
3898 /// elements of the slice move to the end while the last `k` elements move
3899 /// to the front.
3900 ///
3901 /// After calling `rotate_right`, the element previously at index
3902 /// `self.len() - k` will become the first element in the slice.
3903 ///
3904 /// # Panics
3905 ///
3906 /// This function will panic if `k` is greater than the length of the
3907 /// slice. Note that `k == self.len()` does _not_ panic and is a no-op
3908 /// rotation.
3909 ///
3910 /// # Complexity
3911 ///
3912 /// Takes linear (in `self.len()`) time.
3913 ///
3914 /// # Examples
3915 ///
3916 /// ```
3917 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3918 /// a.rotate_right(2);
3919 /// assert_eq!(a, ['e', 'f', 'a', 'b', 'c', 'd']);
3920 /// ```
3921 ///
3922 /// Rotating a subslice:
3923 ///
3924 /// ```
3925 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3926 /// a[1..5].rotate_right(1);
3927 /// assert_eq!(a, ['a', 'e', 'b', 'c', 'd', 'f']);
3928 /// ```
3929 #[stable(feature = "slice_rotate", since = "1.26.0")]
3930 #[rustc_const_stable(feature = "const_slice_rotate", since = "1.92.0")]
3931 pub const fn rotate_right(&mut self, k: usize) {
3932 assert!(k <= self.len());
3933 let mid = self.len() - k;
3934 let p = self.as_mut_ptr();
3935
3936 // SAFETY: The range `[p.add(mid) - mid, p.add(mid) + k)` is trivially
3937 // valid for reading and writing, as required by `ptr_rotate`.
3938 unsafe {
3939 rotate::ptr_rotate(mid, p.add(mid), k);
3940 }
3941 }
3942
3943 /// Moves the elements of this slice `N` places to the left, returning the ones
3944 /// that "fall off" the front, and putting `inserted` at the end.
3945 ///
3946 /// Equivalently, you can think of concatenating `self` and `inserted` into one
3947 /// long sequence, then returning the left-most `N` items and the rest into `self`:
3948 ///
3949 /// ```text
3950 /// self (before) inserted
3951 /// vvvvvvvvvvvvvvv vvv
3952 /// [1, 2, 3, 4, 5] [9]
3953 /// ↙ ↙ ↙ ↙ ↙ ↙
3954 /// [1] [2, 3, 4, 5, 9]
3955 /// ^^^ ^^^^^^^^^^^^^^^
3956 /// returned self (after)
3957 /// ```
3958 ///
3959 /// See also [`Self::shift_right`] and compare [`Self::rotate_left`].
3960 ///
3961 /// # Examples
3962 ///
3963 /// ```
3964 /// #![feature(slice_shift)]
3965 ///
3966 /// // Same as the diagram above
3967 /// let mut a = [1, 2, 3, 4, 5];
3968 /// let inserted = [9];
3969 /// let returned = a.shift_left(inserted);
3970 /// assert_eq!(returned, [1]);
3971 /// assert_eq!(a, [2, 3, 4, 5, 9]);
3972 ///
3973 /// // You can shift multiple items at a time
3974 /// let mut a = *b"Hello world";
3975 /// assert_eq!(a.shift_left(*b" peace"), *b"Hello ");
3976 /// assert_eq!(a, *b"world peace");
3977 ///
3978 /// // The name comes from this operation's similarity to bitshifts
3979 /// let mut a: u8 = 0b10010110;
3980 /// a <<= 3;
3981 /// assert_eq!(a, 0b10110000_u8);
3982 /// let mut a: [_; 8] = [1, 0, 0, 1, 0, 1, 1, 0];
3983 /// a.shift_left([0; 3]);
3984 /// assert_eq!(a, [1, 0, 1, 1, 0, 0, 0, 0]);
3985 ///
3986 /// // Remember you can sub-slice to affect less that the whole slice.
3987 /// // For example, this is similar to `.remove(1)` + `.insert(4, 'Z')`
3988 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3989 /// assert_eq!(a[1..=4].shift_left(['Z']), ['b']);
3990 /// assert_eq!(a, ['a', 'c', 'd', 'e', 'Z', 'f']);
3991 ///
3992 /// // If the size matches it's equivalent to `mem::replace`
3993 /// let mut a = [1, 2, 3];
3994 /// assert_eq!(a.shift_left([7, 8, 9]), [1, 2, 3]);
3995 /// assert_eq!(a, [7, 8, 9]);
3996 ///
3997 /// // Some of the "inserted" elements end up returned if the slice is too short
3998 /// let mut a = [];
3999 /// assert_eq!(a.shift_left([1, 2, 3]), [1, 2, 3]);
4000 /// let mut a = [9];
4001 /// assert_eq!(a.shift_left([1, 2, 3]), [9, 1, 2]);
4002 /// assert_eq!(a, [3]);
4003 /// ```
4004 #[unstable(feature = "slice_shift", issue = "151772")]
4005 pub const fn shift_left<const N: usize>(&mut self, inserted: [T; N]) -> [T; N] {
4006 if let Some(shift) = self.len().checked_sub(N) {
4007 // SAFETY: Having just checked that the inserted/returned arrays are
4008 // shorter than (or the same length as) the slice:
4009 // 1. The read for the items to return is in-bounds
4010 // 2. We can `memmove` the slice over to cover the items we're returning
4011 // to ensure those aren't double-dropped
4012 // 3. Then we write (in-bounds for the same reason as the read) the
4013 // inserted items atop the items of the slice that we just duplicated
4014 //
4015 // And none of this can panic, so there's no risk of intermediate unwinds.
4016 unsafe {
4017 let ptr = self.as_mut_ptr();
4018 let returned = ptr.cast_array::<N>().read();
4019 ptr.copy_from(ptr.add(N), shift);
4020 ptr.add(shift).cast_array::<N>().write(inserted);
4021 returned
4022 }
4023 } else {
4024 // SAFETY: Having checked that the slice is strictly shorter than the
4025 // inserted/returned arrays, it means we'll be copying the whole slice
4026 // into the returned array, but that's not enough on its own. We also
4027 // need to copy some of the inserted array into the returned array,
4028 // with the rest going into the slice. Because `&mut` is exclusive
4029 // and we own both `inserted` and `returned`, they're all disjoint
4030 // allocations from each other as we can use `nonoverlapping` copies.
4031 //
4032 // We avoid double-frees by `ManuallyDrop`ing the inserted items,
4033 // since we always copy them to other locations that will drop them
4034 // instead. Plus nothing in here can panic -- it's just memcpy three
4035 // times -- so there's no intermediate unwinds to worry about.
4036 unsafe {
4037 let len = self.len();
4038 let slice = self.as_mut_ptr();
4039 let inserted = mem::ManuallyDrop::new(inserted);
4040 let inserted = (&raw const inserted).cast::<T>();
4041
4042 let mut returned = MaybeUninit::<[T; N]>::uninit();
4043 let ptr = returned.as_mut_ptr().cast::<T>();
4044 ptr.copy_from_nonoverlapping(slice, len);
4045 ptr.add(len).copy_from_nonoverlapping(inserted, N - len);
4046 slice.copy_from_nonoverlapping(inserted.add(N - len), len);
4047 returned.assume_init()
4048 }
4049 }
4050 }
4051
4052 /// Moves the elements of this slice `N` places to the right, returning the ones
4053 /// that "fall off" the back, and putting `inserted` at the beginning.
4054 ///
4055 /// Equivalently, you can think of concatenating `inserted` and `self` into one
4056 /// long sequence, then returning the right-most `N` items and the rest into `self`:
4057 ///
4058 /// ```text
4059 /// inserted self (before)
4060 /// vvv vvvvvvvvvvvvvvv
4061 /// [0] [5, 6, 7, 8, 9]
4062 /// ↘ ↘ ↘ ↘ ↘ ↘
4063 /// [0, 5, 6, 7, 8] [9]
4064 /// ^^^^^^^^^^^^^^^ ^^^
4065 /// self (after) returned
4066 /// ```
4067 ///
4068 /// See also [`Self::shift_left`] and compare [`Self::rotate_right`].
4069 ///
4070 /// # Examples
4071 ///
4072 /// ```
4073 /// #![feature(slice_shift)]
4074 ///
4075 /// // Same as the diagram above
4076 /// let mut a = [5, 6, 7, 8, 9];
4077 /// let inserted = [0];
4078 /// let returned = a.shift_right(inserted);
4079 /// assert_eq!(returned, [9]);
4080 /// assert_eq!(a, [0, 5, 6, 7, 8]);
4081 ///
4082 /// // The name comes from this operation's similarity to bitshifts
4083 /// let mut a: u8 = 0b10010110;
4084 /// a >>= 3;
4085 /// assert_eq!(a, 0b00010010_u8);
4086 /// let mut a: [_; 8] = [1, 0, 0, 1, 0, 1, 1, 0];
4087 /// a.shift_right([0; 3]);
4088 /// assert_eq!(a, [0, 0, 0, 1, 0, 0, 1, 0]);
4089 ///
4090 /// // Remember you can sub-slice to affect less that the whole slice.
4091 /// // For example, this is similar to `.remove(4)` + `.insert(1, 'Z')`
4092 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
4093 /// assert_eq!(a[1..=4].shift_right(['Z']), ['e']);
4094 /// assert_eq!(a, ['a', 'Z', 'b', 'c', 'd', 'f']);
4095 ///
4096 /// // If the size matches it's equivalent to `mem::replace`
4097 /// let mut a = [1, 2, 3];
4098 /// assert_eq!(a.shift_right([7, 8, 9]), [1, 2, 3]);
4099 /// assert_eq!(a, [7, 8, 9]);
4100 ///
4101 /// // Some of the "inserted" elements end up returned if the slice is too short
4102 /// let mut a = [];
4103 /// assert_eq!(a.shift_right([1, 2, 3]), [1, 2, 3]);
4104 /// let mut a = [9];
4105 /// assert_eq!(a.shift_right([1, 2, 3]), [2, 3, 9]);
4106 /// assert_eq!(a, [1]);
4107 /// ```
4108 #[unstable(feature = "slice_shift", issue = "151772")]
4109 pub const fn shift_right<const N: usize>(&mut self, inserted: [T; N]) -> [T; N] {
4110 if let Some(shift) = self.len().checked_sub(N) {
4111 // SAFETY: Having just checked that the inserted/returned arrays are
4112 // shorter than (or the same length as) the slice:
4113 // 1. The read for the items to return is in-bounds
4114 // 2. We can `memmove` the slice over to cover the items we're returning
4115 // to ensure those aren't double-dropped
4116 // 3. Then we write (in-bounds for the same reason as the read) the
4117 // inserted items atop the items of the slice that we just duplicated
4118 //
4119 // And none of this can panic, so there's no risk of intermediate unwinds.
4120 unsafe {
4121 let ptr = self.as_mut_ptr();
4122 let returned = ptr.add(shift).cast_array::<N>().read();
4123 ptr.add(N).copy_from(ptr, shift);
4124 ptr.cast_array::<N>().write(inserted);
4125 returned
4126 }
4127 } else {
4128 // SAFETY: Having checked that the slice is strictly shorter than the
4129 // inserted/returned arrays, it means we'll be copying the whole slice
4130 // into the returned array, but that's not enough on its own. We also
4131 // need to copy some of the inserted array into the returned array,
4132 // with the rest going into the slice. Because `&mut` is exclusive
4133 // and we own both `inserted` and `returned`, they're all disjoint
4134 // allocations from each other as we can use `nonoverlapping` copies.
4135 //
4136 // We avoid double-frees by `ManuallyDrop`ing the inserted items,
4137 // since we always copy them to other locations that will drop them
4138 // instead. Plus nothing in here can panic -- it's just memcpy three
4139 // times -- so there's no intermediate unwinds to worry about.
4140 unsafe {
4141 let len = self.len();
4142 let slice = self.as_mut_ptr();
4143 let inserted = mem::ManuallyDrop::new(inserted);
4144 let inserted = (&raw const inserted).cast::<T>();
4145
4146 let mut returned = MaybeUninit::<[T; N]>::uninit();
4147 let ptr = returned.as_mut_ptr().cast::<T>();
4148 ptr.add(N - len).copy_from_nonoverlapping(slice, len);
4149 ptr.copy_from_nonoverlapping(inserted.add(len), N - len);
4150 slice.copy_from_nonoverlapping(inserted, len);
4151 returned.assume_init()
4152 }
4153 }
4154 }
4155
4156 /// Fills `self` with elements by cloning `value`.
4157 ///
4158 /// # Examples
4159 ///
4160 /// ```
4161 /// let mut buf = vec![0; 10];
4162 /// buf.fill(1);
4163 /// assert_eq!(buf, vec![1; 10]);
4164 /// ```
4165 #[doc(alias = "memset")]
4166 #[stable(feature = "slice_fill", since = "1.50.0")]
4167 pub fn fill(&mut self, value: T)
4168 where
4169 T: Clone,
4170 {
4171 specialize::SpecFill::spec_fill(self, value);
4172 }
4173
4174 /// Fills `self` with elements returned by calling a closure repeatedly.
4175 ///
4176 /// This method uses a closure to create new values. If you'd rather
4177 /// [`Clone`] a given value, use [`fill`]. If you want to use the [`Default`]
4178 /// trait to generate values, you can pass [`Default::default`] as the
4179 /// argument.
4180 ///
4181 /// [`fill`]: slice::fill
4182 ///
4183 /// # Examples
4184 ///
4185 /// ```
4186 /// let mut buf = vec![1; 10];
4187 /// buf.fill_with(Default::default);
4188 /// assert_eq!(buf, vec![0; 10]);
4189 /// ```
4190 #[stable(feature = "slice_fill_with", since = "1.51.0")]
4191 pub fn fill_with<F>(&mut self, mut f: F)
4192 where
4193 F: FnMut() -> T,
4194 {
4195 for el in self {
4196 *el = f();
4197 }
4198 }
4199
4200 /// Copies the elements from `src` into `self`.
4201 ///
4202 /// The length of `src` must be the same as `self`.
4203 ///
4204 /// # Panics
4205 ///
4206 /// This function will panic if the two slices have different lengths.
4207 ///
4208 /// # Examples
4209 ///
4210 /// Cloning two elements from a slice into another:
4211 ///
4212 /// ```
4213 /// let src = [1, 2, 3, 4];
4214 /// let mut dst = [0, 0];
4215 ///
4216 /// // Because the slices have to be the same length,
4217 /// // we slice the source slice from four elements
4218 /// // to two. It will panic if we don't do this.
4219 /// dst.clone_from_slice(&src[2..]);
4220 ///
4221 /// assert_eq!(src, [1, 2, 3, 4]);
4222 /// assert_eq!(dst, [3, 4]);
4223 /// ```
4224 ///
4225 /// Rust enforces that there can only be one mutable reference with no
4226 /// immutable references to a particular piece of data in a particular
4227 /// scope. Because of this, attempting to use `clone_from_slice` on a
4228 /// single slice will result in a compile failure:
4229 ///
4230 /// ```compile_fail
4231 /// let mut slice = [1, 2, 3, 4, 5];
4232 ///
4233 /// slice[..2].clone_from_slice(&slice[3..]); // compile fail!
4234 /// ```
4235 ///
4236 /// To work around this, we can use [`split_at_mut`] to create two distinct
4237 /// sub-slices from a slice:
4238 ///
4239 /// ```
4240 /// let mut slice = [1, 2, 3, 4, 5];
4241 ///
4242 /// {
4243 /// let (left, right) = slice.split_at_mut(2);
4244 /// left.clone_from_slice(&right[1..]);
4245 /// }
4246 ///
4247 /// assert_eq!(slice, [4, 5, 3, 4, 5]);
4248 /// ```
4249 ///
4250 /// [`copy_from_slice`]: slice::copy_from_slice
4251 /// [`split_at_mut`]: slice::split_at_mut
4252 #[stable(feature = "clone_from_slice", since = "1.7.0")]
4253 #[track_caller]
4254 #[rustc_const_unstable(feature = "const_clone", issue = "142757")]
4255 pub const fn clone_from_slice(&mut self, src: &[T])
4256 where
4257 T: [const] Clone + [const] Destruct,
4258 {
4259 self.spec_clone_from(src);
4260 }
4261
4262 /// Copies all elements from `src` into `self`, using a memcpy.
4263 ///
4264 /// The length of `src` must be the same as `self`.
4265 ///
4266 /// If `T` does not implement `Copy`, use [`clone_from_slice`].
4267 ///
4268 /// # Panics
4269 ///
4270 /// This function will panic if the two slices have different lengths.
4271 ///
4272 /// # Examples
4273 ///
4274 /// Copying two elements from a slice into another:
4275 ///
4276 /// ```
4277 /// let src = [1, 2, 3, 4];
4278 /// let mut dst = [0, 0];
4279 ///
4280 /// // Because the slices have to be the same length,
4281 /// // we slice the source slice from four elements
4282 /// // to two. It will panic if we don't do this.
4283 /// dst.copy_from_slice(&src[2..]);
4284 ///
4285 /// assert_eq!(src, [1, 2, 3, 4]);
4286 /// assert_eq!(dst, [3, 4]);
4287 /// ```
4288 ///
4289 /// Rust enforces that there can only be one mutable reference with no
4290 /// immutable references to a particular piece of data in a particular
4291 /// scope. Because of this, attempting to use `copy_from_slice` on a
4292 /// single slice will result in a compile failure:
4293 ///
4294 /// ```compile_fail
4295 /// let mut slice = [1, 2, 3, 4, 5];
4296 ///
4297 /// slice[..2].copy_from_slice(&slice[3..]); // compile fail!
4298 /// ```
4299 ///
4300 /// To work around this, we can use [`split_at_mut`] to create two distinct
4301 /// sub-slices from a slice:
4302 ///
4303 /// ```
4304 /// let mut slice = [1, 2, 3, 4, 5];
4305 ///
4306 /// {
4307 /// let (left, right) = slice.split_at_mut(2);
4308 /// left.copy_from_slice(&right[1..]);
4309 /// }
4310 ///
4311 /// assert_eq!(slice, [4, 5, 3, 4, 5]);
4312 /// ```
4313 ///
4314 /// [`clone_from_slice`]: slice::clone_from_slice
4315 /// [`split_at_mut`]: slice::split_at_mut
4316 #[doc(alias = "memcpy")]
4317 #[inline]
4318 #[stable(feature = "copy_from_slice", since = "1.9.0")]
4319 #[rustc_const_stable(feature = "const_copy_from_slice", since = "1.87.0")]
4320 #[track_caller]
4321 pub const fn copy_from_slice(&mut self, src: &[T])
4322 where
4323 T: Copy,
4324 {
4325 // SAFETY: `T` implements `Copy`.
4326 unsafe { copy_from_slice_impl(self, src) }
4327 }
4328
4329 /// Copies elements from one part of the slice to another part of itself,
4330 /// using a memmove.
4331 ///
4332 /// `src` is the range within `self` to copy from. `dest` is the starting
4333 /// index of the range within `self` to copy to, which will have the same
4334 /// length as `src`. The two ranges may overlap. The ends of the two ranges
4335 /// must be less than or equal to `self.len()`.
4336 ///
4337 /// # Panics
4338 ///
4339 /// This function will panic if either range exceeds the end of the slice,
4340 /// or if the end of `src` is before the start.
4341 ///
4342 /// # Examples
4343 ///
4344 /// Copying four bytes within a slice:
4345 ///
4346 /// ```
4347 /// let mut bytes = *b"Hello, World!";
4348 ///
4349 /// bytes.copy_within(1..5, 8);
4350 ///
4351 /// assert_eq!(&bytes, b"Hello, Wello!");
4352 /// ```
4353 #[inline]
4354 #[stable(feature = "copy_within", since = "1.37.0")]
4355 #[track_caller]
4356 pub fn copy_within<R: RangeBounds<usize>>(&mut self, src: R, dest: usize)
4357 where
4358 T: Copy,
4359 {
4360 let Range { start: src_start, end: src_end } = slice::range(src, ..self.len());
4361 let count = src_end - src_start;
4362 assert!(dest <= self.len() - count, "dest is out of bounds");
4363 // SAFETY: the conditions for `ptr::copy` have all been checked above,
4364 // as have those for `ptr::add`.
4365 unsafe {
4366 // Derive both `src_ptr` and `dest_ptr` from the same loan
4367 let ptr = self.as_mut_ptr();
4368 let src_ptr = ptr.add(src_start);
4369 let dest_ptr = ptr.add(dest);
4370 ptr::copy(src_ptr, dest_ptr, count);
4371 }
4372 }
4373
4374 /// Swaps all elements in `self` with those in `other`.
4375 ///
4376 /// The length of `other` must be the same as `self`.
4377 ///
4378 /// # Panics
4379 ///
4380 /// This function will panic if the two slices have different lengths.
4381 ///
4382 /// # Example
4383 ///
4384 /// Swapping two elements across slices:
4385 ///
4386 /// ```
4387 /// let mut slice1 = [0, 0];
4388 /// let mut slice2 = [1, 2, 3, 4];
4389 ///
4390 /// slice1.swap_with_slice(&mut slice2[2..]);
4391 ///
4392 /// assert_eq!(slice1, [3, 4]);
4393 /// assert_eq!(slice2, [1, 2, 0, 0]);
4394 /// ```
4395 ///
4396 /// Rust enforces that there can only be one mutable reference to a
4397 /// particular piece of data in a particular scope. Because of this,
4398 /// attempting to use `swap_with_slice` on a single slice will result in
4399 /// a compile failure:
4400 ///
4401 /// ```compile_fail
4402 /// let mut slice = [1, 2, 3, 4, 5];
4403 /// slice[..2].swap_with_slice(&mut slice[3..]); // compile fail!
4404 /// ```
4405 ///
4406 /// To work around this, we can use [`split_at_mut`] to create two distinct
4407 /// mutable sub-slices from a slice:
4408 ///
4409 /// ```
4410 /// let mut slice = [1, 2, 3, 4, 5];
4411 ///
4412 /// {
4413 /// let (left, right) = slice.split_at_mut(2);
4414 /// left.swap_with_slice(&mut right[1..]);
4415 /// }
4416 ///
4417 /// assert_eq!(slice, [4, 5, 3, 1, 2]);
4418 /// ```
4419 ///
4420 /// [`split_at_mut`]: slice::split_at_mut
4421 #[stable(feature = "swap_with_slice", since = "1.27.0")]
4422 #[rustc_const_unstable(feature = "const_swap_with_slice", issue = "142204")]
4423 #[track_caller]
4424 pub const fn swap_with_slice(&mut self, other: &mut [T]) {
4425 assert!(self.len() == other.len(), "destination and source slices have different lengths");
4426 // SAFETY: `self` is valid for `self.len()` elements by definition, and `src` was
4427 // checked to have the same length. The slices cannot overlap because
4428 // mutable references are exclusive.
4429 unsafe {
4430 ptr::swap_nonoverlapping(self.as_mut_ptr(), other.as_mut_ptr(), self.len());
4431 }
4432 }
4433
4434 /// Function to calculate lengths of the middle and trailing slice for `align_to{,_mut}`.
4435 fn align_to_offsets<U>(&self) -> (usize, usize) {
4436 // What we gonna do about `rest` is figure out what multiple of `U`s we can put in a
4437 // lowest number of `T`s. And how many `T`s we need for each such "multiple".
4438 //
4439 // Consider for example T=u8 U=u16. Then we can put 1 U in 2 Ts. Simple. Now, consider
4440 // for example a case where size_of::<T> = 16, size_of::<U> = 24. We can put 2 Us in
4441 // place of every 3 Ts in the `rest` slice. A bit more complicated.
4442 //
4443 // Formula to calculate this is:
4444 //
4445 // Us = lcm(size_of::<T>, size_of::<U>) / size_of::<U>
4446 // Ts = lcm(size_of::<T>, size_of::<U>) / size_of::<T>
4447 //
4448 // Expanded and simplified:
4449 //
4450 // Us = size_of::<T> / gcd(size_of::<T>, size_of::<U>)
4451 // Ts = size_of::<U> / gcd(size_of::<T>, size_of::<U>)
4452 //
4453 // Luckily since all this is constant-evaluated... performance here matters not!
4454 const fn gcd(a: usize, b: usize) -> usize {
4455 if b == 0 { a } else { gcd(b, a % b) }
4456 }
4457
4458 // Explicitly wrap the function call in a const block so it gets
4459 // constant-evaluated even in debug mode.
4460 let gcd: usize = const { gcd(size_of::<T>(), size_of::<U>()) };
4461 let ts: usize = size_of::<U>() / gcd;
4462 let us: usize = size_of::<T>() / gcd;
4463
4464 // Armed with this knowledge, we can find how many `U`s we can fit!
4465 let us_len = self.len() / ts * us;
4466 // And how many `T`s will be in the trailing slice!
4467 let ts_len = self.len() % ts;
4468 (us_len, ts_len)
4469 }
4470
4471 /// Transmutes the slice to a slice of another type, ensuring alignment of the types is
4472 /// maintained.
4473 ///
4474 /// This method splits the slice into three distinct slices: prefix, correctly aligned middle
4475 /// slice of a new type, and the suffix slice. The middle part will be as big as possible under
4476 /// the given alignment constraint and element size.
4477 ///
4478 /// This method has no purpose when either input element `T` or output element `U` are
4479 /// zero-sized and will return the original slice without splitting anything.
4480 ///
4481 /// # Safety
4482 ///
4483 /// This method is essentially a `transmute` with respect to the elements in the returned
4484 /// middle slice, so all the usual caveats pertaining to `transmute::<T, U>` also apply here.
4485 ///
4486 /// # Examples
4487 ///
4488 /// Basic usage:
4489 ///
4490 /// ```
4491 /// unsafe {
4492 /// let bytes: [u8; 7] = [1, 2, 3, 4, 5, 6, 7];
4493 /// let (prefix, shorts, suffix) = bytes.align_to::<u16>();
4494 /// // less_efficient_algorithm_for_bytes(prefix);
4495 /// // more_efficient_algorithm_for_aligned_shorts(shorts);
4496 /// // less_efficient_algorithm_for_bytes(suffix);
4497 /// }
4498 /// ```
4499 #[stable(feature = "slice_align_to", since = "1.30.0")]
4500 #[must_use]
4501 pub unsafe fn align_to<U>(&self) -> (&[T], &[U], &[T]) {
4502 // Note that most of this function will be constant-evaluated,
4503 if U::IS_ZST || T::IS_ZST {
4504 // handle ZSTs specially, which is – don't handle them at all.
4505 return (self, &[], &[]);
4506 }
4507
4508 // First, find at what point do we split between the first and 2nd slice. Easy with
4509 // ptr.align_offset.
4510 let ptr = self.as_ptr();
4511 // SAFETY: See the `align_to_mut` method for the detailed safety comment.
4512 let offset = unsafe { crate::ptr::align_offset(ptr, align_of::<U>()) };
4513 if offset > self.len() {
4514 (self, &[], &[])
4515 } else {
4516 let (left, rest) = self.split_at(offset);
4517 let (us_len, ts_len) = rest.align_to_offsets::<U>();
4518 // Inform Miri that we want to consider the "middle" pointer to be suitably aligned.
4519 #[cfg(miri)]
4520 crate::intrinsics::miri_promise_symbolic_alignment(
4521 rest.as_ptr().cast(),
4522 align_of::<U>(),
4523 );
4524 // SAFETY: now `rest` is definitely aligned, so `from_raw_parts` below is okay,
4525 // since the caller guarantees that we can transmute `T` to `U` safely.
4526 unsafe {
4527 (
4528 left,
4529 from_raw_parts(rest.as_ptr() as *const U, us_len),
4530 from_raw_parts(rest.as_ptr().add(rest.len() - ts_len), ts_len),
4531 )
4532 }
4533 }
4534 }
4535
4536 /// Transmutes the mutable slice to a mutable slice of another type, ensuring alignment of the
4537 /// types is maintained.
4538 ///
4539 /// This method splits the slice into three distinct slices: prefix, correctly aligned middle
4540 /// slice of a new type, and the suffix slice. The middle part will be as big as possible under
4541 /// the given alignment constraint and element size.
4542 ///
4543 /// This method has no purpose when either input element `T` or output element `U` are
4544 /// zero-sized and will return the original slice without splitting anything.
4545 ///
4546 /// # Safety
4547 ///
4548 /// This method is essentially a `transmute` with respect to the elements in the returned
4549 /// middle slice, so all the usual caveats pertaining to `transmute::<T, U>` also apply here.
4550 ///
4551 /// # Examples
4552 ///
4553 /// Basic usage:
4554 ///
4555 /// ```
4556 /// unsafe {
4557 /// let mut bytes: [u8; 7] = [1, 2, 3, 4, 5, 6, 7];
4558 /// let (prefix, shorts, suffix) = bytes.align_to_mut::<u16>();
4559 /// // less_efficient_algorithm_for_bytes(prefix);
4560 /// // more_efficient_algorithm_for_aligned_shorts(shorts);
4561 /// // less_efficient_algorithm_for_bytes(suffix);
4562 /// }
4563 /// ```
4564 #[stable(feature = "slice_align_to", since = "1.30.0")]
4565 #[must_use]
4566 pub unsafe fn align_to_mut<U>(&mut self) -> (&mut [T], &mut [U], &mut [T]) {
4567 // Note that most of this function will be constant-evaluated,
4568 if U::IS_ZST || T::IS_ZST {
4569 // handle ZSTs specially, which is – don't handle them at all.
4570 return (self, &mut [], &mut []);
4571 }
4572
4573 // First, find at what point do we split between the first and 2nd slice. Easy with
4574 // ptr.align_offset.
4575 let ptr = self.as_ptr();
4576 // SAFETY: Here we are ensuring we will use aligned pointers for U for the
4577 // rest of the method. This is done by passing a pointer to &[T] with an
4578 // alignment targeted for U.
4579 // `crate::ptr::align_offset` is called with a correctly aligned and
4580 // valid pointer `ptr` (it comes from a reference to `self`) and with
4581 // a size that is a power of two (since it comes from the alignment for U),
4582 // satisfying its safety constraints.
4583 let offset = unsafe { crate::ptr::align_offset(ptr, align_of::<U>()) };
4584 if offset > self.len() {
4585 (self, &mut [], &mut [])
4586 } else {
4587 let (left, rest) = self.split_at_mut(offset);
4588 let (us_len, ts_len) = rest.align_to_offsets::<U>();
4589 let rest_len = rest.len();
4590 let mut_ptr = rest.as_mut_ptr();
4591 // Inform Miri that we want to consider the "middle" pointer to be suitably aligned.
4592 #[cfg(miri)]
4593 crate::intrinsics::miri_promise_symbolic_alignment(
4594 mut_ptr.cast() as *const (),
4595 align_of::<U>(),
4596 );
4597 // We can't use `rest` again after this, that would invalidate its alias `mut_ptr`!
4598 // SAFETY: see comments for `align_to`.
4599 unsafe {
4600 (
4601 left,
4602 from_raw_parts_mut(mut_ptr as *mut U, us_len),
4603 from_raw_parts_mut(mut_ptr.add(rest_len - ts_len), ts_len),
4604 )
4605 }
4606 }
4607 }
4608
4609 /// Splits a slice into a prefix, a middle of aligned SIMD types, and a suffix.
4610 ///
4611 /// This is a safe wrapper around [`slice::align_to`], so inherits the same
4612 /// guarantees as that method.
4613 ///
4614 /// # Panics
4615 ///
4616 /// This will panic if the size of the SIMD type is different from
4617 /// `LANES` times that of the scalar.
4618 ///
4619 /// At the time of writing, the trait restrictions on `Simd<T, LANES>` keeps
4620 /// that from ever happening, as only power-of-two numbers of lanes are
4621 /// supported. It's possible that, in the future, those restrictions might
4622 /// be lifted in a way that would make it possible to see panics from this
4623 /// method for something like `LANES == 3`.
4624 ///
4625 /// # Examples
4626 ///
4627 /// ```
4628 /// #![feature(portable_simd)]
4629 /// use core::simd::prelude::*;
4630 ///
4631 /// let short = &[1, 2, 3];
4632 /// let (prefix, middle, suffix) = short.as_simd::<4>();
4633 /// assert_eq!(middle, []); // Not enough elements for anything in the middle
4634 ///
4635 /// // They might be split in any possible way between prefix and suffix
4636 /// let it = prefix.iter().chain(suffix).copied();
4637 /// assert_eq!(it.collect::<Vec<_>>(), vec![1, 2, 3]);
4638 ///
4639 /// fn basic_simd_sum(x: &[f32]) -> f32 {
4640 /// use std::ops::Add;
4641 /// let (prefix, middle, suffix) = x.as_simd();
4642 /// let sums = f32x4::from_array([
4643 /// prefix.iter().copied().sum(),
4644 /// 0.0,
4645 /// 0.0,
4646 /// suffix.iter().copied().sum(),
4647 /// ]);
4648 /// let sums = middle.iter().copied().fold(sums, f32x4::add);
4649 /// sums.reduce_sum()
4650 /// }
4651 ///
4652 /// let numbers: Vec<f32> = (1..101).map(|x| x as _).collect();
4653 /// assert_eq!(basic_simd_sum(&numbers[1..99]), 4949.0);
4654 /// ```
4655 #[unstable(feature = "portable_simd", issue = "86656")]
4656 #[must_use]
4657 pub fn as_simd<const LANES: usize>(&self) -> (&[T], &[Simd<T, LANES>], &[T])
4658 where
4659 Simd<T, LANES>: AsRef<[T; LANES]>,
4660 T: simd::SimdElement,
4661 {
4662 // These are expected to always match, as vector types are laid out like
4663 // arrays per <https://llvm.org/docs/LangRef.html#vector-type>, but we
4664 // might as well double-check since it'll optimize away anyhow.
4665 assert_eq!(size_of::<Simd<T, LANES>>(), size_of::<[T; LANES]>());
4666
4667 // SAFETY: The simd types have the same layout as arrays, just with
4668 // potentially-higher alignment, so the de-facto transmutes are sound.
4669 unsafe { self.align_to() }
4670 }
4671
4672 /// Splits a mutable slice into a mutable prefix, a middle of aligned SIMD types,
4673 /// and a mutable suffix.
4674 ///
4675 /// This is a safe wrapper around [`slice::align_to_mut`], so inherits the same
4676 /// guarantees as that method.
4677 ///
4678 /// This is the mutable version of [`slice::as_simd`]; see that for examples.
4679 ///
4680 /// # Panics
4681 ///
4682 /// This will panic if the size of the SIMD type is different from
4683 /// `LANES` times that of the scalar.
4684 ///
4685 /// At the time of writing, the trait restrictions on `Simd<T, LANES>` keeps
4686 /// that from ever happening, as only power-of-two numbers of lanes are
4687 /// supported. It's possible that, in the future, those restrictions might
4688 /// be lifted in a way that would make it possible to see panics from this
4689 /// method for something like `LANES == 3`.
4690 #[unstable(feature = "portable_simd", issue = "86656")]
4691 #[must_use]
4692 pub fn as_simd_mut<const LANES: usize>(&mut self) -> (&mut [T], &mut [Simd<T, LANES>], &mut [T])
4693 where
4694 Simd<T, LANES>: AsMut<[T; LANES]>,
4695 T: simd::SimdElement,
4696 {
4697 // These are expected to always match, as vector types are laid out like
4698 // arrays per <https://llvm.org/docs/LangRef.html#vector-type>, but we
4699 // might as well double-check since it'll optimize away anyhow.
4700 assert_eq!(size_of::<Simd<T, LANES>>(), size_of::<[T; LANES]>());
4701
4702 // SAFETY: The simd types have the same layout as arrays, just with
4703 // potentially-higher alignment, so the de-facto transmutes are sound.
4704 unsafe { self.align_to_mut() }
4705 }
4706
4707 /// Checks if the elements of this slice are sorted.
4708 ///
4709 /// That is, for each element `a` and its following element `b`, `a <= b` must hold. If the
4710 /// slice yields exactly zero or one element, `true` is returned.
4711 ///
4712 /// Note that if `Self::Item` is only `PartialOrd`, but not `Ord`, the above definition
4713 /// implies that this function returns `false` if any two consecutive items are not
4714 /// comparable.
4715 ///
4716 /// # Examples
4717 ///
4718 /// ```
4719 /// let empty: [i32; 0] = [];
4720 ///
4721 /// assert!([1, 2, 2, 9].is_sorted());
4722 /// assert!(![1, 3, 2, 4].is_sorted());
4723 /// assert!([0].is_sorted());
4724 /// assert!(empty.is_sorted());
4725 /// assert!(![0.0, 1.0, f32::NAN].is_sorted());
4726 /// ```
4727 #[inline]
4728 #[stable(feature = "is_sorted", since = "1.82.0")]
4729 #[must_use]
4730 pub fn is_sorted(&self) -> bool
4731 where
4732 T: PartialOrd,
4733 {
4734 // This odd number works the best. 32 + 1 extra due to overlapping chunk boundaries.
4735 const CHUNK_SIZE: usize = 33;
4736 if self.len() < CHUNK_SIZE {
4737 return self.windows(2).all(|w| w[0] <= w[1]);
4738 }
4739 let mut i = 0;
4740 // Check in chunks for autovectorization.
4741 while i < self.len() - CHUNK_SIZE {
4742 let chunk = &self[i..i + CHUNK_SIZE];
4743 if !chunk.windows(2).fold(true, |acc, w| acc & (w[0] <= w[1])) {
4744 return false;
4745 }
4746 // We need to ensure that chunk boundaries are also sorted.
4747 // Overlap the next chunk with the last element of our last chunk.
4748 i += CHUNK_SIZE - 1;
4749 }
4750 self[i..].windows(2).all(|w| w[0] <= w[1])
4751 }
4752
4753 /// Checks if the elements of this slice are sorted using the given comparator function.
4754 ///
4755 /// Instead of using `PartialOrd::partial_cmp`, this function uses the given `compare`
4756 /// function to determine whether two elements are to be considered in sorted order.
4757 ///
4758 /// # Examples
4759 ///
4760 /// ```
4761 /// assert!([1, 2, 2, 9].is_sorted_by(|a, b| a <= b));
4762 /// assert!(![1, 2, 2, 9].is_sorted_by(|a, b| a < b));
4763 ///
4764 /// assert!([0].is_sorted_by(|a, b| true));
4765 /// assert!([0].is_sorted_by(|a, b| false));
4766 ///
4767 /// let empty: [i32; 0] = [];
4768 /// assert!(empty.is_sorted_by(|a, b| false));
4769 /// assert!(empty.is_sorted_by(|a, b| true));
4770 /// ```
4771 #[stable(feature = "is_sorted", since = "1.82.0")]
4772 #[must_use]
4773 pub fn is_sorted_by<'a, F>(&'a self, mut compare: F) -> bool
4774 where
4775 F: FnMut(&'a T, &'a T) -> bool,
4776 {
4777 self.array_windows().all(|[a, b]| compare(a, b))
4778 }
4779
4780 /// Checks if the elements of this slice are sorted using the given key extraction function.
4781 ///
4782 /// Instead of comparing the slice's elements directly, this function compares the keys of the
4783 /// elements, as determined by `f`. Apart from that, it's equivalent to [`is_sorted`]; see its
4784 /// documentation for more information.
4785 ///
4786 /// [`is_sorted`]: slice::is_sorted
4787 ///
4788 /// # Examples
4789 ///
4790 /// ```
4791 /// assert!(["c", "bb", "aaa"].is_sorted_by_key(|s| s.len()));
4792 /// assert!(![-2i32, -1, 0, 3].is_sorted_by_key(|n| n.abs()));
4793 /// ```
4794 #[inline]
4795 #[stable(feature = "is_sorted", since = "1.82.0")]
4796 #[must_use]
4797 pub fn is_sorted_by_key<'a, F, K>(&'a self, f: F) -> bool
4798 where
4799 F: FnMut(&'a T) -> K,
4800 K: PartialOrd,
4801 {
4802 self.iter().is_sorted_by_key(f)
4803 }
4804
4805 /// Returns the index of the partition point according to the given predicate
4806 /// (the index of the first element of the second partition).
4807 ///
4808 /// The slice is assumed to be partitioned according to the given predicate.
4809 /// This means that all elements for which the predicate returns true are at the start of the slice
4810 /// and all elements for which the predicate returns false are at the end.
4811 /// For example, `[7, 15, 3, 5, 4, 12, 6]` is partitioned under the predicate `x % 2 != 0`
4812 /// (all odd numbers are at the start, all even at the end).
4813 ///
4814 /// If this slice is not partitioned, the returned result is unspecified and meaningless,
4815 /// as this method performs a kind of binary search.
4816 ///
4817 /// See also [`binary_search`], [`binary_search_by`], and [`binary_search_by_key`].
4818 ///
4819 /// [`binary_search`]: slice::binary_search
4820 /// [`binary_search_by`]: slice::binary_search_by
4821 /// [`binary_search_by_key`]: slice::binary_search_by_key
4822 ///
4823 /// # Examples
4824 ///
4825 /// ```
4826 /// let v = [1, 2, 3, 3, 5, 6, 7];
4827 /// let i = v.partition_point(|&x| x < 5);
4828 ///
4829 /// assert_eq!(i, 4);
4830 /// assert!(v[..i].iter().all(|&x| x < 5));
4831 /// assert!(v[i..].iter().all(|&x| !(x < 5)));
4832 /// ```
4833 ///
4834 /// If all elements of the slice match the predicate, including if the slice
4835 /// is empty, then the length of the slice will be returned:
4836 ///
4837 /// ```
4838 /// let a = [2, 4, 8];
4839 /// assert_eq!(a.partition_point(|x| x < &100), a.len());
4840 /// let a: [i32; 0] = [];
4841 /// assert_eq!(a.partition_point(|x| x < &100), 0);
4842 /// ```
4843 ///
4844 /// If you want to insert an item to a sorted vector, while maintaining
4845 /// sort order:
4846 ///
4847 /// ```
4848 /// let mut s = vec![0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
4849 /// let num = 42;
4850 /// let idx = s.partition_point(|&x| x <= num);
4851 /// s.insert(idx, num);
4852 /// assert_eq!(s, [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 42, 55]);
4853 /// ```
4854 #[stable(feature = "partition_point", since = "1.52.0")]
4855 #[must_use]
4856 pub fn partition_point<P>(&self, mut pred: P) -> usize
4857 where
4858 P: FnMut(&T) -> bool,
4859 {
4860 self.binary_search_by(|x| if pred(x) { Less } else { Greater }).unwrap_or_else(|i| i)
4861 }
4862
4863 /// Removes the subslice corresponding to the given range
4864 /// and returns a reference to it.
4865 ///
4866 /// Returns `None` and does not modify the slice if the given
4867 /// range is out of bounds.
4868 ///
4869 /// Note that this method only accepts one-sided ranges such as
4870 /// `2..` or `..6`, but not `2..6`.
4871 ///
4872 /// # Examples
4873 ///
4874 /// Splitting off the first three elements of a slice:
4875 ///
4876 /// ```
4877 /// let mut slice: &[_] = &['a', 'b', 'c', 'd'];
4878 /// let mut first_three = slice.split_off(..3).unwrap();
4879 ///
4880 /// assert_eq!(slice, &['d']);
4881 /// assert_eq!(first_three, &['a', 'b', 'c']);
4882 /// ```
4883 ///
4884 /// Splitting off a slice starting with the third element:
4885 ///
4886 /// ```
4887 /// let mut slice: &[_] = &['a', 'b', 'c', 'd'];
4888 /// let mut tail = slice.split_off(2..).unwrap();
4889 ///
4890 /// assert_eq!(slice, &['a', 'b']);
4891 /// assert_eq!(tail, &['c', 'd']);
4892 /// ```
4893 ///
4894 /// Getting `None` when `range` is out of bounds:
4895 ///
4896 /// ```
4897 /// let mut slice: &[_] = &['a', 'b', 'c', 'd'];
4898 ///
4899 /// assert_eq!(None, slice.split_off(5..));
4900 /// assert_eq!(None, slice.split_off(..5));
4901 /// assert_eq!(None, slice.split_off(..=4));
4902 /// let expected: &[char] = &['a', 'b', 'c', 'd'];
4903 /// assert_eq!(Some(expected), slice.split_off(..4));
4904 /// ```
4905 #[inline]
4906 #[must_use = "method does not modify the slice if the range is out of bounds"]
4907 #[stable(feature = "slice_take", since = "1.87.0")]
4908 pub fn split_off<'a, R: OneSidedRange<usize>>(
4909 self: &mut &'a Self,
4910 range: R,
4911 ) -> Option<&'a Self> {
4912 let (direction, split_index) = split_point_of(range)?;
4913 if split_index > self.len() {
4914 return None;
4915 }
4916 let (front, back) = self.split_at(split_index);
4917 match direction {
4918 Direction::Front => {
4919 *self = back;
4920 Some(front)
4921 }
4922 Direction::Back => {
4923 *self = front;
4924 Some(back)
4925 }
4926 }
4927 }
4928
4929 /// Removes the subslice corresponding to the given range
4930 /// and returns a mutable reference to it.
4931 ///
4932 /// Returns `None` and does not modify the slice if the given
4933 /// range is out of bounds.
4934 ///
4935 /// Note that this method only accepts one-sided ranges such as
4936 /// `2..` or `..6`, but not `2..6`.
4937 ///
4938 /// # Examples
4939 ///
4940 /// Splitting off the first three elements of a slice:
4941 ///
4942 /// ```
4943 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c', 'd'];
4944 /// let mut first_three = slice.split_off_mut(..3).unwrap();
4945 ///
4946 /// assert_eq!(slice, &mut ['d']);
4947 /// assert_eq!(first_three, &mut ['a', 'b', 'c']);
4948 /// ```
4949 ///
4950 /// Splitting off a slice starting with the third element:
4951 ///
4952 /// ```
4953 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c', 'd'];
4954 /// let mut tail = slice.split_off_mut(2..).unwrap();
4955 ///
4956 /// assert_eq!(slice, &mut ['a', 'b']);
4957 /// assert_eq!(tail, &mut ['c', 'd']);
4958 /// ```
4959 ///
4960 /// Getting `None` when `range` is out of bounds:
4961 ///
4962 /// ```
4963 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c', 'd'];
4964 ///
4965 /// assert_eq!(None, slice.split_off_mut(5..));
4966 /// assert_eq!(None, slice.split_off_mut(..5));
4967 /// assert_eq!(None, slice.split_off_mut(..=4));
4968 /// let expected: &mut [_] = &mut ['a', 'b', 'c', 'd'];
4969 /// assert_eq!(Some(expected), slice.split_off_mut(..4));
4970 /// ```
4971 #[inline]
4972 #[must_use = "method does not modify the slice if the range is out of bounds"]
4973 #[stable(feature = "slice_take", since = "1.87.0")]
4974 pub fn split_off_mut<'a, R: OneSidedRange<usize>>(
4975 self: &mut &'a mut Self,
4976 range: R,
4977 ) -> Option<&'a mut Self> {
4978 let (direction, split_index) = split_point_of(range)?;
4979 if split_index > self.len() {
4980 return None;
4981 }
4982 let (front, back) = mem::take(self).split_at_mut(split_index);
4983 match direction {
4984 Direction::Front => {
4985 *self = back;
4986 Some(front)
4987 }
4988 Direction::Back => {
4989 *self = front;
4990 Some(back)
4991 }
4992 }
4993 }
4994
4995 /// Removes the first element of the slice and returns a reference
4996 /// to it.
4997 ///
4998 /// Returns `None` if the slice is empty.
4999 ///
5000 /// # Examples
5001 ///
5002 /// ```
5003 /// let mut slice: &[_] = &['a', 'b', 'c'];
5004 /// let first = slice.split_off_first().unwrap();
5005 ///
5006 /// assert_eq!(slice, &['b', 'c']);
5007 /// assert_eq!(first, &'a');
5008 /// ```
5009 #[inline]
5010 #[stable(feature = "slice_take", since = "1.87.0")]
5011 #[rustc_const_unstable(feature = "const_split_off_first_last", issue = "138539")]
5012 pub const fn split_off_first<'a>(self: &mut &'a Self) -> Option<&'a T> {
5013 // FIXME(const-hack): Use `?` when available in const instead of `let-else`.
5014 let Some((first, rem)) = self.split_first() else { return None };
5015 *self = rem;
5016 Some(first)
5017 }
5018
5019 /// Removes the first element of the slice and returns a mutable
5020 /// reference to it.
5021 ///
5022 /// Returns `None` if the slice is empty.
5023 ///
5024 /// # Examples
5025 ///
5026 /// ```
5027 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c'];
5028 /// let first = slice.split_off_first_mut().unwrap();
5029 /// *first = 'd';
5030 ///
5031 /// assert_eq!(slice, &['b', 'c']);
5032 /// assert_eq!(first, &'d');
5033 /// ```
5034 #[inline]
5035 #[stable(feature = "slice_take", since = "1.87.0")]
5036 #[rustc_const_unstable(feature = "const_split_off_first_last", issue = "138539")]
5037 pub const fn split_off_first_mut<'a>(self: &mut &'a mut Self) -> Option<&'a mut T> {
5038 // FIXME(const-hack): Use `mem::take` and `?` when available in const.
5039 // Original: `mem::take(self).split_first_mut()?`
5040 let Some((first, rem)) = mem::replace(self, &mut []).split_first_mut() else { return None };
5041 *self = rem;
5042 Some(first)
5043 }
5044
5045 /// Removes the last element of the slice and returns a reference
5046 /// to it.
5047 ///
5048 /// Returns `None` if the slice is empty.
5049 ///
5050 /// # Examples
5051 ///
5052 /// ```
5053 /// let mut slice: &[_] = &['a', 'b', 'c'];
5054 /// let last = slice.split_off_last().unwrap();
5055 ///
5056 /// assert_eq!(slice, &['a', 'b']);
5057 /// assert_eq!(last, &'c');
5058 /// ```
5059 #[inline]
5060 #[stable(feature = "slice_take", since = "1.87.0")]
5061 #[rustc_const_unstable(feature = "const_split_off_first_last", issue = "138539")]
5062 pub const fn split_off_last<'a>(self: &mut &'a Self) -> Option<&'a T> {
5063 // FIXME(const-hack): Use `?` when available in const instead of `let-else`.
5064 let Some((last, rem)) = self.split_last() else { return None };
5065 *self = rem;
5066 Some(last)
5067 }
5068
5069 /// Removes the last element of the slice and returns a mutable
5070 /// reference to it.
5071 ///
5072 /// Returns `None` if the slice is empty.
5073 ///
5074 /// # Examples
5075 ///
5076 /// ```
5077 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c'];
5078 /// let last = slice.split_off_last_mut().unwrap();
5079 /// *last = 'd';
5080 ///
5081 /// assert_eq!(slice, &['a', 'b']);
5082 /// assert_eq!(last, &'d');
5083 /// ```
5084 #[inline]
5085 #[stable(feature = "slice_take", since = "1.87.0")]
5086 #[rustc_const_unstable(feature = "const_split_off_first_last", issue = "138539")]
5087 pub const fn split_off_last_mut<'a>(self: &mut &'a mut Self) -> Option<&'a mut T> {
5088 // FIXME(const-hack): Use `mem::take` and `?` when available in const.
5089 // Original: `mem::take(self).split_last_mut()?`
5090 let Some((last, rem)) = mem::replace(self, &mut []).split_last_mut() else { return None };
5091 *self = rem;
5092 Some(last)
5093 }
5094
5095 /// Returns mutable references to many indices at once, without doing any checks.
5096 ///
5097 /// An index can be either a `usize`, a [`Range`] or a [`RangeInclusive`]. Note
5098 /// that this method takes an array, so all indices must be of the same type.
5099 /// If passed an array of `usize`s this method gives back an array of mutable references
5100 /// to single elements, while if passed an array of ranges it gives back an array of
5101 /// mutable references to slices.
5102 ///
5103 /// For a safe alternative see [`get_disjoint_mut`].
5104 ///
5105 /// # Safety
5106 ///
5107 /// Calling this method with overlapping or out-of-bounds indices is *[undefined behavior]*
5108 /// even if the resulting references are not used.
5109 ///
5110 /// # Examples
5111 ///
5112 /// ```
5113 /// let x = &mut [1, 2, 4];
5114 ///
5115 /// unsafe {
5116 /// let [a, b] = x.get_disjoint_unchecked_mut([0, 2]);
5117 /// *a *= 10;
5118 /// *b *= 100;
5119 /// }
5120 /// assert_eq!(x, &[10, 2, 400]);
5121 ///
5122 /// unsafe {
5123 /// let [a, b] = x.get_disjoint_unchecked_mut([0..1, 1..3]);
5124 /// a[0] = 8;
5125 /// b[0] = 88;
5126 /// b[1] = 888;
5127 /// }
5128 /// assert_eq!(x, &[8, 88, 888]);
5129 ///
5130 /// unsafe {
5131 /// let [a, b] = x.get_disjoint_unchecked_mut([1..=2, 0..=0]);
5132 /// a[0] = 11;
5133 /// a[1] = 111;
5134 /// b[0] = 1;
5135 /// }
5136 /// assert_eq!(x, &[1, 11, 111]);
5137 /// ```
5138 ///
5139 /// [`get_disjoint_mut`]: slice::get_disjoint_mut
5140 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
5141 #[stable(feature = "get_many_mut", since = "1.86.0")]
5142 #[inline]
5143 #[track_caller]
5144 pub unsafe fn get_disjoint_unchecked_mut<I, const N: usize>(
5145 &mut self,
5146 indices: [I; N],
5147 ) -> [&mut I::Output; N]
5148 where
5149 I: GetDisjointMutIndex + SliceIndex<Self>,
5150 {
5151 // NB: This implementation is written as it is because any variation of
5152 // `indices.map(|i| self.get_unchecked_mut(i))` would make miri unhappy,
5153 // or generate worse code otherwise. This is also why we need to go
5154 // through a raw pointer here.
5155 let slice: *mut [T] = self;
5156 let mut arr: MaybeUninit<[&mut I::Output; N]> = MaybeUninit::uninit();
5157 let arr_ptr = arr.as_mut_ptr();
5158
5159 // SAFETY: We expect `indices` to contain disjunct values that are
5160 // in bounds of `self`.
5161 unsafe {
5162 for i in 0..N {
5163 let idx = indices.get_unchecked(i).clone();
5164 arr_ptr.cast::<&mut I::Output>().add(i).write(&mut *slice.get_unchecked_mut(idx));
5165 }
5166 arr.assume_init()
5167 }
5168 }
5169
5170 /// Returns mutable references to many indices at once.
5171 ///
5172 /// An index can be either a `usize`, a [`Range`] or a [`RangeInclusive`]. Note
5173 /// that this method takes an array, so all indices must be of the same type.
5174 /// If passed an array of `usize`s this method gives back an array of mutable references
5175 /// to single elements, while if passed an array of ranges it gives back an array of
5176 /// mutable references to slices.
5177 ///
5178 /// Returns an error if any index is out-of-bounds, or if there are overlapping indices.
5179 /// An empty range is not considered to overlap if it is located at the beginning or at
5180 /// the end of another range, but is considered to overlap if it is located in the middle.
5181 ///
5182 /// This method does a O(n^2) check to check that there are no overlapping indices, so be careful
5183 /// when passing many indices.
5184 ///
5185 /// # Examples
5186 ///
5187 /// ```
5188 /// let v = &mut [1, 2, 3];
5189 /// if let Ok([a, b]) = v.get_disjoint_mut([0, 2]) {
5190 /// *a = 413;
5191 /// *b = 612;
5192 /// }
5193 /// assert_eq!(v, &[413, 2, 612]);
5194 ///
5195 /// if let Ok([a, b]) = v.get_disjoint_mut([0..1, 1..3]) {
5196 /// a[0] = 8;
5197 /// b[0] = 88;
5198 /// b[1] = 888;
5199 /// }
5200 /// assert_eq!(v, &[8, 88, 888]);
5201 ///
5202 /// if let Ok([a, b]) = v.get_disjoint_mut([1..=2, 0..=0]) {
5203 /// a[0] = 11;
5204 /// a[1] = 111;
5205 /// b[0] = 1;
5206 /// }
5207 /// assert_eq!(v, &[1, 11, 111]);
5208 /// ```
5209 #[stable(feature = "get_many_mut", since = "1.86.0")]
5210 #[inline]
5211 pub fn get_disjoint_mut<I, const N: usize>(
5212 &mut self,
5213 indices: [I; N],
5214 ) -> Result<[&mut I::Output; N], GetDisjointMutError>
5215 where
5216 I: GetDisjointMutIndex + SliceIndex<Self>,
5217 {
5218 get_disjoint_check_valid(&indices, self.len())?;
5219 // SAFETY: The `get_disjoint_check_valid()` call checked that all indices
5220 // are disjunct and in bounds.
5221 unsafe { Ok(self.get_disjoint_unchecked_mut(indices)) }
5222 }
5223
5224 /// Returns the index that an element reference points to.
5225 ///
5226 /// Returns `None` if `element` does not point to the start of an element within the slice.
5227 ///
5228 /// This method is useful for extending slice iterators like [`slice::split`].
5229 ///
5230 /// Note that this uses pointer arithmetic and **does not compare elements**.
5231 /// To find the index of an element via comparison, use
5232 /// [`.iter().position()`](crate::iter::Iterator::position) instead.
5233 ///
5234 /// # Panics
5235 /// Panics if `T` is zero-sized.
5236 ///
5237 /// # Examples
5238 /// Basic usage:
5239 /// ```
5240 /// let nums: &[u32] = &[1, 7, 1, 1];
5241 /// let num = &nums[2];
5242 ///
5243 /// assert_eq!(num, &1);
5244 /// assert_eq!(nums.element_offset(num), Some(2));
5245 /// ```
5246 /// Returning `None` with an unaligned element:
5247 /// ```
5248 /// let arr: &[[u32; 2]] = &[[0, 1], [2, 3]];
5249 /// let flat_arr: &[u32] = arr.as_flattened();
5250 ///
5251 /// let ok_elm: &[u32; 2] = flat_arr[0..2].try_into().unwrap();
5252 /// let weird_elm: &[u32; 2] = flat_arr[1..3].try_into().unwrap();
5253 ///
5254 /// assert_eq!(ok_elm, &[0, 1]);
5255 /// assert_eq!(weird_elm, &[1, 2]);
5256 ///
5257 /// assert_eq!(arr.element_offset(ok_elm), Some(0)); // Points to element 0
5258 /// assert_eq!(arr.element_offset(weird_elm), None); // Points between element 0 and 1
5259 /// ```
5260 #[must_use]
5261 #[stable(feature = "element_offset", since = "1.94.0")]
5262 pub fn element_offset(&self, element: &T) -> Option<usize> {
5263 if T::IS_ZST {
5264 panic!("elements are zero-sized");
5265 }
5266
5267 let self_start = self.as_ptr().addr();
5268 let elem_start = ptr::from_ref(element).addr();
5269
5270 let byte_offset = elem_start.wrapping_sub(self_start);
5271
5272 if !byte_offset.is_multiple_of(size_of::<T>()) {
5273 return None;
5274 }
5275
5276 let offset = byte_offset / size_of::<T>();
5277
5278 if offset < self.len() { Some(offset) } else { None }
5279 }
5280
5281 /// Returns the range of indices that a subslice points to.
5282 ///
5283 /// Returns `None` if `subslice` does not point within the slice or if it is not aligned with the
5284 /// elements in the slice.
5285 ///
5286 /// This method **does not compare elements**. Instead, this method finds the location in the slice that
5287 /// `subslice` was obtained from. To find the index of a subslice via comparison, instead use
5288 /// [`.windows()`](slice::windows)[`.position()`](crate::iter::Iterator::position).
5289 ///
5290 /// This method is useful for extending slice iterators like [`slice::split`].
5291 ///
5292 /// Note that this may return a false positive (either `Some(0..0)` or `Some(self.len()..self.len())`)
5293 /// if `subslice` has a length of zero and points to the beginning or end of another, separate, slice.
5294 ///
5295 /// # Panics
5296 /// Panics if `T` is zero-sized.
5297 ///
5298 /// # Examples
5299 /// Basic usage:
5300 /// ```
5301 /// use core::range::Range;
5302 ///
5303 /// let nums = &[0, 5, 10, 0, 0, 5];
5304 ///
5305 /// let mut iter = nums
5306 /// .split(|t| *t == 0)
5307 /// .map(|n| nums.subslice_range(n).unwrap());
5308 ///
5309 /// assert_eq!(iter.next(), Some(Range { start: 0, end: 0 }));
5310 /// assert_eq!(iter.next(), Some(Range { start: 1, end: 3 }));
5311 /// assert_eq!(iter.next(), Some(Range { start: 4, end: 4 }));
5312 /// assert_eq!(iter.next(), Some(Range { start: 5, end: 6 }));
5313 /// ```
5314 #[must_use]
5315 #[stable(feature = "substr_range", since = "CURRENT_RUSTC_VERSION")]
5316 pub fn subslice_range(&self, subslice: &[T]) -> Option<core::range::Range<usize>> {
5317 if T::IS_ZST {
5318 panic!("elements are zero-sized");
5319 }
5320
5321 let self_start = self.as_ptr().addr();
5322 let subslice_start = subslice.as_ptr().addr();
5323
5324 let byte_start = subslice_start.wrapping_sub(self_start);
5325
5326 if !byte_start.is_multiple_of(size_of::<T>()) {
5327 return None;
5328 }
5329
5330 let start = byte_start / size_of::<T>();
5331 let end = start.wrapping_add(subslice.len());
5332
5333 if start <= self.len() && end <= self.len() {
5334 Some(core::range::Range { start, end })
5335 } else {
5336 None
5337 }
5338 }
5339
5340 /// Returns the same slice `&[T]`.
5341 ///
5342 /// This method is redundant when used directly on `&[T]`, but
5343 /// it helps dereferencing other "container" types to slices,
5344 /// for example `Box<[T]>` or `Arc<[T]>`.
5345 #[inline]
5346 #[unstable(feature = "str_as_str", issue = "130366")]
5347 pub const fn as_slice(&self) -> &[T] {
5348 self
5349 }
5350
5351 /// Returns the same slice `&mut [T]`.
5352 ///
5353 /// This method is redundant when used directly on `&mut [T]`, but
5354 /// it helps dereferencing other "container" types to slices,
5355 /// for example `Box<[T]>` or `MutexGuard<[T]>`.
5356 #[inline]
5357 #[unstable(feature = "str_as_str", issue = "130366")]
5358 pub const fn as_mut_slice(&mut self) -> &mut [T] {
5359 self
5360 }
5361}
5362
5363impl<T> [MaybeUninit<T>] {
5364 /// Transmutes the mutable uninitialized slice to a mutable uninitialized slice of
5365 /// another type, ensuring alignment of the types is maintained.
5366 ///
5367 /// This is a safe wrapper around [`slice::align_to_mut`], so inherits the same
5368 /// guarantees as that method.
5369 ///
5370 /// # Examples
5371 ///
5372 /// ```
5373 /// #![feature(align_to_uninit_mut)]
5374 /// use std::mem::MaybeUninit;
5375 ///
5376 /// pub struct BumpAllocator<'scope> {
5377 /// memory: &'scope mut [MaybeUninit<u8>],
5378 /// }
5379 ///
5380 /// impl<'scope> BumpAllocator<'scope> {
5381 /// pub fn new(memory: &'scope mut [MaybeUninit<u8>]) -> Self {
5382 /// Self { memory }
5383 /// }
5384 /// pub fn try_alloc_uninit<T>(&mut self) -> Option<&'scope mut MaybeUninit<T>> {
5385 /// let first_end = self.memory.as_ptr().align_offset(align_of::<T>()) + size_of::<T>();
5386 /// let prefix = self.memory.split_off_mut(..first_end)?;
5387 /// Some(&mut prefix.align_to_uninit_mut::<T>().1[0])
5388 /// }
5389 /// pub fn try_alloc_u32(&mut self, value: u32) -> Option<&'scope mut u32> {
5390 /// let uninit = self.try_alloc_uninit()?;
5391 /// Some(uninit.write(value))
5392 /// }
5393 /// }
5394 ///
5395 /// let mut memory = [MaybeUninit::<u8>::uninit(); 10];
5396 /// let mut allocator = BumpAllocator::new(&mut memory);
5397 /// let v = allocator.try_alloc_u32(42);
5398 /// assert_eq!(v, Some(&mut 42));
5399 /// ```
5400 #[unstable(feature = "align_to_uninit_mut", issue = "139062")]
5401 #[inline]
5402 #[must_use]
5403 pub fn align_to_uninit_mut<U>(&mut self) -> (&mut Self, &mut [MaybeUninit<U>], &mut Self) {
5404 // SAFETY: `MaybeUninit` is transparent. Correct size and alignment are guaranteed by
5405 // `align_to_mut` itself. Therefore the only thing that we have to ensure for a safe
5406 // `transmute` is that the values are valid for the types involved. But for `MaybeUninit`
5407 // any values are valid, so this operation is safe.
5408 unsafe { self.align_to_mut() }
5409 }
5410}
5411
5412impl<T, const N: usize> [[T; N]] {
5413 /// Takes a `&[[T; N]]`, and flattens it to a `&[T]`.
5414 ///
5415 /// For the opposite operation, see [`as_chunks`] and [`as_rchunks`].
5416 ///
5417 /// [`as_chunks`]: slice::as_chunks
5418 /// [`as_rchunks`]: slice::as_rchunks
5419 ///
5420 /// # Panics
5421 ///
5422 /// This panics if the length of the resulting slice would overflow a `usize`.
5423 ///
5424 /// This is only possible when flattening a slice of arrays of zero-sized
5425 /// types, and thus tends to be irrelevant in practice. If
5426 /// `size_of::<T>() > 0`, this will never panic.
5427 ///
5428 /// # Examples
5429 ///
5430 /// ```
5431 /// assert_eq!([[1, 2, 3], [4, 5, 6]].as_flattened(), &[1, 2, 3, 4, 5, 6]);
5432 ///
5433 /// assert_eq!(
5434 /// [[1, 2, 3], [4, 5, 6]].as_flattened(),
5435 /// [[1, 2], [3, 4], [5, 6]].as_flattened(),
5436 /// );
5437 ///
5438 /// let slice_of_empty_arrays: &[[i32; 0]] = &[[], [], [], [], []];
5439 /// assert!(slice_of_empty_arrays.as_flattened().is_empty());
5440 ///
5441 /// let empty_slice_of_arrays: &[[u32; 10]] = &[];
5442 /// assert!(empty_slice_of_arrays.as_flattened().is_empty());
5443 /// ```
5444 #[stable(feature = "slice_flatten", since = "1.80.0")]
5445 #[rustc_const_stable(feature = "const_slice_flatten", since = "1.87.0")]
5446 pub const fn as_flattened(&self) -> &[T] {
5447 let len = if T::IS_ZST {
5448 self.len().checked_mul(N).expect("slice len overflow")
5449 } else {
5450 // SAFETY: `self.len() * N` cannot overflow because `self` is
5451 // already in the address space.
5452 unsafe { self.len().unchecked_mul(N) }
5453 };
5454 // SAFETY: `[T]` is layout-identical to `[T; N]`
5455 unsafe { from_raw_parts(self.as_ptr().cast(), len) }
5456 }
5457
5458 /// Takes a `&mut [[T; N]]`, and flattens it to a `&mut [T]`.
5459 ///
5460 /// For the opposite operation, see [`as_chunks_mut`] and [`as_rchunks_mut`].
5461 ///
5462 /// [`as_chunks_mut`]: slice::as_chunks_mut
5463 /// [`as_rchunks_mut`]: slice::as_rchunks_mut
5464 ///
5465 /// # Panics
5466 ///
5467 /// This panics if the length of the resulting slice would overflow a `usize`.
5468 ///
5469 /// This is only possible when flattening a slice of arrays of zero-sized
5470 /// types, and thus tends to be irrelevant in practice. If
5471 /// `size_of::<T>() > 0`, this will never panic.
5472 ///
5473 /// # Examples
5474 ///
5475 /// ```
5476 /// fn add_5_to_all(slice: &mut [i32]) {
5477 /// for i in slice {
5478 /// *i += 5;
5479 /// }
5480 /// }
5481 ///
5482 /// let mut array = [[1, 2, 3], [4, 5, 6], [7, 8, 9]];
5483 /// add_5_to_all(array.as_flattened_mut());
5484 /// assert_eq!(array, [[6, 7, 8], [9, 10, 11], [12, 13, 14]]);
5485 /// ```
5486 #[stable(feature = "slice_flatten", since = "1.80.0")]
5487 #[rustc_const_stable(feature = "const_slice_flatten", since = "1.87.0")]
5488 pub const fn as_flattened_mut(&mut self) -> &mut [T] {
5489 let len = if T::IS_ZST {
5490 self.len().checked_mul(N).expect("slice len overflow")
5491 } else {
5492 // SAFETY: `self.len() * N` cannot overflow because `self` is
5493 // already in the address space.
5494 unsafe { self.len().unchecked_mul(N) }
5495 };
5496 // SAFETY: `[T]` is layout-identical to `[T; N]`
5497 unsafe { from_raw_parts_mut(self.as_mut_ptr().cast(), len) }
5498 }
5499}
5500
5501impl [f32] {
5502 /// Sorts the slice of floats.
5503 ///
5504 /// This sort is in-place (i.e. does not allocate), *O*(*n* \* log(*n*)) worst-case, and uses
5505 /// the ordering defined by [`f32::total_cmp`].
5506 ///
5507 /// # Current implementation
5508 ///
5509 /// This uses the same sorting algorithm as [`sort_unstable_by`](slice::sort_unstable_by).
5510 ///
5511 /// # Examples
5512 ///
5513 /// ```
5514 /// #![feature(sort_floats)]
5515 /// let mut v = [2.6, -5e-8, f32::NAN, 8.29, f32::INFINITY, -1.0, 0.0, -f32::INFINITY, -0.0];
5516 ///
5517 /// v.sort_floats();
5518 /// let sorted = [-f32::INFINITY, -1.0, -5e-8, -0.0, 0.0, 2.6, 8.29, f32::INFINITY, f32::NAN];
5519 /// assert_eq!(&v[..8], &sorted[..8]);
5520 /// assert!(v[8].is_nan());
5521 /// ```
5522 #[unstable(feature = "sort_floats", issue = "93396")]
5523 #[inline]
5524 pub fn sort_floats(&mut self) {
5525 self.sort_unstable_by(f32::total_cmp);
5526 }
5527}
5528
5529impl [f64] {
5530 /// Sorts the slice of floats.
5531 ///
5532 /// This sort is in-place (i.e. does not allocate), *O*(*n* \* log(*n*)) worst-case, and uses
5533 /// the ordering defined by [`f64::total_cmp`].
5534 ///
5535 /// # Current implementation
5536 ///
5537 /// This uses the same sorting algorithm as [`sort_unstable_by`](slice::sort_unstable_by).
5538 ///
5539 /// # Examples
5540 ///
5541 /// ```
5542 /// #![feature(sort_floats)]
5543 /// let mut v = [2.6, -5e-8, f64::NAN, 8.29, f64::INFINITY, -1.0, 0.0, -f64::INFINITY, -0.0];
5544 ///
5545 /// v.sort_floats();
5546 /// let sorted = [-f64::INFINITY, -1.0, -5e-8, -0.0, 0.0, 2.6, 8.29, f64::INFINITY, f64::NAN];
5547 /// assert_eq!(&v[..8], &sorted[..8]);
5548 /// assert!(v[8].is_nan());
5549 /// ```
5550 #[unstable(feature = "sort_floats", issue = "93396")]
5551 #[inline]
5552 pub fn sort_floats(&mut self) {
5553 self.sort_unstable_by(f64::total_cmp);
5554 }
5555}
5556
5557/// Copies `src` to `dest`.
5558///
5559/// # Safety
5560/// `T` must implement one of `Copy` or `TrivialClone`.
5561#[track_caller]
5562const unsafe fn copy_from_slice_impl<T: Clone>(dest: &mut [T], src: &[T]) {
5563 // The panic code path was put into a cold function to not bloat the
5564 // call site.
5565 #[cfg_attr(not(panic = "immediate-abort"), inline(never), cold)]
5566 #[cfg_attr(panic = "immediate-abort", inline)]
5567 #[track_caller]
5568 const fn len_mismatch_fail(dst_len: usize, src_len: usize) -> ! {
5569 const_panic!(
5570 "copy_from_slice: source slice length does not match destination slice length",
5571 "copy_from_slice: source slice length ({src_len}) does not match destination slice length ({dst_len})",
5572 src_len: usize,
5573 dst_len: usize,
5574 )
5575 }
5576
5577 if dest.len() != src.len() {
5578 len_mismatch_fail(dest.len(), src.len());
5579 }
5580
5581 // SAFETY: `self` is valid for `self.len()` elements by definition, and `src` was
5582 // checked to have the same length. The slices cannot overlap because
5583 // mutable references are exclusive.
5584 unsafe {
5585 ptr::copy_nonoverlapping(src.as_ptr(), dest.as_mut_ptr(), dest.len());
5586 }
5587}
5588
5589#[rustc_const_unstable(feature = "const_clone", issue = "142757")]
5590const trait CloneFromSpec<T> {
5591 fn spec_clone_from(&mut self, src: &[T])
5592 where
5593 T: [const] Destruct;
5594}
5595
5596#[rustc_const_unstable(feature = "const_clone", issue = "142757")]
5597const impl<T> CloneFromSpec<T> for [T]
5598where
5599 T: [const] Clone + [const] Destruct,
5600{
5601 #[track_caller]
5602 default fn spec_clone_from(&mut self, src: &[T]) {
5603 assert!(self.len() == src.len(), "destination and source slices have different lengths");
5604 // NOTE: We need to explicitly slice them to the same length
5605 // to make it easier for the optimizer to elide bounds checking.
5606 // But since it can't be relied on we also have an explicit specialization for T: Copy.
5607 let len = self.len();
5608 let src = &src[..len];
5609 // FIXME(const_hack): make this a `for idx in 0..self.len()` loop.
5610 let mut idx = 0;
5611 while idx < self.len() {
5612 self[idx].clone_from(&src[idx]);
5613 idx += 1;
5614 }
5615 }
5616}
5617
5618#[rustc_const_unstable(feature = "const_clone", issue = "142757")]
5619const impl<T> CloneFromSpec<T> for [T]
5620where
5621 T: [const] TrivialClone + [const] Destruct,
5622{
5623 #[track_caller]
5624 fn spec_clone_from(&mut self, src: &[T]) {
5625 // SAFETY: `T` implements `TrivialClone`.
5626 unsafe {
5627 copy_from_slice_impl(self, src);
5628 }
5629 }
5630}
5631
5632#[stable(feature = "rust1", since = "1.0.0")]
5633#[rustc_const_unstable(feature = "const_default", issue = "143894")]
5634const impl<T> Default for &[T] {
5635 /// Creates an empty slice.
5636 fn default() -> Self {
5637 &[]
5638 }
5639}
5640
5641#[stable(feature = "mut_slice_default", since = "1.5.0")]
5642#[rustc_const_unstable(feature = "const_default", issue = "143894")]
5643const impl<T> Default for &mut [T] {
5644 /// Creates a mutable empty slice.
5645 fn default() -> Self {
5646 &mut []
5647 }
5648}
5649
5650#[unstable(feature = "slice_pattern", reason = "stopgap trait for slice patterns", issue = "56345")]
5651/// Patterns in slices - currently, only used by `strip_prefix` and `strip_suffix`. At a future
5652/// point, we hope to generalise `core::str::Pattern` (which at the time of writing is limited to
5653/// `str`) to slices, and then this trait will be replaced or abolished.
5654pub trait SlicePattern {
5655 /// The element type of the slice being matched on.
5656 type Item;
5657
5658 /// Currently, the consumers of `SlicePattern` need a slice.
5659 fn as_slice(&self) -> &[Self::Item];
5660}
5661
5662#[stable(feature = "slice_strip", since = "1.51.0")]
5663impl<T> SlicePattern for [T] {
5664 type Item = T;
5665
5666 #[inline]
5667 fn as_slice(&self) -> &[Self::Item] {
5668 self
5669 }
5670}
5671
5672#[stable(feature = "slice_strip", since = "1.51.0")]
5673impl<T, const N: usize> SlicePattern for [T; N] {
5674 type Item = T;
5675
5676 #[inline]
5677 fn as_slice(&self) -> &[Self::Item] {
5678 self
5679 }
5680}
5681
5682/// This checks every index against each other, and against `len`.
5683///
5684/// This will do `binomial(N + 1, 2) = N * (N + 1) / 2 = 0, 1, 3, 6, 10, ..`
5685/// comparison operations.
5686#[inline]
5687fn get_disjoint_check_valid<I: GetDisjointMutIndex, const N: usize>(
5688 indices: &[I; N],
5689 len: usize,
5690) -> Result<(), GetDisjointMutError> {
5691 // NB: The optimizer should inline the loops into a sequence
5692 // of instructions without additional branching.
5693 for (i, idx) in indices.iter().enumerate() {
5694 if !idx.is_in_bounds(len) {
5695 return Err(GetDisjointMutError::IndexOutOfBounds);
5696 }
5697 for idx2 in &indices[..i] {
5698 if idx.is_overlapping(idx2) {
5699 return Err(GetDisjointMutError::OverlappingIndices);
5700 }
5701 }
5702 }
5703 Ok(())
5704}
5705
5706/// The error type returned by [`get_disjoint_mut`][`slice::get_disjoint_mut`].
5707///
5708/// It indicates one of two possible errors:
5709/// - An index is out-of-bounds.
5710/// - The same index appeared multiple times in the array
5711/// (or different but overlapping indices when ranges are provided).
5712///
5713/// # Examples
5714///
5715/// ```
5716/// use std::slice::GetDisjointMutError;
5717///
5718/// let v = &mut [1, 2, 3];
5719/// assert_eq!(v.get_disjoint_mut([0, 999]), Err(GetDisjointMutError::IndexOutOfBounds));
5720/// assert_eq!(v.get_disjoint_mut([1, 1]), Err(GetDisjointMutError::OverlappingIndices));
5721/// ```
5722#[stable(feature = "get_many_mut", since = "1.86.0")]
5723#[derive(Debug, Clone, PartialEq, Eq)]
5724pub enum GetDisjointMutError {
5725 /// An index provided was out-of-bounds for the slice.
5726 IndexOutOfBounds,
5727 /// Two indices provided were overlapping.
5728 OverlappingIndices,
5729}
5730
5731#[stable(feature = "get_many_mut", since = "1.86.0")]
5732impl fmt::Display for GetDisjointMutError {
5733 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
5734 let msg = match self {
5735 GetDisjointMutError::IndexOutOfBounds => "an index is out of bounds",
5736 GetDisjointMutError::OverlappingIndices => "there were overlapping indices",
5737 };
5738 fmt::Display::fmt(msg, f)
5739 }
5740}
5741
5742/// A helper trait for `<[T]>::get_disjoint_mut()`.
5743///
5744/// # Safety
5745///
5746/// If `is_in_bounds()` returns `true` and `is_overlapping()` returns `false`,
5747/// it must be safe to index the slice with the indices.
5748#[unstable(feature = "get_disjoint_mut_helpers", issue = "none")]
5749pub impl(self) unsafe trait GetDisjointMutIndex: Clone {
5750 /// Returns `true` if `self` is in bounds for `len` slice elements.
5751 #[unstable(feature = "get_disjoint_mut_helpers", issue = "none")]
5752 fn is_in_bounds(&self, len: usize) -> bool;
5753
5754 /// Returns `true` if `self` overlaps with `other`.
5755 ///
5756 /// Note that we don't consider zero-length ranges to overlap at the beginning or the end,
5757 /// but do consider them to overlap in the middle.
5758 #[unstable(feature = "get_disjoint_mut_helpers", issue = "none")]
5759 fn is_overlapping(&self, other: &Self) -> bool;
5760}
5761
5762#[unstable(feature = "get_disjoint_mut_helpers", issue = "none")]
5763// SAFETY: We implement `is_in_bounds()` and `is_overlapping()` correctly.
5764unsafe impl GetDisjointMutIndex for usize {
5765 #[inline]
5766 fn is_in_bounds(&self, len: usize) -> bool {
5767 *self < len
5768 }
5769
5770 #[inline]
5771 fn is_overlapping(&self, other: &Self) -> bool {
5772 *self == *other
5773 }
5774}
5775
5776#[unstable(feature = "get_disjoint_mut_helpers", issue = "none")]
5777// SAFETY: We implement `is_in_bounds()` and `is_overlapping()` correctly.
5778unsafe impl GetDisjointMutIndex for Range<usize> {
5779 #[inline]
5780 fn is_in_bounds(&self, len: usize) -> bool {
5781 (self.start <= self.end) & (self.end <= len)
5782 }
5783
5784 #[inline]
5785 fn is_overlapping(&self, other: &Self) -> bool {
5786 (self.start < other.end) & (other.start < self.end)
5787 }
5788}
5789
5790#[unstable(feature = "get_disjoint_mut_helpers", issue = "none")]
5791// SAFETY: We implement `is_in_bounds()` and `is_overlapping()` correctly.
5792unsafe impl GetDisjointMutIndex for RangeInclusive<usize> {
5793 #[inline]
5794 fn is_in_bounds(&self, len: usize) -> bool {
5795 (self.start <= self.end) & (self.end < len)
5796 }
5797
5798 #[inline]
5799 fn is_overlapping(&self, other: &Self) -> bool {
5800 (self.start <= other.end) & (other.start <= self.end)
5801 }
5802}
5803
5804#[unstable(feature = "get_disjoint_mut_helpers", issue = "none")]
5805// SAFETY: We implement `is_in_bounds()` and `is_overlapping()` correctly.
5806unsafe impl GetDisjointMutIndex for range::Range<usize> {
5807 #[inline]
5808 fn is_in_bounds(&self, len: usize) -> bool {
5809 Range::from(*self).is_in_bounds(len)
5810 }
5811
5812 #[inline]
5813 fn is_overlapping(&self, other: &Self) -> bool {
5814 Range::from(*self).is_overlapping(&Range::from(*other))
5815 }
5816}
5817
5818#[unstable(feature = "get_disjoint_mut_helpers", issue = "none")]
5819// SAFETY: We implement `is_in_bounds()` and `is_overlapping()` correctly.
5820unsafe impl GetDisjointMutIndex for range::RangeInclusive<usize> {
5821 #[inline]
5822 fn is_in_bounds(&self, len: usize) -> bool {
5823 RangeInclusive::from(*self).is_in_bounds(len)
5824 }
5825
5826 #[inline]
5827 fn is_overlapping(&self, other: &Self) -> bool {
5828 RangeInclusive::from(*self).is_overlapping(&RangeInclusive::from(*other))
5829 }
5830}