1 //! Slice management and manipulation.
3 //! For more details see [`std::slice`].
5 //! [`std::slice`]: ../../std/slice/index.html
7 #![stable(feature = "rust1", since = "1.0.0")]
9 use crate::cmp::Ordering::{self, Greater, Less};
10 use crate::intrinsics::{assert_unsafe_precondition, exact_div};
11 use crate::marker::Copy;
12 use crate::mem::{self, SizedTypeProperties};
13 use crate::num::NonZeroUsize;
14 use crate::ops::{Bound, FnMut, OneSidedRange, Range, RangeBounds};
15 use crate::option::Option;
16 use crate::option::Option::{None, Some};
18 use crate::result::Result;
19 use crate::result::Result::{Err, Ok};
20 use crate::simd::{self, Simd};
24 feature = "slice_internals",
26 reason = "exposed from core to be reused in std; use the memchr crate"
28 /// Pure rust memchr implementation, taken from rust-memchr
40 #[stable(feature = "rust1", since = "1.0.0")]
41 pub use iter::{Chunks, ChunksMut, Windows};
42 #[stable(feature = "rust1", since = "1.0.0")]
43 pub use iter::{Iter, IterMut};
44 #[stable(feature = "rust1", since = "1.0.0")]
45 pub use iter::{RSplitN, RSplitNMut, Split, SplitMut, SplitN, SplitNMut};
47 #[stable(feature = "slice_rsplit", since = "1.27.0")]
48 pub use iter::{RSplit, RSplitMut};
50 #[stable(feature = "chunks_exact", since = "1.31.0")]
51 pub use iter::{ChunksExact, ChunksExactMut};
53 #[stable(feature = "rchunks", since = "1.31.0")]
54 pub use iter::{RChunks, RChunksExact, RChunksExactMut, RChunksMut};
56 #[unstable(feature = "array_chunks", issue = "74985")]
57 pub use iter::{ArrayChunks, ArrayChunksMut};
59 #[unstable(feature = "array_windows", issue = "75027")]
60 pub use iter::ArrayWindows;
62 #[unstable(feature = "slice_group_by", issue = "80552")]
63 pub use iter::{GroupBy, GroupByMut};
65 #[stable(feature = "split_inclusive", since = "1.51.0")]
66 pub use iter::{SplitInclusive, SplitInclusiveMut};
68 #[stable(feature = "rust1", since = "1.0.0")]
69 pub use raw::{from_raw_parts, from_raw_parts_mut};
71 #[stable(feature = "from_ref", since = "1.28.0")]
72 pub use raw::{from_mut, from_ref};
74 #[unstable(feature = "slice_from_ptr_range", issue = "89792")]
75 pub use raw::{from_mut_ptr_range, from_ptr_range};
77 // This function is public only because there is no other way to unit test heapsort.
78 #[unstable(feature = "sort_internals", reason = "internal to sort module", issue = "none")]
79 pub use sort::heapsort;
81 #[stable(feature = "slice_get_slice", since = "1.28.0")]
82 pub use index::SliceIndex;
84 #[unstable(feature = "slice_range", issue = "76393")]
87 #[stable(feature = "inherent_ascii_escape", since = "1.60.0")]
88 pub use ascii::EscapeAscii;
90 /// Calculates the direction and split point of a one-sided range.
92 /// This is a helper function for `take` and `take_mut` that returns
93 /// the direction of the split (front or back) as well as the index at
94 /// which to split. Returns `None` if the split index would overflow.
96 fn split_point_of(range: impl OneSidedRange<usize>) -> Option<(Direction, usize)> {
99 Some(match (range.start_bound(), range.end_bound()) {
100 (Unbounded, Excluded(i)) => (Direction::Front, *i),
101 (Unbounded, Included(i)) => (Direction::Front, i.checked_add(1)?),
102 (Excluded(i), Unbounded) => (Direction::Back, i.checked_add(1)?),
103 (Included(i), Unbounded) => (Direction::Back, *i),
115 /// Returns the number of elements in the slice.
120 /// let a = [1, 2, 3];
121 /// assert_eq!(a.len(), 3);
123 #[lang = "slice_len_fn"]
124 #[stable(feature = "rust1", since = "1.0.0")]
125 #[rustc_const_stable(feature = "const_slice_len", since = "1.39.0")]
128 // SAFETY: const sound because we transmute out the length field as a usize (which it must be)
129 pub const fn len(&self) -> usize {
130 // FIXME: Replace with `crate::ptr::metadata(self)` when that is const-stable.
131 // As of this writing this causes a "Const-stable functions can only call other
132 // const-stable functions" error.
134 // SAFETY: Accessing the value from the `PtrRepr` union is safe since *const T
135 // and PtrComponents<T> have the same memory layouts. Only std can make this
137 unsafe { crate::ptr::PtrRepr { const_ptr: self }.components.metadata }
140 /// Returns `true` if the slice has a length of 0.
145 /// let a = [1, 2, 3];
146 /// assert!(!a.is_empty());
148 #[stable(feature = "rust1", since = "1.0.0")]
149 #[rustc_const_stable(feature = "const_slice_is_empty", since = "1.39.0")]
152 pub const fn is_empty(&self) -> bool {
156 /// Returns the first element of the slice, or `None` if it is empty.
161 /// let v = [10, 40, 30];
162 /// assert_eq!(Some(&10), v.first());
164 /// let w: &[i32] = &[];
165 /// assert_eq!(None, w.first());
167 #[stable(feature = "rust1", since = "1.0.0")]
168 #[rustc_const_stable(feature = "const_slice_first_last_not_mut", since = "1.56.0")]
171 pub const fn first(&self) -> Option<&T> {
172 if let [first, ..] = self { Some(first) } else { None }
175 /// Returns a mutable pointer to the first element of the slice, or `None` if it is empty.
180 /// let x = &mut [0, 1, 2];
182 /// if let Some(first) = x.first_mut() {
185 /// assert_eq!(x, &[5, 1, 2]);
187 #[stable(feature = "rust1", since = "1.0.0")]
188 #[rustc_const_unstable(feature = "const_slice_first_last", issue = "83570")]
191 pub const fn first_mut(&mut self) -> Option<&mut T> {
192 if let [first, ..] = self { Some(first) } else { None }
195 /// Returns the first and all the rest of the elements of the slice, or `None` if it is empty.
200 /// let x = &[0, 1, 2];
202 /// if let Some((first, elements)) = x.split_first() {
203 /// assert_eq!(first, &0);
204 /// assert_eq!(elements, &[1, 2]);
207 #[stable(feature = "slice_splits", since = "1.5.0")]
208 #[rustc_const_stable(feature = "const_slice_first_last_not_mut", since = "1.56.0")]
211 pub const fn split_first(&self) -> Option<(&T, &[T])> {
212 if let [first, tail @ ..] = self { Some((first, tail)) } else { None }
215 /// Returns the first and all the rest of the elements of the slice, or `None` if it is empty.
220 /// let x = &mut [0, 1, 2];
222 /// if let Some((first, elements)) = x.split_first_mut() {
227 /// assert_eq!(x, &[3, 4, 5]);
229 #[stable(feature = "slice_splits", since = "1.5.0")]
230 #[rustc_const_unstable(feature = "const_slice_first_last", issue = "83570")]
233 pub const fn split_first_mut(&mut self) -> Option<(&mut T, &mut [T])> {
234 if let [first, tail @ ..] = self { Some((first, tail)) } else { None }
237 /// Returns the last and all the rest of the elements of the slice, or `None` if it is empty.
242 /// let x = &[0, 1, 2];
244 /// if let Some((last, elements)) = x.split_last() {
245 /// assert_eq!(last, &2);
246 /// assert_eq!(elements, &[0, 1]);
249 #[stable(feature = "slice_splits", since = "1.5.0")]
250 #[rustc_const_stable(feature = "const_slice_first_last_not_mut", since = "1.56.0")]
253 pub const fn split_last(&self) -> Option<(&T, &[T])> {
254 if let [init @ .., last] = self { Some((last, init)) } else { None }
257 /// Returns the last and all the rest of the elements of the slice, or `None` if it is empty.
262 /// let x = &mut [0, 1, 2];
264 /// if let Some((last, elements)) = x.split_last_mut() {
269 /// assert_eq!(x, &[4, 5, 3]);
271 #[stable(feature = "slice_splits", since = "1.5.0")]
272 #[rustc_const_unstable(feature = "const_slice_first_last", issue = "83570")]
275 pub const fn split_last_mut(&mut self) -> Option<(&mut T, &mut [T])> {
276 if let [init @ .., last] = self { Some((last, init)) } else { None }
279 /// Returns the last element of the slice, or `None` if it is empty.
284 /// let v = [10, 40, 30];
285 /// assert_eq!(Some(&30), v.last());
287 /// let w: &[i32] = &[];
288 /// assert_eq!(None, w.last());
290 #[stable(feature = "rust1", since = "1.0.0")]
291 #[rustc_const_stable(feature = "const_slice_first_last_not_mut", since = "1.56.0")]
294 pub const fn last(&self) -> Option<&T> {
295 if let [.., last] = self { Some(last) } else { None }
298 /// Returns a mutable pointer to the last item in the slice.
303 /// let x = &mut [0, 1, 2];
305 /// if let Some(last) = x.last_mut() {
308 /// assert_eq!(x, &[0, 1, 10]);
310 #[stable(feature = "rust1", since = "1.0.0")]
311 #[rustc_const_unstable(feature = "const_slice_first_last", issue = "83570")]
314 pub const fn last_mut(&mut self) -> Option<&mut T> {
315 if let [.., last] = self { Some(last) } else { None }
318 /// Returns a reference to an element or subslice depending on the type of
321 /// - If given a position, returns a reference to the element at that
322 /// position or `None` if out of bounds.
323 /// - If given a range, returns the subslice corresponding to that range,
324 /// or `None` if out of bounds.
329 /// let v = [10, 40, 30];
330 /// assert_eq!(Some(&40), v.get(1));
331 /// assert_eq!(Some(&[10, 40][..]), v.get(0..2));
332 /// assert_eq!(None, v.get(3));
333 /// assert_eq!(None, v.get(0..4));
335 #[stable(feature = "rust1", since = "1.0.0")]
336 #[rustc_const_unstable(feature = "const_slice_index", issue = "none")]
339 pub const fn get<I>(&self, index: I) -> Option<&I::Output>
341 I: ~const SliceIndex<Self>,
346 /// Returns a mutable reference to an element or subslice depending on the
347 /// type of index (see [`get`]) or `None` if the index is out of bounds.
349 /// [`get`]: slice::get
354 /// let x = &mut [0, 1, 2];
356 /// if let Some(elem) = x.get_mut(1) {
359 /// assert_eq!(x, &[0, 42, 2]);
361 #[stable(feature = "rust1", since = "1.0.0")]
362 #[rustc_const_unstable(feature = "const_slice_index", issue = "none")]
365 pub const fn get_mut<I>(&mut self, index: I) -> Option<&mut I::Output>
367 I: ~const SliceIndex<Self>,
372 /// Returns a reference to an element or subslice, without doing bounds
375 /// For a safe alternative see [`get`].
379 /// Calling this method with an out-of-bounds index is *[undefined behavior]*
380 /// even if the resulting reference is not used.
382 /// [`get`]: slice::get
383 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
388 /// let x = &[1, 2, 4];
391 /// assert_eq!(x.get_unchecked(1), &2);
394 #[stable(feature = "rust1", since = "1.0.0")]
395 #[rustc_const_unstable(feature = "const_slice_index", issue = "none")]
398 pub const unsafe fn get_unchecked<I>(&self, index: I) -> &I::Output
400 I: ~const SliceIndex<Self>,
402 // SAFETY: the caller must uphold most of the safety requirements for `get_unchecked`;
403 // the slice is dereferenceable because `self` is a safe reference.
404 // The returned pointer is safe because impls of `SliceIndex` have to guarantee that it is.
405 unsafe { &*index.get_unchecked(self) }
408 /// Returns a mutable reference to an element or subslice, without doing
411 /// For a safe alternative see [`get_mut`].
415 /// Calling this method with an out-of-bounds index is *[undefined behavior]*
416 /// even if the resulting reference is not used.
418 /// [`get_mut`]: slice::get_mut
419 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
424 /// let x = &mut [1, 2, 4];
427 /// let elem = x.get_unchecked_mut(1);
430 /// assert_eq!(x, &[1, 13, 4]);
432 #[stable(feature = "rust1", since = "1.0.0")]
433 #[rustc_const_unstable(feature = "const_slice_index", issue = "none")]
436 pub const unsafe fn get_unchecked_mut<I>(&mut self, index: I) -> &mut I::Output
438 I: ~const SliceIndex<Self>,
440 // SAFETY: the caller must uphold the safety requirements for `get_unchecked_mut`;
441 // the slice is dereferenceable because `self` is a safe reference.
442 // The returned pointer is safe because impls of `SliceIndex` have to guarantee that it is.
443 unsafe { &mut *index.get_unchecked_mut(self) }
446 /// Returns a raw pointer to the slice's buffer.
448 /// The caller must ensure that the slice outlives the pointer this
449 /// function returns, or else it will end up pointing to garbage.
451 /// The caller must also ensure that the memory the pointer (non-transitively) points to
452 /// is never written to (except inside an `UnsafeCell`) using this pointer or any pointer
453 /// derived from it. If you need to mutate the contents of the slice, use [`as_mut_ptr`].
455 /// Modifying the container referenced by this slice may cause its buffer
456 /// to be reallocated, which would also make any pointers to it invalid.
461 /// let x = &[1, 2, 4];
462 /// let x_ptr = x.as_ptr();
465 /// for i in 0..x.len() {
466 /// assert_eq!(x.get_unchecked(i), &*x_ptr.add(i));
471 /// [`as_mut_ptr`]: slice::as_mut_ptr
472 #[stable(feature = "rust1", since = "1.0.0")]
473 #[rustc_const_stable(feature = "const_slice_as_ptr", since = "1.32.0")]
476 pub const fn as_ptr(&self) -> *const T {
477 self as *const [T] as *const T
480 /// Returns an unsafe mutable pointer to the slice's buffer.
482 /// The caller must ensure that the slice outlives the pointer this
483 /// function returns, or else it will end up pointing to garbage.
485 /// Modifying the container referenced by this slice may cause its buffer
486 /// to be reallocated, which would also make any pointers to it invalid.
491 /// let x = &mut [1, 2, 4];
492 /// let x_ptr = x.as_mut_ptr();
495 /// for i in 0..x.len() {
496 /// *x_ptr.add(i) += 2;
499 /// assert_eq!(x, &[3, 4, 6]);
501 #[stable(feature = "rust1", since = "1.0.0")]
502 #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")]
503 #[rustc_allow_const_fn_unstable(const_mut_refs)]
506 pub const fn as_mut_ptr(&mut self) -> *mut T {
507 self as *mut [T] as *mut T
510 /// Returns the two raw pointers spanning the slice.
512 /// The returned range is half-open, which means that the end pointer
513 /// points *one past* the last element of the slice. This way, an empty
514 /// slice is represented by two equal pointers, and the difference between
515 /// the two pointers represents the size of the slice.
517 /// See [`as_ptr`] for warnings on using these pointers. The end pointer
518 /// requires extra caution, as it does not point to a valid element in the
521 /// This function is useful for interacting with foreign interfaces which
522 /// use two pointers to refer to a range of elements in memory, as is
525 /// It can also be useful to check if a pointer to an element refers to an
526 /// element of this slice:
529 /// let a = [1, 2, 3];
530 /// let x = &a[1] as *const _;
531 /// let y = &5 as *const _;
533 /// assert!(a.as_ptr_range().contains(&x));
534 /// assert!(!a.as_ptr_range().contains(&y));
537 /// [`as_ptr`]: slice::as_ptr
538 #[stable(feature = "slice_ptr_range", since = "1.48.0")]
539 #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")]
542 pub const fn as_ptr_range(&self) -> Range<*const T> {
543 let start = self.as_ptr();
544 // SAFETY: The `add` here is safe, because:
546 // - Both pointers are part of the same object, as pointing directly
547 // past the object also counts.
549 // - The size of the slice is never larger than isize::MAX bytes, as
551 // - https://github.com/rust-lang/unsafe-code-guidelines/issues/102#issuecomment-473340447
552 // - https://doc.rust-lang.org/reference/behavior-considered-undefined.html
553 // - https://doc.rust-lang.org/core/slice/fn.from_raw_parts.html#safety
554 // (This doesn't seem normative yet, but the very same assumption is
555 // made in many places, including the Index implementation of slices.)
557 // - There is no wrapping around involved, as slices do not wrap past
558 // the end of the address space.
560 // See the documentation of pointer::add.
561 let end = unsafe { start.add(self.len()) };
565 /// Returns the two unsafe mutable pointers spanning the slice.
567 /// The returned range is half-open, which means that the end pointer
568 /// points *one past* the last element of the slice. This way, an empty
569 /// slice is represented by two equal pointers, and the difference between
570 /// the two pointers represents the size of the slice.
572 /// See [`as_mut_ptr`] for warnings on using these pointers. The end
573 /// pointer requires extra caution, as it does not point to a valid element
576 /// This function is useful for interacting with foreign interfaces which
577 /// use two pointers to refer to a range of elements in memory, as is
580 /// [`as_mut_ptr`]: slice::as_mut_ptr
581 #[stable(feature = "slice_ptr_range", since = "1.48.0")]
582 #[rustc_const_stable(feature = "const_ptr_offset", since = "1.61.0")]
583 #[rustc_allow_const_fn_unstable(const_mut_refs)]
586 pub const fn as_mut_ptr_range(&mut self) -> Range<*mut T> {
587 let start = self.as_mut_ptr();
588 // SAFETY: See as_ptr_range() above for why `add` here is safe.
589 let end = unsafe { start.add(self.len()) };
593 /// Swaps two elements in the slice.
597 /// * a - The index of the first element
598 /// * b - The index of the second element
602 /// Panics if `a` or `b` are out of bounds.
607 /// let mut v = ["a", "b", "c", "d", "e"];
609 /// assert!(v == ["a", "b", "e", "d", "c"]);
611 #[stable(feature = "rust1", since = "1.0.0")]
612 #[rustc_const_unstable(feature = "const_swap", issue = "83163")]
615 pub const fn swap(&mut self, a: usize, b: usize) {
616 // FIXME: use swap_unchecked here (https://github.com/rust-lang/rust/pull/88540#issuecomment-944344343)
617 // Can't take two mutable loans from one vector, so instead use raw pointers.
618 let pa = ptr::addr_of_mut!(self[a]);
619 let pb = ptr::addr_of_mut!(self[b]);
620 // SAFETY: `pa` and `pb` have been created from safe mutable references and refer
621 // to elements in the slice and therefore are guaranteed to be valid and aligned.
622 // Note that accessing the elements behind `a` and `b` is checked and will
623 // panic when out of bounds.
629 /// Swaps two elements in the slice, without doing bounds checking.
631 /// For a safe alternative see [`swap`].
635 /// * a - The index of the first element
636 /// * b - The index of the second element
640 /// Calling this method with an out-of-bounds index is *[undefined behavior]*.
641 /// The caller has to ensure that `a < self.len()` and `b < self.len()`.
646 /// #![feature(slice_swap_unchecked)]
648 /// let mut v = ["a", "b", "c", "d"];
649 /// // SAFETY: we know that 1 and 3 are both indices of the slice
650 /// unsafe { v.swap_unchecked(1, 3) };
651 /// assert!(v == ["a", "d", "c", "b"]);
654 /// [`swap`]: slice::swap
655 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
656 #[unstable(feature = "slice_swap_unchecked", issue = "88539")]
657 #[rustc_const_unstable(feature = "const_swap", issue = "83163")]
658 pub const unsafe fn swap_unchecked(&mut self, a: usize, b: usize) {
660 let ptr = this.as_mut_ptr();
661 // SAFETY: caller has to guarantee that `a < self.len()` and `b < self.len()`
663 assert_unsafe_precondition!([T](a: usize, b: usize, this: &mut [T]) => a < this.len() && b < this.len());
664 ptr::swap(ptr.add(a), ptr.add(b));
668 /// Reverses the order of elements in the slice, in place.
673 /// let mut v = [1, 2, 3];
675 /// assert!(v == [3, 2, 1]);
677 #[stable(feature = "rust1", since = "1.0.0")]
678 #[rustc_const_unstable(feature = "const_reverse", issue = "100784")]
680 pub const fn reverse(&mut self) {
681 let half_len = self.len() / 2;
682 let Range { start, end } = self.as_mut_ptr_range();
684 // These slices will skip the middle item for an odd length,
685 // since that one doesn't need to move.
686 let (front_half, back_half) =
687 // SAFETY: Both are subparts of the original slice, so the memory
688 // range is valid, and they don't overlap because they're each only
689 // half (or less) of the original slice.
692 slice::from_raw_parts_mut(start, half_len),
693 slice::from_raw_parts_mut(end.sub(half_len), half_len),
697 // Introducing a function boundary here means that the two halves
698 // get `noalias` markers, allowing better optimization as LLVM
699 // knows that they're disjoint, unlike in the original slice.
700 revswap(front_half, back_half, half_len);
703 const fn revswap<T>(a: &mut [T], b: &mut [T], n: usize) {
704 debug_assert!(a.len() == n);
705 debug_assert!(b.len() == n);
707 // Because this function is first compiled in isolation,
708 // this check tells LLVM that the indexing below is
709 // in-bounds. Then after inlining -- once the actual
710 // lengths of the slices are known -- it's removed.
711 let (a, b) = (&mut a[..n], &mut b[..n]);
715 mem::swap(&mut a[i], &mut b[n - 1 - i]);
721 /// Returns an iterator over the slice.
723 /// The iterator yields all items from start to end.
728 /// let x = &[1, 2, 4];
729 /// let mut iterator = x.iter();
731 /// assert_eq!(iterator.next(), Some(&1));
732 /// assert_eq!(iterator.next(), Some(&2));
733 /// assert_eq!(iterator.next(), Some(&4));
734 /// assert_eq!(iterator.next(), None);
736 #[stable(feature = "rust1", since = "1.0.0")]
738 pub fn iter(&self) -> Iter<'_, T> {
742 /// Returns an iterator that allows modifying each value.
744 /// The iterator yields all items from start to end.
749 /// let x = &mut [1, 2, 4];
750 /// for elem in x.iter_mut() {
753 /// assert_eq!(x, &[3, 4, 6]);
755 #[stable(feature = "rust1", since = "1.0.0")]
757 pub fn iter_mut(&mut self) -> IterMut<'_, T> {
761 /// Returns an iterator over all contiguous windows of length
762 /// `size`. The windows overlap. If the slice is shorter than
763 /// `size`, the iterator returns no values.
767 /// Panics if `size` is 0.
772 /// let slice = ['r', 'u', 's', 't'];
773 /// let mut iter = slice.windows(2);
774 /// assert_eq!(iter.next().unwrap(), &['r', 'u']);
775 /// assert_eq!(iter.next().unwrap(), &['u', 's']);
776 /// assert_eq!(iter.next().unwrap(), &['s', 't']);
777 /// assert!(iter.next().is_none());
780 /// If the slice is shorter than `size`:
783 /// let slice = ['f', 'o', 'o'];
784 /// let mut iter = slice.windows(4);
785 /// assert!(iter.next().is_none());
787 #[stable(feature = "rust1", since = "1.0.0")]
789 pub fn windows(&self, size: usize) -> Windows<'_, T> {
790 let size = NonZeroUsize::new(size).expect("size is zero");
791 Windows::new(self, size)
794 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
795 /// beginning of the slice.
797 /// The chunks are slices and do not overlap. If `chunk_size` does not divide the length of the
798 /// slice, then the last chunk will not have length `chunk_size`.
800 /// See [`chunks_exact`] for a variant of this iterator that returns chunks of always exactly
801 /// `chunk_size` elements, and [`rchunks`] for the same iterator but starting at the end of the
806 /// Panics if `chunk_size` is 0.
811 /// let slice = ['l', 'o', 'r', 'e', 'm'];
812 /// let mut iter = slice.chunks(2);
813 /// assert_eq!(iter.next().unwrap(), &['l', 'o']);
814 /// assert_eq!(iter.next().unwrap(), &['r', 'e']);
815 /// assert_eq!(iter.next().unwrap(), &['m']);
816 /// assert!(iter.next().is_none());
819 /// [`chunks_exact`]: slice::chunks_exact
820 /// [`rchunks`]: slice::rchunks
821 #[stable(feature = "rust1", since = "1.0.0")]
823 pub fn chunks(&self, chunk_size: usize) -> Chunks<'_, T> {
824 assert_ne!(chunk_size, 0, "chunks cannot have a size of zero");
825 Chunks::new(self, chunk_size)
828 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
829 /// beginning of the slice.
831 /// The chunks are mutable slices, and do not overlap. If `chunk_size` does not divide the
832 /// length of the slice, then the last chunk will not have length `chunk_size`.
834 /// See [`chunks_exact_mut`] for a variant of this iterator that returns chunks of always
835 /// exactly `chunk_size` elements, and [`rchunks_mut`] for the same iterator but starting at
836 /// the end of the slice.
840 /// Panics if `chunk_size` is 0.
845 /// let v = &mut [0, 0, 0, 0, 0];
846 /// let mut count = 1;
848 /// for chunk in v.chunks_mut(2) {
849 /// for elem in chunk.iter_mut() {
854 /// assert_eq!(v, &[1, 1, 2, 2, 3]);
857 /// [`chunks_exact_mut`]: slice::chunks_exact_mut
858 /// [`rchunks_mut`]: slice::rchunks_mut
859 #[stable(feature = "rust1", since = "1.0.0")]
861 pub fn chunks_mut(&mut self, chunk_size: usize) -> ChunksMut<'_, T> {
862 assert_ne!(chunk_size, 0, "chunks cannot have a size of zero");
863 ChunksMut::new(self, chunk_size)
866 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
867 /// beginning of the slice.
869 /// The chunks are slices and do not overlap. If `chunk_size` does not divide the length of the
870 /// slice, then the last up to `chunk_size-1` elements will be omitted and can be retrieved
871 /// from the `remainder` function of the iterator.
873 /// Due to each chunk having exactly `chunk_size` elements, the compiler can often optimize the
874 /// resulting code better than in the case of [`chunks`].
876 /// See [`chunks`] for a variant of this iterator that also returns the remainder as a smaller
877 /// chunk, and [`rchunks_exact`] for the same iterator but starting at the end of the slice.
881 /// Panics if `chunk_size` is 0.
886 /// let slice = ['l', 'o', 'r', 'e', 'm'];
887 /// let mut iter = slice.chunks_exact(2);
888 /// assert_eq!(iter.next().unwrap(), &['l', 'o']);
889 /// assert_eq!(iter.next().unwrap(), &['r', 'e']);
890 /// assert!(iter.next().is_none());
891 /// assert_eq!(iter.remainder(), &['m']);
894 /// [`chunks`]: slice::chunks
895 /// [`rchunks_exact`]: slice::rchunks_exact
896 #[stable(feature = "chunks_exact", since = "1.31.0")]
898 pub fn chunks_exact(&self, chunk_size: usize) -> ChunksExact<'_, T> {
899 assert_ne!(chunk_size, 0);
900 ChunksExact::new(self, chunk_size)
903 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
904 /// beginning of the slice.
906 /// The chunks are mutable slices, and do not overlap. If `chunk_size` does not divide the
907 /// length of the slice, then the last up to `chunk_size-1` elements will be omitted and can be
908 /// retrieved from the `into_remainder` function of the iterator.
910 /// Due to each chunk having exactly `chunk_size` elements, the compiler can often optimize the
911 /// resulting code better than in the case of [`chunks_mut`].
913 /// See [`chunks_mut`] for a variant of this iterator that also returns the remainder as a
914 /// smaller chunk, and [`rchunks_exact_mut`] for the same iterator but starting at the end of
919 /// Panics if `chunk_size` is 0.
924 /// let v = &mut [0, 0, 0, 0, 0];
925 /// let mut count = 1;
927 /// for chunk in v.chunks_exact_mut(2) {
928 /// for elem in chunk.iter_mut() {
933 /// assert_eq!(v, &[1, 1, 2, 2, 0]);
936 /// [`chunks_mut`]: slice::chunks_mut
937 /// [`rchunks_exact_mut`]: slice::rchunks_exact_mut
938 #[stable(feature = "chunks_exact", since = "1.31.0")]
940 pub fn chunks_exact_mut(&mut self, chunk_size: usize) -> ChunksExactMut<'_, T> {
941 assert_ne!(chunk_size, 0);
942 ChunksExactMut::new(self, chunk_size)
945 /// Splits the slice into a slice of `N`-element arrays,
946 /// assuming that there's no remainder.
950 /// This may only be called when
951 /// - The slice splits exactly into `N`-element chunks (aka `self.len() % N == 0`).
957 /// #![feature(slice_as_chunks)]
958 /// let slice: &[char] = &['l', 'o', 'r', 'e', 'm', '!'];
959 /// let chunks: &[[char; 1]] =
960 /// // SAFETY: 1-element chunks never have remainder
961 /// unsafe { slice.as_chunks_unchecked() };
962 /// assert_eq!(chunks, &[['l'], ['o'], ['r'], ['e'], ['m'], ['!']]);
963 /// let chunks: &[[char; 3]] =
964 /// // SAFETY: The slice length (6) is a multiple of 3
965 /// unsafe { slice.as_chunks_unchecked() };
966 /// assert_eq!(chunks, &[['l', 'o', 'r'], ['e', 'm', '!']]);
968 /// // These would be unsound:
969 /// // let chunks: &[[_; 5]] = slice.as_chunks_unchecked() // The slice length is not a multiple of 5
970 /// // let chunks: &[[_; 0]] = slice.as_chunks_unchecked() // Zero-length chunks are never allowed
972 #[unstable(feature = "slice_as_chunks", issue = "74985")]
975 pub unsafe fn as_chunks_unchecked<const N: usize>(&self) -> &[[T; N]] {
977 // SAFETY: Caller must guarantee that `N` is nonzero and exactly divides the slice length
978 let new_len = unsafe {
979 assert_unsafe_precondition!([T](this: &[T], N: usize) => N != 0 && this.len() % N == 0);
980 exact_div(self.len(), N)
982 // SAFETY: We cast a slice of `new_len * N` elements into
983 // a slice of `new_len` many `N` elements chunks.
984 unsafe { from_raw_parts(self.as_ptr().cast(), new_len) }
987 /// Splits the slice into a slice of `N`-element arrays,
988 /// starting at the beginning of the slice,
989 /// and a remainder slice with length strictly less than `N`.
993 /// Panics if `N` is 0. This check will most probably get changed to a compile time
994 /// error before this method gets stabilized.
999 /// #![feature(slice_as_chunks)]
1000 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1001 /// let (chunks, remainder) = slice.as_chunks();
1002 /// assert_eq!(chunks, &[['l', 'o'], ['r', 'e']]);
1003 /// assert_eq!(remainder, &['m']);
1005 #[unstable(feature = "slice_as_chunks", issue = "74985")]
1008 pub fn as_chunks<const N: usize>(&self) -> (&[[T; N]], &[T]) {
1010 let len = self.len() / N;
1011 let (multiple_of_n, remainder) = self.split_at(len * N);
1012 // SAFETY: We already panicked for zero, and ensured by construction
1013 // that the length of the subslice is a multiple of N.
1014 let array_slice = unsafe { multiple_of_n.as_chunks_unchecked() };
1015 (array_slice, remainder)
1018 /// Splits the slice into a slice of `N`-element arrays,
1019 /// starting at the end of the slice,
1020 /// and a remainder slice with length strictly less than `N`.
1024 /// Panics if `N` is 0. This check will most probably get changed to a compile time
1025 /// error before this method gets stabilized.
1030 /// #![feature(slice_as_chunks)]
1031 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1032 /// let (remainder, chunks) = slice.as_rchunks();
1033 /// assert_eq!(remainder, &['l']);
1034 /// assert_eq!(chunks, &[['o', 'r'], ['e', 'm']]);
1036 #[unstable(feature = "slice_as_chunks", issue = "74985")]
1039 pub fn as_rchunks<const N: usize>(&self) -> (&[T], &[[T; N]]) {
1041 let len = self.len() / N;
1042 let (remainder, multiple_of_n) = self.split_at(self.len() - len * N);
1043 // SAFETY: We already panicked for zero, and ensured by construction
1044 // that the length of the subslice is a multiple of N.
1045 let array_slice = unsafe { multiple_of_n.as_chunks_unchecked() };
1046 (remainder, array_slice)
1049 /// Returns an iterator over `N` elements of the slice at a time, starting at the
1050 /// beginning of the slice.
1052 /// The chunks are array references and do not overlap. If `N` does not divide the
1053 /// length of the slice, then the last up to `N-1` elements will be omitted and can be
1054 /// retrieved from the `remainder` function of the iterator.
1056 /// This method is the const generic equivalent of [`chunks_exact`].
1060 /// Panics if `N` is 0. This check will most probably get changed to a compile time
1061 /// error before this method gets stabilized.
1066 /// #![feature(array_chunks)]
1067 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1068 /// let mut iter = slice.array_chunks();
1069 /// assert_eq!(iter.next().unwrap(), &['l', 'o']);
1070 /// assert_eq!(iter.next().unwrap(), &['r', 'e']);
1071 /// assert!(iter.next().is_none());
1072 /// assert_eq!(iter.remainder(), &['m']);
1075 /// [`chunks_exact`]: slice::chunks_exact
1076 #[unstable(feature = "array_chunks", issue = "74985")]
1078 pub fn array_chunks<const N: usize>(&self) -> ArrayChunks<'_, T, N> {
1080 ArrayChunks::new(self)
1083 /// Splits the slice into a slice of `N`-element arrays,
1084 /// assuming that there's no remainder.
1088 /// This may only be called when
1089 /// - The slice splits exactly into `N`-element chunks (aka `self.len() % N == 0`).
1095 /// #![feature(slice_as_chunks)]
1096 /// let slice: &mut [char] = &mut ['l', 'o', 'r', 'e', 'm', '!'];
1097 /// let chunks: &mut [[char; 1]] =
1098 /// // SAFETY: 1-element chunks never have remainder
1099 /// unsafe { slice.as_chunks_unchecked_mut() };
1100 /// chunks[0] = ['L'];
1101 /// assert_eq!(chunks, &[['L'], ['o'], ['r'], ['e'], ['m'], ['!']]);
1102 /// let chunks: &mut [[char; 3]] =
1103 /// // SAFETY: The slice length (6) is a multiple of 3
1104 /// unsafe { slice.as_chunks_unchecked_mut() };
1105 /// chunks[1] = ['a', 'x', '?'];
1106 /// assert_eq!(slice, &['L', 'o', 'r', 'a', 'x', '?']);
1108 /// // These would be unsound:
1109 /// // let chunks: &[[_; 5]] = slice.as_chunks_unchecked_mut() // The slice length is not a multiple of 5
1110 /// // let chunks: &[[_; 0]] = slice.as_chunks_unchecked_mut() // Zero-length chunks are never allowed
1112 #[unstable(feature = "slice_as_chunks", issue = "74985")]
1115 pub unsafe fn as_chunks_unchecked_mut<const N: usize>(&mut self) -> &mut [[T; N]] {
1117 // SAFETY: Caller must guarantee that `N` is nonzero and exactly divides the slice length
1118 let new_len = unsafe {
1119 assert_unsafe_precondition!([T](this: &[T], N: usize) => N != 0 && this.len() % N == 0);
1120 exact_div(this.len(), N)
1122 // SAFETY: We cast a slice of `new_len * N` elements into
1123 // a slice of `new_len` many `N` elements chunks.
1124 unsafe { from_raw_parts_mut(self.as_mut_ptr().cast(), new_len) }
1127 /// Splits the slice into a slice of `N`-element arrays,
1128 /// starting at the beginning of the slice,
1129 /// and a remainder slice with length strictly less than `N`.
1133 /// Panics if `N` is 0. This check will most probably get changed to a compile time
1134 /// error before this method gets stabilized.
1139 /// #![feature(slice_as_chunks)]
1140 /// let v = &mut [0, 0, 0, 0, 0];
1141 /// let mut count = 1;
1143 /// let (chunks, remainder) = v.as_chunks_mut();
1144 /// remainder[0] = 9;
1145 /// for chunk in chunks {
1146 /// *chunk = [count; 2];
1149 /// assert_eq!(v, &[1, 1, 2, 2, 9]);
1151 #[unstable(feature = "slice_as_chunks", issue = "74985")]
1154 pub fn as_chunks_mut<const N: usize>(&mut self) -> (&mut [[T; N]], &mut [T]) {
1156 let len = self.len() / N;
1157 let (multiple_of_n, remainder) = self.split_at_mut(len * N);
1158 // SAFETY: We already panicked for zero, and ensured by construction
1159 // that the length of the subslice is a multiple of N.
1160 let array_slice = unsafe { multiple_of_n.as_chunks_unchecked_mut() };
1161 (array_slice, remainder)
1164 /// Splits the slice into a slice of `N`-element arrays,
1165 /// starting at the end of the slice,
1166 /// and a remainder slice with length strictly less than `N`.
1170 /// Panics if `N` is 0. This check will most probably get changed to a compile time
1171 /// error before this method gets stabilized.
1176 /// #![feature(slice_as_chunks)]
1177 /// let v = &mut [0, 0, 0, 0, 0];
1178 /// let mut count = 1;
1180 /// let (remainder, chunks) = v.as_rchunks_mut();
1181 /// remainder[0] = 9;
1182 /// for chunk in chunks {
1183 /// *chunk = [count; 2];
1186 /// assert_eq!(v, &[9, 1, 1, 2, 2]);
1188 #[unstable(feature = "slice_as_chunks", issue = "74985")]
1191 pub fn as_rchunks_mut<const N: usize>(&mut self) -> (&mut [T], &mut [[T; N]]) {
1193 let len = self.len() / N;
1194 let (remainder, multiple_of_n) = self.split_at_mut(self.len() - len * N);
1195 // SAFETY: We already panicked for zero, and ensured by construction
1196 // that the length of the subslice is a multiple of N.
1197 let array_slice = unsafe { multiple_of_n.as_chunks_unchecked_mut() };
1198 (remainder, array_slice)
1201 /// Returns an iterator over `N` elements of the slice at a time, starting at the
1202 /// beginning of the slice.
1204 /// The chunks are mutable array references and do not overlap. If `N` does not divide
1205 /// the length of the slice, then the last up to `N-1` elements will be omitted and
1206 /// can be retrieved from the `into_remainder` function of the iterator.
1208 /// This method is the const generic equivalent of [`chunks_exact_mut`].
1212 /// Panics if `N` is 0. This check will most probably get changed to a compile time
1213 /// error before this method gets stabilized.
1218 /// #![feature(array_chunks)]
1219 /// let v = &mut [0, 0, 0, 0, 0];
1220 /// let mut count = 1;
1222 /// for chunk in v.array_chunks_mut() {
1223 /// *chunk = [count; 2];
1226 /// assert_eq!(v, &[1, 1, 2, 2, 0]);
1229 /// [`chunks_exact_mut`]: slice::chunks_exact_mut
1230 #[unstable(feature = "array_chunks", issue = "74985")]
1232 pub fn array_chunks_mut<const N: usize>(&mut self) -> ArrayChunksMut<'_, T, N> {
1234 ArrayChunksMut::new(self)
1237 /// Returns an iterator over overlapping windows of `N` elements of a slice,
1238 /// starting at the beginning of the slice.
1240 /// This is the const generic equivalent of [`windows`].
1242 /// If `N` is greater than the size of the slice, it will return no windows.
1246 /// Panics if `N` is 0. This check will most probably get changed to a compile time
1247 /// error before this method gets stabilized.
1252 /// #![feature(array_windows)]
1253 /// let slice = [0, 1, 2, 3];
1254 /// let mut iter = slice.array_windows();
1255 /// assert_eq!(iter.next().unwrap(), &[0, 1]);
1256 /// assert_eq!(iter.next().unwrap(), &[1, 2]);
1257 /// assert_eq!(iter.next().unwrap(), &[2, 3]);
1258 /// assert!(iter.next().is_none());
1261 /// [`windows`]: slice::windows
1262 #[unstable(feature = "array_windows", issue = "75027")]
1264 pub fn array_windows<const N: usize>(&self) -> ArrayWindows<'_, T, N> {
1266 ArrayWindows::new(self)
1269 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the end
1272 /// The chunks are slices and do not overlap. If `chunk_size` does not divide the length of the
1273 /// slice, then the last chunk will not have length `chunk_size`.
1275 /// See [`rchunks_exact`] for a variant of this iterator that returns chunks of always exactly
1276 /// `chunk_size` elements, and [`chunks`] for the same iterator but starting at the beginning
1281 /// Panics if `chunk_size` is 0.
1286 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1287 /// let mut iter = slice.rchunks(2);
1288 /// assert_eq!(iter.next().unwrap(), &['e', 'm']);
1289 /// assert_eq!(iter.next().unwrap(), &['o', 'r']);
1290 /// assert_eq!(iter.next().unwrap(), &['l']);
1291 /// assert!(iter.next().is_none());
1294 /// [`rchunks_exact`]: slice::rchunks_exact
1295 /// [`chunks`]: slice::chunks
1296 #[stable(feature = "rchunks", since = "1.31.0")]
1298 pub fn rchunks(&self, chunk_size: usize) -> RChunks<'_, T> {
1299 assert!(chunk_size != 0);
1300 RChunks::new(self, chunk_size)
1303 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the end
1306 /// The chunks are mutable slices, and do not overlap. If `chunk_size` does not divide the
1307 /// length of the slice, then the last chunk will not have length `chunk_size`.
1309 /// See [`rchunks_exact_mut`] for a variant of this iterator that returns chunks of always
1310 /// exactly `chunk_size` elements, and [`chunks_mut`] for the same iterator but starting at the
1311 /// beginning of the slice.
1315 /// Panics if `chunk_size` is 0.
1320 /// let v = &mut [0, 0, 0, 0, 0];
1321 /// let mut count = 1;
1323 /// for chunk in v.rchunks_mut(2) {
1324 /// for elem in chunk.iter_mut() {
1329 /// assert_eq!(v, &[3, 2, 2, 1, 1]);
1332 /// [`rchunks_exact_mut`]: slice::rchunks_exact_mut
1333 /// [`chunks_mut`]: slice::chunks_mut
1334 #[stable(feature = "rchunks", since = "1.31.0")]
1336 pub fn rchunks_mut(&mut self, chunk_size: usize) -> RChunksMut<'_, T> {
1337 assert!(chunk_size != 0);
1338 RChunksMut::new(self, chunk_size)
1341 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the
1342 /// end of the slice.
1344 /// The chunks are slices and do not overlap. If `chunk_size` does not divide the length of the
1345 /// slice, then the last up to `chunk_size-1` elements will be omitted and can be retrieved
1346 /// from the `remainder` function of the iterator.
1348 /// Due to each chunk having exactly `chunk_size` elements, the compiler can often optimize the
1349 /// resulting code better than in the case of [`rchunks`].
1351 /// See [`rchunks`] for a variant of this iterator that also returns the remainder as a smaller
1352 /// chunk, and [`chunks_exact`] for the same iterator but starting at the beginning of the
1357 /// Panics if `chunk_size` is 0.
1362 /// let slice = ['l', 'o', 'r', 'e', 'm'];
1363 /// let mut iter = slice.rchunks_exact(2);
1364 /// assert_eq!(iter.next().unwrap(), &['e', 'm']);
1365 /// assert_eq!(iter.next().unwrap(), &['o', 'r']);
1366 /// assert!(iter.next().is_none());
1367 /// assert_eq!(iter.remainder(), &['l']);
1370 /// [`chunks`]: slice::chunks
1371 /// [`rchunks`]: slice::rchunks
1372 /// [`chunks_exact`]: slice::chunks_exact
1373 #[stable(feature = "rchunks", since = "1.31.0")]
1375 pub fn rchunks_exact(&self, chunk_size: usize) -> RChunksExact<'_, T> {
1376 assert!(chunk_size != 0);
1377 RChunksExact::new(self, chunk_size)
1380 /// Returns an iterator over `chunk_size` elements of the slice at a time, starting at the end
1383 /// The chunks are mutable slices, and do not overlap. If `chunk_size` does not divide the
1384 /// length of the slice, then the last up to `chunk_size-1` elements will be omitted and can be
1385 /// retrieved from the `into_remainder` function of the iterator.
1387 /// Due to each chunk having exactly `chunk_size` elements, the compiler can often optimize the
1388 /// resulting code better than in the case of [`chunks_mut`].
1390 /// See [`rchunks_mut`] for a variant of this iterator that also returns the remainder as a
1391 /// smaller chunk, and [`chunks_exact_mut`] for the same iterator but starting at the beginning
1396 /// Panics if `chunk_size` is 0.
1401 /// let v = &mut [0, 0, 0, 0, 0];
1402 /// let mut count = 1;
1404 /// for chunk in v.rchunks_exact_mut(2) {
1405 /// for elem in chunk.iter_mut() {
1410 /// assert_eq!(v, &[0, 2, 2, 1, 1]);
1413 /// [`chunks_mut`]: slice::chunks_mut
1414 /// [`rchunks_mut`]: slice::rchunks_mut
1415 /// [`chunks_exact_mut`]: slice::chunks_exact_mut
1416 #[stable(feature = "rchunks", since = "1.31.0")]
1418 pub fn rchunks_exact_mut(&mut self, chunk_size: usize) -> RChunksExactMut<'_, T> {
1419 assert!(chunk_size != 0);
1420 RChunksExactMut::new(self, chunk_size)
1423 /// Returns an iterator over the slice producing non-overlapping runs
1424 /// of elements using the predicate to separate them.
1426 /// The predicate is called on two elements following themselves,
1427 /// it means the predicate is called on `slice[0]` and `slice[1]`
1428 /// then on `slice[1]` and `slice[2]` and so on.
1433 /// #![feature(slice_group_by)]
1435 /// let slice = &[1, 1, 1, 3, 3, 2, 2, 2];
1437 /// let mut iter = slice.group_by(|a, b| a == b);
1439 /// assert_eq!(iter.next(), Some(&[1, 1, 1][..]));
1440 /// assert_eq!(iter.next(), Some(&[3, 3][..]));
1441 /// assert_eq!(iter.next(), Some(&[2, 2, 2][..]));
1442 /// assert_eq!(iter.next(), None);
1445 /// This method can be used to extract the sorted subslices:
1448 /// #![feature(slice_group_by)]
1450 /// let slice = &[1, 1, 2, 3, 2, 3, 2, 3, 4];
1452 /// let mut iter = slice.group_by(|a, b| a <= b);
1454 /// assert_eq!(iter.next(), Some(&[1, 1, 2, 3][..]));
1455 /// assert_eq!(iter.next(), Some(&[2, 3][..]));
1456 /// assert_eq!(iter.next(), Some(&[2, 3, 4][..]));
1457 /// assert_eq!(iter.next(), None);
1459 #[unstable(feature = "slice_group_by", issue = "80552")]
1461 pub fn group_by<F>(&self, pred: F) -> GroupBy<'_, T, F>
1463 F: FnMut(&T, &T) -> bool,
1465 GroupBy::new(self, pred)
1468 /// Returns an iterator over the slice producing non-overlapping mutable
1469 /// runs of elements using the predicate to separate them.
1471 /// The predicate is called on two elements following themselves,
1472 /// it means the predicate is called on `slice[0]` and `slice[1]`
1473 /// then on `slice[1]` and `slice[2]` and so on.
1478 /// #![feature(slice_group_by)]
1480 /// let slice = &mut [1, 1, 1, 3, 3, 2, 2, 2];
1482 /// let mut iter = slice.group_by_mut(|a, b| a == b);
1484 /// assert_eq!(iter.next(), Some(&mut [1, 1, 1][..]));
1485 /// assert_eq!(iter.next(), Some(&mut [3, 3][..]));
1486 /// assert_eq!(iter.next(), Some(&mut [2, 2, 2][..]));
1487 /// assert_eq!(iter.next(), None);
1490 /// This method can be used to extract the sorted subslices:
1493 /// #![feature(slice_group_by)]
1495 /// let slice = &mut [1, 1, 2, 3, 2, 3, 2, 3, 4];
1497 /// let mut iter = slice.group_by_mut(|a, b| a <= b);
1499 /// assert_eq!(iter.next(), Some(&mut [1, 1, 2, 3][..]));
1500 /// assert_eq!(iter.next(), Some(&mut [2, 3][..]));
1501 /// assert_eq!(iter.next(), Some(&mut [2, 3, 4][..]));
1502 /// assert_eq!(iter.next(), None);
1504 #[unstable(feature = "slice_group_by", issue = "80552")]
1506 pub fn group_by_mut<F>(&mut self, pred: F) -> GroupByMut<'_, T, F>
1508 F: FnMut(&T, &T) -> bool,
1510 GroupByMut::new(self, pred)
1513 /// Divides one slice into two at an index.
1515 /// The first will contain all indices from `[0, mid)` (excluding
1516 /// the index `mid` itself) and the second will contain all
1517 /// indices from `[mid, len)` (excluding the index `len` itself).
1521 /// Panics if `mid > len`.
1526 /// let v = [1, 2, 3, 4, 5, 6];
1529 /// let (left, right) = v.split_at(0);
1530 /// assert_eq!(left, []);
1531 /// assert_eq!(right, [1, 2, 3, 4, 5, 6]);
1535 /// let (left, right) = v.split_at(2);
1536 /// assert_eq!(left, [1, 2]);
1537 /// assert_eq!(right, [3, 4, 5, 6]);
1541 /// let (left, right) = v.split_at(6);
1542 /// assert_eq!(left, [1, 2, 3, 4, 5, 6]);
1543 /// assert_eq!(right, []);
1546 #[stable(feature = "rust1", since = "1.0.0")]
1547 #[rustc_const_unstable(feature = "const_slice_split_at_not_mut", issue = "101158")]
1551 pub const fn split_at(&self, mid: usize) -> (&[T], &[T]) {
1552 assert!(mid <= self.len());
1553 // SAFETY: `[ptr; mid]` and `[mid; len]` are inside `self`, which
1554 // fulfills the requirements of `split_at_unchecked`.
1555 unsafe { self.split_at_unchecked(mid) }
1558 /// Divides one mutable slice into two at an index.
1560 /// The first will contain all indices from `[0, mid)` (excluding
1561 /// the index `mid` itself) and the second will contain all
1562 /// indices from `[mid, len)` (excluding the index `len` itself).
1566 /// Panics if `mid > len`.
1571 /// let mut v = [1, 0, 3, 0, 5, 6];
1572 /// let (left, right) = v.split_at_mut(2);
1573 /// assert_eq!(left, [1, 0]);
1574 /// assert_eq!(right, [3, 0, 5, 6]);
1577 /// assert_eq!(v, [1, 2, 3, 4, 5, 6]);
1579 #[stable(feature = "rust1", since = "1.0.0")]
1583 #[rustc_const_unstable(feature = "const_slice_split_at_mut", issue = "101804")]
1584 pub const fn split_at_mut(&mut self, mid: usize) -> (&mut [T], &mut [T]) {
1585 assert!(mid <= self.len());
1586 // SAFETY: `[ptr; mid]` and `[mid; len]` are inside `self`, which
1587 // fulfills the requirements of `from_raw_parts_mut`.
1588 unsafe { self.split_at_mut_unchecked(mid) }
1591 /// Divides one slice into two at an index, without doing bounds checking.
1593 /// The first will contain all indices from `[0, mid)` (excluding
1594 /// the index `mid` itself) and the second will contain all
1595 /// indices from `[mid, len)` (excluding the index `len` itself).
1597 /// For a safe alternative see [`split_at`].
1601 /// Calling this method with an out-of-bounds index is *[undefined behavior]*
1602 /// even if the resulting reference is not used. The caller has to ensure that
1603 /// `0 <= mid <= self.len()`.
1605 /// [`split_at`]: slice::split_at
1606 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
1611 /// #![feature(slice_split_at_unchecked)]
1613 /// let v = [1, 2, 3, 4, 5, 6];
1616 /// let (left, right) = v.split_at_unchecked(0);
1617 /// assert_eq!(left, []);
1618 /// assert_eq!(right, [1, 2, 3, 4, 5, 6]);
1622 /// let (left, right) = v.split_at_unchecked(2);
1623 /// assert_eq!(left, [1, 2]);
1624 /// assert_eq!(right, [3, 4, 5, 6]);
1628 /// let (left, right) = v.split_at_unchecked(6);
1629 /// assert_eq!(left, [1, 2, 3, 4, 5, 6]);
1630 /// assert_eq!(right, []);
1633 #[unstable(feature = "slice_split_at_unchecked", reason = "new API", issue = "76014")]
1634 #[rustc_const_unstable(feature = "slice_split_at_unchecked", issue = "76014")]
1637 pub const unsafe fn split_at_unchecked(&self, mid: usize) -> (&[T], &[T]) {
1638 // HACK: the const function `from_raw_parts` is used to make this
1639 // function const; previously the implementation used
1640 // `(self.get_unchecked(..mid), self.get_unchecked(mid..))`
1642 let len = self.len();
1643 let ptr = self.as_ptr();
1645 // SAFETY: Caller has to check that `0 <= mid <= self.len()`
1646 unsafe { (from_raw_parts(ptr, mid), from_raw_parts(ptr.add(mid), len - mid)) }
1649 /// Divides one mutable slice into two at an index, without doing bounds checking.
1651 /// The first will contain all indices from `[0, mid)` (excluding
1652 /// the index `mid` itself) and the second will contain all
1653 /// indices from `[mid, len)` (excluding the index `len` itself).
1655 /// For a safe alternative see [`split_at_mut`].
1659 /// Calling this method with an out-of-bounds index is *[undefined behavior]*
1660 /// even if the resulting reference is not used. The caller has to ensure that
1661 /// `0 <= mid <= self.len()`.
1663 /// [`split_at_mut`]: slice::split_at_mut
1664 /// [undefined behavior]: https://doc.rust-lang.org/reference/behavior-considered-undefined.html
1669 /// #![feature(slice_split_at_unchecked)]
1671 /// let mut v = [1, 0, 3, 0, 5, 6];
1672 /// // scoped to restrict the lifetime of the borrows
1674 /// let (left, right) = v.split_at_mut_unchecked(2);
1675 /// assert_eq!(left, [1, 0]);
1676 /// assert_eq!(right, [3, 0, 5, 6]);
1680 /// assert_eq!(v, [1, 2, 3, 4, 5, 6]);
1682 #[unstable(feature = "slice_split_at_unchecked", reason = "new API", issue = "76014")]
1683 #[rustc_const_unstable(feature = "const_slice_split_at_mut", issue = "101804")]
1686 pub const unsafe fn split_at_mut_unchecked(&mut self, mid: usize) -> (&mut [T], &mut [T]) {
1687 let len = self.len();
1688 let ptr = self.as_mut_ptr();
1690 // SAFETY: Caller has to check that `0 <= mid <= self.len()`.
1692 // `[ptr; mid]` and `[mid; len]` are not overlapping, so returning a mutable reference
1695 assert_unsafe_precondition!((mid: usize, len: usize) => mid <= len);
1696 (from_raw_parts_mut(ptr, mid), from_raw_parts_mut(ptr.add(mid), len - mid))
1700 /// Divides one slice into an array and a remainder slice at an index.
1702 /// The array will contain all indices from `[0, N)` (excluding
1703 /// the index `N` itself) and the slice will contain all
1704 /// indices from `[N, len)` (excluding the index `len` itself).
1708 /// Panics if `N > len`.
1713 /// #![feature(split_array)]
1715 /// let v = &[1, 2, 3, 4, 5, 6][..];
1718 /// let (left, right) = v.split_array_ref::<0>();
1719 /// assert_eq!(left, &[]);
1720 /// assert_eq!(right, [1, 2, 3, 4, 5, 6]);
1724 /// let (left, right) = v.split_array_ref::<2>();
1725 /// assert_eq!(left, &[1, 2]);
1726 /// assert_eq!(right, [3, 4, 5, 6]);
1730 /// let (left, right) = v.split_array_ref::<6>();
1731 /// assert_eq!(left, &[1, 2, 3, 4, 5, 6]);
1732 /// assert_eq!(right, []);
1735 #[unstable(feature = "split_array", reason = "new API", issue = "90091")]
1739 pub fn split_array_ref<const N: usize>(&self) -> (&[T; N], &[T]) {
1740 let (a, b) = self.split_at(N);
1741 // SAFETY: a points to [T; N]? Yes it's [T] of length N (checked by split_at)
1742 unsafe { (&*(a.as_ptr() as *const [T; N]), b) }
1745 /// Divides one mutable slice into an array and a remainder slice at an index.
1747 /// The array will contain all indices from `[0, N)` (excluding
1748 /// the index `N` itself) and the slice will contain all
1749 /// indices from `[N, len)` (excluding the index `len` itself).
1753 /// Panics if `N > len`.
1758 /// #![feature(split_array)]
1760 /// let mut v = &mut [1, 0, 3, 0, 5, 6][..];
1761 /// let (left, right) = v.split_array_mut::<2>();
1762 /// assert_eq!(left, &mut [1, 0]);
1763 /// assert_eq!(right, [3, 0, 5, 6]);
1766 /// assert_eq!(v, [1, 2, 3, 4, 5, 6]);
1768 #[unstable(feature = "split_array", reason = "new API", issue = "90091")]
1772 pub fn split_array_mut<const N: usize>(&mut self) -> (&mut [T; N], &mut [T]) {
1773 let (a, b) = self.split_at_mut(N);
1774 // SAFETY: a points to [T; N]? Yes it's [T] of length N (checked by split_at_mut)
1775 unsafe { (&mut *(a.as_mut_ptr() as *mut [T; N]), b) }
1778 /// Divides one slice into an array and a remainder slice at an index from
1781 /// The slice will contain all indices from `[0, len - N)` (excluding
1782 /// the index `len - N` itself) and the array will contain all
1783 /// indices from `[len - N, len)` (excluding the index `len` itself).
1787 /// Panics if `N > len`.
1792 /// #![feature(split_array)]
1794 /// let v = &[1, 2, 3, 4, 5, 6][..];
1797 /// let (left, right) = v.rsplit_array_ref::<0>();
1798 /// assert_eq!(left, [1, 2, 3, 4, 5, 6]);
1799 /// assert_eq!(right, &[]);
1803 /// let (left, right) = v.rsplit_array_ref::<2>();
1804 /// assert_eq!(left, [1, 2, 3, 4]);
1805 /// assert_eq!(right, &[5, 6]);
1809 /// let (left, right) = v.rsplit_array_ref::<6>();
1810 /// assert_eq!(left, []);
1811 /// assert_eq!(right, &[1, 2, 3, 4, 5, 6]);
1814 #[unstable(feature = "split_array", reason = "new API", issue = "90091")]
1817 pub fn rsplit_array_ref<const N: usize>(&self) -> (&[T], &[T; N]) {
1818 assert!(N <= self.len());
1819 let (a, b) = self.split_at(self.len() - N);
1820 // SAFETY: b points to [T; N]? Yes it's [T] of length N (checked by split_at)
1821 unsafe { (a, &*(b.as_ptr() as *const [T; N])) }
1824 /// Divides one mutable slice into an array and a remainder slice at an
1825 /// index from the end.
1827 /// The slice will contain all indices from `[0, len - N)` (excluding
1828 /// the index `N` itself) and the array will contain all
1829 /// indices from `[len - N, len)` (excluding the index `len` itself).
1833 /// Panics if `N > len`.
1838 /// #![feature(split_array)]
1840 /// let mut v = &mut [1, 0, 3, 0, 5, 6][..];
1841 /// let (left, right) = v.rsplit_array_mut::<4>();
1842 /// assert_eq!(left, [1, 0]);
1843 /// assert_eq!(right, &mut [3, 0, 5, 6]);
1846 /// assert_eq!(v, [1, 2, 3, 4, 5, 6]);
1848 #[unstable(feature = "split_array", reason = "new API", issue = "90091")]
1851 pub fn rsplit_array_mut<const N: usize>(&mut self) -> (&mut [T], &mut [T; N]) {
1852 assert!(N <= self.len());
1853 let (a, b) = self.split_at_mut(self.len() - N);
1854 // SAFETY: b points to [T; N]? Yes it's [T] of length N (checked by split_at_mut)
1855 unsafe { (a, &mut *(b.as_mut_ptr() as *mut [T; N])) }
1858 /// Returns an iterator over subslices separated by elements that match
1859 /// `pred`. The matched element is not contained in the subslices.
1864 /// let slice = [10, 40, 33, 20];
1865 /// let mut iter = slice.split(|num| num % 3 == 0);
1867 /// assert_eq!(iter.next().unwrap(), &[10, 40]);
1868 /// assert_eq!(iter.next().unwrap(), &[20]);
1869 /// assert!(iter.next().is_none());
1872 /// If the first element is matched, an empty slice will be the first item
1873 /// returned by the iterator. Similarly, if the last element in the slice
1874 /// is matched, an empty slice will be the last item returned by the
1878 /// let slice = [10, 40, 33];
1879 /// let mut iter = slice.split(|num| num % 3 == 0);
1881 /// assert_eq!(iter.next().unwrap(), &[10, 40]);
1882 /// assert_eq!(iter.next().unwrap(), &[]);
1883 /// assert!(iter.next().is_none());
1886 /// If two matched elements are directly adjacent, an empty slice will be
1887 /// present between them:
1890 /// let slice = [10, 6, 33, 20];
1891 /// let mut iter = slice.split(|num| num % 3 == 0);
1893 /// assert_eq!(iter.next().unwrap(), &[10]);
1894 /// assert_eq!(iter.next().unwrap(), &[]);
1895 /// assert_eq!(iter.next().unwrap(), &[20]);
1896 /// assert!(iter.next().is_none());
1898 #[stable(feature = "rust1", since = "1.0.0")]
1900 pub fn split<F>(&self, pred: F) -> Split<'_, T, F>
1902 F: FnMut(&T) -> bool,
1904 Split::new(self, pred)
1907 /// Returns an iterator over mutable subslices separated by elements that
1908 /// match `pred`. The matched element is not contained in the subslices.
1913 /// let mut v = [10, 40, 30, 20, 60, 50];
1915 /// for group in v.split_mut(|num| *num % 3 == 0) {
1918 /// assert_eq!(v, [1, 40, 30, 1, 60, 1]);
1920 #[stable(feature = "rust1", since = "1.0.0")]
1922 pub fn split_mut<F>(&mut self, pred: F) -> SplitMut<'_, T, F>
1924 F: FnMut(&T) -> bool,
1926 SplitMut::new(self, pred)
1929 /// Returns an iterator over subslices separated by elements that match
1930 /// `pred`. The matched element is contained in the end of the previous
1931 /// subslice as a terminator.
1936 /// let slice = [10, 40, 33, 20];
1937 /// let mut iter = slice.split_inclusive(|num| num % 3 == 0);
1939 /// assert_eq!(iter.next().unwrap(), &[10, 40, 33]);
1940 /// assert_eq!(iter.next().unwrap(), &[20]);
1941 /// assert!(iter.next().is_none());
1944 /// If the last element of the slice is matched,
1945 /// that element will be considered the terminator of the preceding slice.
1946 /// That slice will be the last item returned by the iterator.
1949 /// let slice = [3, 10, 40, 33];
1950 /// let mut iter = slice.split_inclusive(|num| num % 3 == 0);
1952 /// assert_eq!(iter.next().unwrap(), &[3]);
1953 /// assert_eq!(iter.next().unwrap(), &[10, 40, 33]);
1954 /// assert!(iter.next().is_none());
1956 #[stable(feature = "split_inclusive", since = "1.51.0")]
1958 pub fn split_inclusive<F>(&self, pred: F) -> SplitInclusive<'_, T, F>
1960 F: FnMut(&T) -> bool,
1962 SplitInclusive::new(self, pred)
1965 /// Returns an iterator over mutable subslices separated by elements that
1966 /// match `pred`. The matched element is contained in the previous
1967 /// subslice as a terminator.
1972 /// let mut v = [10, 40, 30, 20, 60, 50];
1974 /// for group in v.split_inclusive_mut(|num| *num % 3 == 0) {
1975 /// let terminator_idx = group.len()-1;
1976 /// group[terminator_idx] = 1;
1978 /// assert_eq!(v, [10, 40, 1, 20, 1, 1]);
1980 #[stable(feature = "split_inclusive", since = "1.51.0")]
1982 pub fn split_inclusive_mut<F>(&mut self, pred: F) -> SplitInclusiveMut<'_, T, F>
1984 F: FnMut(&T) -> bool,
1986 SplitInclusiveMut::new(self, pred)
1989 /// Returns an iterator over subslices separated by elements that match
1990 /// `pred`, starting at the end of the slice and working backwards.
1991 /// The matched element is not contained in the subslices.
1996 /// let slice = [11, 22, 33, 0, 44, 55];
1997 /// let mut iter = slice.rsplit(|num| *num == 0);
1999 /// assert_eq!(iter.next().unwrap(), &[44, 55]);
2000 /// assert_eq!(iter.next().unwrap(), &[11, 22, 33]);
2001 /// assert_eq!(iter.next(), None);
2004 /// As with `split()`, if the first or last element is matched, an empty
2005 /// slice will be the first (or last) item returned by the iterator.
2008 /// let v = &[0, 1, 1, 2, 3, 5, 8];
2009 /// let mut it = v.rsplit(|n| *n % 2 == 0);
2010 /// assert_eq!(it.next().unwrap(), &[]);
2011 /// assert_eq!(it.next().unwrap(), &[3, 5]);
2012 /// assert_eq!(it.next().unwrap(), &[1, 1]);
2013 /// assert_eq!(it.next().unwrap(), &[]);
2014 /// assert_eq!(it.next(), None);
2016 #[stable(feature = "slice_rsplit", since = "1.27.0")]
2018 pub fn rsplit<F>(&self, pred: F) -> RSplit<'_, T, F>
2020 F: FnMut(&T) -> bool,
2022 RSplit::new(self, pred)
2025 /// Returns an iterator over mutable subslices separated by elements that
2026 /// match `pred`, starting at the end of the slice and working
2027 /// backwards. The matched element is not contained in the subslices.
2032 /// let mut v = [100, 400, 300, 200, 600, 500];
2034 /// let mut count = 0;
2035 /// for group in v.rsplit_mut(|num| *num % 3 == 0) {
2037 /// group[0] = count;
2039 /// assert_eq!(v, [3, 400, 300, 2, 600, 1]);
2042 #[stable(feature = "slice_rsplit", since = "1.27.0")]
2044 pub fn rsplit_mut<F>(&mut self, pred: F) -> RSplitMut<'_, T, F>
2046 F: FnMut(&T) -> bool,
2048 RSplitMut::new(self, pred)
2051 /// Returns an iterator over subslices separated by elements that match
2052 /// `pred`, limited to returning at most `n` items. The matched element is
2053 /// not contained in the subslices.
2055 /// The last element returned, if any, will contain the remainder of the
2060 /// Print the slice split once by numbers divisible by 3 (i.e., `[10, 40]`,
2061 /// `[20, 60, 50]`):
2064 /// let v = [10, 40, 30, 20, 60, 50];
2066 /// for group in v.splitn(2, |num| *num % 3 == 0) {
2067 /// println!("{group:?}");
2070 #[stable(feature = "rust1", since = "1.0.0")]
2072 pub fn splitn<F>(&self, n: usize, pred: F) -> SplitN<'_, T, F>
2074 F: FnMut(&T) -> bool,
2076 SplitN::new(self.split(pred), n)
2079 /// Returns an iterator over mutable subslices separated by elements that match
2080 /// `pred`, limited to returning at most `n` items. The matched element is
2081 /// not contained in the subslices.
2083 /// The last element returned, if any, will contain the remainder of the
2089 /// let mut v = [10, 40, 30, 20, 60, 50];
2091 /// for group in v.splitn_mut(2, |num| *num % 3 == 0) {
2094 /// assert_eq!(v, [1, 40, 30, 1, 60, 50]);
2096 #[stable(feature = "rust1", since = "1.0.0")]
2098 pub fn splitn_mut<F>(&mut self, n: usize, pred: F) -> SplitNMut<'_, T, F>
2100 F: FnMut(&T) -> bool,
2102 SplitNMut::new(self.split_mut(pred), n)
2105 /// Returns an iterator over subslices separated by elements that match
2106 /// `pred` limited to returning at most `n` items. This starts at the end of
2107 /// the slice and works backwards. The matched element is not contained in
2110 /// The last element returned, if any, will contain the remainder of the
2115 /// Print the slice split once, starting from the end, by numbers divisible
2116 /// by 3 (i.e., `[50]`, `[10, 40, 30, 20]`):
2119 /// let v = [10, 40, 30, 20, 60, 50];
2121 /// for group in v.rsplitn(2, |num| *num % 3 == 0) {
2122 /// println!("{group:?}");
2125 #[stable(feature = "rust1", since = "1.0.0")]
2127 pub fn rsplitn<F>(&self, n: usize, pred: F) -> RSplitN<'_, T, F>
2129 F: FnMut(&T) -> bool,
2131 RSplitN::new(self.rsplit(pred), n)
2134 /// Returns an iterator over subslices separated by elements that match
2135 /// `pred` limited to returning at most `n` items. This starts at the end of
2136 /// the slice and works backwards. The matched element is not contained in
2139 /// The last element returned, if any, will contain the remainder of the
2145 /// let mut s = [10, 40, 30, 20, 60, 50];
2147 /// for group in s.rsplitn_mut(2, |num| *num % 3 == 0) {
2150 /// assert_eq!(s, [1, 40, 30, 20, 60, 1]);
2152 #[stable(feature = "rust1", since = "1.0.0")]
2154 pub fn rsplitn_mut<F>(&mut self, n: usize, pred: F) -> RSplitNMut<'_, T, F>
2156 F: FnMut(&T) -> bool,
2158 RSplitNMut::new(self.rsplit_mut(pred), n)
2161 /// Returns `true` if the slice contains an element with the given value.
2163 /// This operation is *O*(*n*).
2165 /// Note that if you have a sorted slice, [`binary_search`] may be faster.
2167 /// [`binary_search`]: slice::binary_search
2172 /// let v = [10, 40, 30];
2173 /// assert!(v.contains(&30));
2174 /// assert!(!v.contains(&50));
2177 /// If you do not have a `&T`, but some other value that you can compare
2178 /// with one (for example, `String` implements `PartialEq<str>`), you can
2179 /// use `iter().any`:
2182 /// let v = [String::from("hello"), String::from("world")]; // slice of `String`
2183 /// assert!(v.iter().any(|e| e == "hello")); // search with `&str`
2184 /// assert!(!v.iter().any(|e| e == "hi"));
2186 #[stable(feature = "rust1", since = "1.0.0")]
2189 pub fn contains(&self, x: &T) -> bool
2193 cmp::SliceContains::slice_contains(x, self)
2196 /// Returns `true` if `needle` is a prefix of the slice.
2201 /// let v = [10, 40, 30];
2202 /// assert!(v.starts_with(&[10]));
2203 /// assert!(v.starts_with(&[10, 40]));
2204 /// assert!(!v.starts_with(&[50]));
2205 /// assert!(!v.starts_with(&[10, 50]));
2208 /// Always returns `true` if `needle` is an empty slice:
2211 /// let v = &[10, 40, 30];
2212 /// assert!(v.starts_with(&[]));
2213 /// let v: &[u8] = &[];
2214 /// assert!(v.starts_with(&[]));
2216 #[stable(feature = "rust1", since = "1.0.0")]
2218 pub fn starts_with(&self, needle: &[T]) -> bool
2222 let n = needle.len();
2223 self.len() >= n && needle == &self[..n]
2226 /// Returns `true` if `needle` is a suffix of the slice.
2231 /// let v = [10, 40, 30];
2232 /// assert!(v.ends_with(&[30]));
2233 /// assert!(v.ends_with(&[40, 30]));
2234 /// assert!(!v.ends_with(&[50]));
2235 /// assert!(!v.ends_with(&[50, 30]));
2238 /// Always returns `true` if `needle` is an empty slice:
2241 /// let v = &[10, 40, 30];
2242 /// assert!(v.ends_with(&[]));
2243 /// let v: &[u8] = &[];
2244 /// assert!(v.ends_with(&[]));
2246 #[stable(feature = "rust1", since = "1.0.0")]
2248 pub fn ends_with(&self, needle: &[T]) -> bool
2252 let (m, n) = (self.len(), needle.len());
2253 m >= n && needle == &self[m - n..]
2256 /// Returns a subslice with the prefix removed.
2258 /// If the slice starts with `prefix`, returns the subslice after the prefix, wrapped in `Some`.
2259 /// If `prefix` is empty, simply returns the original slice.
2261 /// If the slice does not start with `prefix`, returns `None`.
2266 /// let v = &[10, 40, 30];
2267 /// assert_eq!(v.strip_prefix(&[10]), Some(&[40, 30][..]));
2268 /// assert_eq!(v.strip_prefix(&[10, 40]), Some(&[30][..]));
2269 /// assert_eq!(v.strip_prefix(&[50]), None);
2270 /// assert_eq!(v.strip_prefix(&[10, 50]), None);
2272 /// let prefix : &str = "he";
2273 /// assert_eq!(b"hello".strip_prefix(prefix.as_bytes()),
2274 /// Some(b"llo".as_ref()));
2276 #[must_use = "returns the subslice without modifying the original"]
2277 #[stable(feature = "slice_strip", since = "1.51.0")]
2278 pub fn strip_prefix<P: SlicePattern<Item = T> + ?Sized>(&self, prefix: &P) -> Option<&[T]>
2282 // This function will need rewriting if and when SlicePattern becomes more sophisticated.
2283 let prefix = prefix.as_slice();
2284 let n = prefix.len();
2285 if n <= self.len() {
2286 let (head, tail) = self.split_at(n);
2294 /// Returns a subslice with the suffix removed.
2296 /// If the slice ends with `suffix`, returns the subslice before the suffix, wrapped in `Some`.
2297 /// If `suffix` is empty, simply returns the original slice.
2299 /// If the slice does not end with `suffix`, returns `None`.
2304 /// let v = &[10, 40, 30];
2305 /// assert_eq!(v.strip_suffix(&[30]), Some(&[10, 40][..]));
2306 /// assert_eq!(v.strip_suffix(&[40, 30]), Some(&[10][..]));
2307 /// assert_eq!(v.strip_suffix(&[50]), None);
2308 /// assert_eq!(v.strip_suffix(&[50, 30]), None);
2310 #[must_use = "returns the subslice without modifying the original"]
2311 #[stable(feature = "slice_strip", since = "1.51.0")]
2312 pub fn strip_suffix<P: SlicePattern<Item = T> + ?Sized>(&self, suffix: &P) -> Option<&[T]>
2316 // This function will need rewriting if and when SlicePattern becomes more sophisticated.
2317 let suffix = suffix.as_slice();
2318 let (len, n) = (self.len(), suffix.len());
2320 let (head, tail) = self.split_at(len - n);
2328 /// Binary searches this slice for a given element.
2329 /// This behaves similarly to [`contains`] if this slice is sorted.
2331 /// If the value is found then [`Result::Ok`] is returned, containing the
2332 /// index of the matching element. If there are multiple matches, then any
2333 /// one of the matches could be returned. The index is chosen
2334 /// deterministically, but is subject to change in future versions of Rust.
2335 /// If the value is not found then [`Result::Err`] is returned, containing
2336 /// the index where a matching element could be inserted while maintaining
2339 /// See also [`binary_search_by`], [`binary_search_by_key`], and [`partition_point`].
2341 /// [`contains`]: slice::contains
2342 /// [`binary_search_by`]: slice::binary_search_by
2343 /// [`binary_search_by_key`]: slice::binary_search_by_key
2344 /// [`partition_point`]: slice::partition_point
2348 /// Looks up a series of four elements. The first is found, with a
2349 /// uniquely determined position; the second and third are not
2350 /// found; the fourth could match any position in `[1, 4]`.
2353 /// let s = [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
2355 /// assert_eq!(s.binary_search(&13), Ok(9));
2356 /// assert_eq!(s.binary_search(&4), Err(7));
2357 /// assert_eq!(s.binary_search(&100), Err(13));
2358 /// let r = s.binary_search(&1);
2359 /// assert!(match r { Ok(1..=4) => true, _ => false, });
2362 /// If you want to find that whole *range* of matching items, rather than
2363 /// an arbitrary matching one, that can be done using [`partition_point`]:
2365 /// let s = [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
2367 /// let low = s.partition_point(|x| x < &1);
2368 /// assert_eq!(low, 1);
2369 /// let high = s.partition_point(|x| x <= &1);
2370 /// assert_eq!(high, 5);
2371 /// let r = s.binary_search(&1);
2372 /// assert!((low..high).contains(&r.unwrap()));
2374 /// assert!(s[..low].iter().all(|&x| x < 1));
2375 /// assert!(s[low..high].iter().all(|&x| x == 1));
2376 /// assert!(s[high..].iter().all(|&x| x > 1));
2378 /// // For something not found, the "range" of equal items is empty
2379 /// assert_eq!(s.partition_point(|x| x < &11), 9);
2380 /// assert_eq!(s.partition_point(|x| x <= &11), 9);
2381 /// assert_eq!(s.binary_search(&11), Err(9));
2384 /// If you want to insert an item to a sorted vector, while maintaining
2385 /// sort order, consider using [`partition_point`]:
2388 /// let mut s = vec![0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
2390 /// let idx = s.partition_point(|&x| x < num);
2391 /// // The above is equivalent to `let idx = s.binary_search(&num).unwrap_or_else(|x| x);`
2392 /// s.insert(idx, num);
2393 /// assert_eq!(s, [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 42, 55]);
2395 #[stable(feature = "rust1", since = "1.0.0")]
2396 pub fn binary_search(&self, x: &T) -> Result<usize, usize>
2400 self.binary_search_by(|p| p.cmp(x))
2403 /// Binary searches this slice with a comparator function.
2404 /// This behaves similarly to [`contains`] if this slice is sorted.
2406 /// The comparator function should implement an order consistent
2407 /// with the sort order of the underlying slice, returning an
2408 /// order code that indicates whether its argument is `Less`,
2409 /// `Equal` or `Greater` the desired target.
2411 /// If the value is found then [`Result::Ok`] is returned, containing the
2412 /// index of the matching element. If there are multiple matches, then any
2413 /// one of the matches could be returned. The index is chosen
2414 /// deterministically, but is subject to change in future versions of Rust.
2415 /// If the value is not found then [`Result::Err`] is returned, containing
2416 /// the index where a matching element could be inserted while maintaining
2419 /// See also [`binary_search`], [`binary_search_by_key`], and [`partition_point`].
2421 /// [`contains`]: slice::contains
2422 /// [`binary_search`]: slice::binary_search
2423 /// [`binary_search_by_key`]: slice::binary_search_by_key
2424 /// [`partition_point`]: slice::partition_point
2428 /// Looks up a series of four elements. The first is found, with a
2429 /// uniquely determined position; the second and third are not
2430 /// found; the fourth could match any position in `[1, 4]`.
2433 /// let s = [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
2436 /// assert_eq!(s.binary_search_by(|probe| probe.cmp(&seek)), Ok(9));
2438 /// assert_eq!(s.binary_search_by(|probe| probe.cmp(&seek)), Err(7));
2440 /// assert_eq!(s.binary_search_by(|probe| probe.cmp(&seek)), Err(13));
2442 /// let r = s.binary_search_by(|probe| probe.cmp(&seek));
2443 /// assert!(match r { Ok(1..=4) => true, _ => false, });
2445 #[stable(feature = "rust1", since = "1.0.0")]
2447 pub fn binary_search_by<'a, F>(&'a self, mut f: F) -> Result<usize, usize>
2449 F: FnMut(&'a T) -> Ordering,
2452 // - 0 <= left <= left + size = right <= self.len()
2453 // - f returns Less for everything in self[..left]
2454 // - f returns Greater for everything in self[right..]
2455 let mut size = self.len();
2457 let mut right = size;
2458 while left < right {
2459 let mid = left + size / 2;
2461 // SAFETY: the while condition means `size` is strictly positive, so
2462 // `size/2 < size`. Thus `left + size/2 < left + size`, which
2463 // coupled with the `left + size <= self.len()` invariant means
2464 // we have `left + size/2 < self.len()`, and this is in-bounds.
2465 let cmp = f(unsafe { self.get_unchecked(mid) });
2467 // The reason why we use if/else control flow rather than match
2468 // is because match reorders comparison operations, which is perf sensitive.
2469 // This is x86 asm for u8: https://rust.godbolt.org/z/8Y8Pra.
2472 } else if cmp == Greater {
2475 // SAFETY: same as the `get_unchecked` above
2476 unsafe { crate::intrinsics::assume(mid < self.len()) };
2480 size = right - left;
2483 // SAFETY: directly true from the overall invariant.
2484 // Note that this is `<=`, unlike the assume in the `Ok` path.
2485 unsafe { crate::intrinsics::assume(left <= self.len()) };
2489 /// Binary searches this slice with a key extraction function.
2490 /// This behaves similarly to [`contains`] if this slice is sorted.
2492 /// Assumes that the slice is sorted by the key, for instance with
2493 /// [`sort_by_key`] using the same key extraction function.
2495 /// If the value is found then [`Result::Ok`] is returned, containing the
2496 /// index of the matching element. If there are multiple matches, then any
2497 /// one of the matches could be returned. The index is chosen
2498 /// deterministically, but is subject to change in future versions of Rust.
2499 /// If the value is not found then [`Result::Err`] is returned, containing
2500 /// the index where a matching element could be inserted while maintaining
2503 /// See also [`binary_search`], [`binary_search_by`], and [`partition_point`].
2505 /// [`contains`]: slice::contains
2506 /// [`sort_by_key`]: slice::sort_by_key
2507 /// [`binary_search`]: slice::binary_search
2508 /// [`binary_search_by`]: slice::binary_search_by
2509 /// [`partition_point`]: slice::partition_point
2513 /// Looks up a series of four elements in a slice of pairs sorted by
2514 /// their second elements. The first is found, with a uniquely
2515 /// determined position; the second and third are not found; the
2516 /// fourth could match any position in `[1, 4]`.
2519 /// let s = [(0, 0), (2, 1), (4, 1), (5, 1), (3, 1),
2520 /// (1, 2), (2, 3), (4, 5), (5, 8), (3, 13),
2521 /// (1, 21), (2, 34), (4, 55)];
2523 /// assert_eq!(s.binary_search_by_key(&13, |&(a, b)| b), Ok(9));
2524 /// assert_eq!(s.binary_search_by_key(&4, |&(a, b)| b), Err(7));
2525 /// assert_eq!(s.binary_search_by_key(&100, |&(a, b)| b), Err(13));
2526 /// let r = s.binary_search_by_key(&1, |&(a, b)| b);
2527 /// assert!(match r { Ok(1..=4) => true, _ => false, });
2529 // Lint rustdoc::broken_intra_doc_links is allowed as `slice::sort_by_key` is
2530 // in crate `alloc`, and as such doesn't exists yet when building `core`: #74481.
2531 // This breaks links when slice is displayed in core, but changing it to use relative links
2532 // would break when the item is re-exported. So allow the core links to be broken for now.
2533 #[allow(rustdoc::broken_intra_doc_links)]
2534 #[stable(feature = "slice_binary_search_by_key", since = "1.10.0")]
2536 pub fn binary_search_by_key<'a, B, F>(&'a self, b: &B, mut f: F) -> Result<usize, usize>
2538 F: FnMut(&'a T) -> B,
2541 self.binary_search_by(|k| f(k).cmp(b))
2544 /// Sorts the slice, but might not preserve the order of equal elements.
2546 /// This sort is unstable (i.e., may reorder equal elements), in-place
2547 /// (i.e., does not allocate), and *O*(*n* \* log(*n*)) worst-case.
2549 /// # Current implementation
2551 /// The current algorithm is based on [pattern-defeating quicksort][pdqsort] by Orson Peters,
2552 /// which combines the fast average case of randomized quicksort with the fast worst case of
2553 /// heapsort, while achieving linear time on slices with certain patterns. It uses some
2554 /// randomization to avoid degenerate cases, but with a fixed seed to always provide
2555 /// deterministic behavior.
2557 /// It is typically faster than stable sorting, except in a few special cases, e.g., when the
2558 /// slice consists of several concatenated sorted sequences.
2563 /// let mut v = [-5, 4, 1, -3, 2];
2565 /// v.sort_unstable();
2566 /// assert!(v == [-5, -3, 1, 2, 4]);
2569 /// [pdqsort]: https://github.com/orlp/pdqsort
2570 #[stable(feature = "sort_unstable", since = "1.20.0")]
2572 pub fn sort_unstable(&mut self)
2576 sort::quicksort(self, T::lt);
2579 /// Sorts the slice with a comparator function, but might not preserve the order of equal
2582 /// This sort is unstable (i.e., may reorder equal elements), in-place
2583 /// (i.e., does not allocate), and *O*(*n* \* log(*n*)) worst-case.
2585 /// The comparator function must define a total ordering for the elements in the slice. If
2586 /// the ordering is not total, the order of the elements is unspecified. An order is a
2587 /// total order if it is (for all `a`, `b` and `c`):
2589 /// * total and antisymmetric: exactly one of `a < b`, `a == b` or `a > b` is true, and
2590 /// * transitive, `a < b` and `b < c` implies `a < c`. The same must hold for both `==` and `>`.
2592 /// For example, while [`f64`] doesn't implement [`Ord`] because `NaN != NaN`, we can use
2593 /// `partial_cmp` as our sort function when we know the slice doesn't contain a `NaN`.
2596 /// let mut floats = [5f64, 4.0, 1.0, 3.0, 2.0];
2597 /// floats.sort_unstable_by(|a, b| a.partial_cmp(b).unwrap());
2598 /// assert_eq!(floats, [1.0, 2.0, 3.0, 4.0, 5.0]);
2601 /// # Current implementation
2603 /// The current algorithm is based on [pattern-defeating quicksort][pdqsort] by Orson Peters,
2604 /// which combines the fast average case of randomized quicksort with the fast worst case of
2605 /// heapsort, while achieving linear time on slices with certain patterns. It uses some
2606 /// randomization to avoid degenerate cases, but with a fixed seed to always provide
2607 /// deterministic behavior.
2609 /// It is typically faster than stable sorting, except in a few special cases, e.g., when the
2610 /// slice consists of several concatenated sorted sequences.
2615 /// let mut v = [5, 4, 1, 3, 2];
2616 /// v.sort_unstable_by(|a, b| a.cmp(b));
2617 /// assert!(v == [1, 2, 3, 4, 5]);
2619 /// // reverse sorting
2620 /// v.sort_unstable_by(|a, b| b.cmp(a));
2621 /// assert!(v == [5, 4, 3, 2, 1]);
2624 /// [pdqsort]: https://github.com/orlp/pdqsort
2625 #[stable(feature = "sort_unstable", since = "1.20.0")]
2627 pub fn sort_unstable_by<F>(&mut self, mut compare: F)
2629 F: FnMut(&T, &T) -> Ordering,
2631 sort::quicksort(self, |a, b| compare(a, b) == Ordering::Less);
2634 /// Sorts the slice with a key extraction function, but might not preserve the order of equal
2637 /// This sort is unstable (i.e., may reorder equal elements), in-place
2638 /// (i.e., does not allocate), and *O*(m \* *n* \* log(*n*)) worst-case, where the key function is
2641 /// # Current implementation
2643 /// The current algorithm is based on [pattern-defeating quicksort][pdqsort] by Orson Peters,
2644 /// which combines the fast average case of randomized quicksort with the fast worst case of
2645 /// heapsort, while achieving linear time on slices with certain patterns. It uses some
2646 /// randomization to avoid degenerate cases, but with a fixed seed to always provide
2647 /// deterministic behavior.
2649 /// Due to its key calling strategy, [`sort_unstable_by_key`](#method.sort_unstable_by_key)
2650 /// is likely to be slower than [`sort_by_cached_key`](#method.sort_by_cached_key) in
2651 /// cases where the key function is expensive.
2656 /// let mut v = [-5i32, 4, 1, -3, 2];
2658 /// v.sort_unstable_by_key(|k| k.abs());
2659 /// assert!(v == [1, 2, -3, 4, -5]);
2662 /// [pdqsort]: https://github.com/orlp/pdqsort
2663 #[stable(feature = "sort_unstable", since = "1.20.0")]
2665 pub fn sort_unstable_by_key<K, F>(&mut self, mut f: F)
2670 sort::quicksort(self, |a, b| f(a).lt(&f(b)));
2673 /// Reorder the slice such that the element at `index` is at its final sorted position.
2675 /// This reordering has the additional property that any value at position `i < index` will be
2676 /// less than or equal to any value at a position `j > index`. Additionally, this reordering is
2677 /// unstable (i.e. any number of equal elements may end up at position `index`), in-place
2678 /// (i.e. does not allocate), and *O*(*n*) worst-case. This function is also/ known as "kth
2679 /// element" in other libraries. It returns a triplet of the following from the reordered slice:
2680 /// the subslice prior to `index`, the element at `index`, and the subslice after `index`;
2681 /// accordingly, the values in those two subslices will respectively all be less-than-or-equal-to
2682 /// and greater-than-or-equal-to the value of the element at `index`.
2684 /// # Current implementation
2686 /// The current algorithm is based on the quickselect portion of the same quicksort algorithm
2687 /// used for [`sort_unstable`].
2689 /// [`sort_unstable`]: slice::sort_unstable
2693 /// Panics when `index >= len()`, meaning it always panics on empty slices.
2698 /// let mut v = [-5i32, 4, 1, -3, 2];
2700 /// // Find the median
2701 /// v.select_nth_unstable(2);
2703 /// // We are only guaranteed the slice will be one of the following, based on the way we sort
2704 /// // about the specified index.
2705 /// assert!(v == [-3, -5, 1, 2, 4] ||
2706 /// v == [-5, -3, 1, 2, 4] ||
2707 /// v == [-3, -5, 1, 4, 2] ||
2708 /// v == [-5, -3, 1, 4, 2]);
2710 #[stable(feature = "slice_select_nth_unstable", since = "1.49.0")]
2712 pub fn select_nth_unstable(&mut self, index: usize) -> (&mut [T], &mut T, &mut [T])
2716 sort::partition_at_index(self, index, T::lt)
2719 /// Reorder the slice with a comparator function such that the element at `index` is at its
2720 /// final sorted position.
2722 /// This reordering has the additional property that any value at position `i < index` will be
2723 /// less than or equal to any value at a position `j > index` using the comparator function.
2724 /// Additionally, this reordering is unstable (i.e. any number of equal elements may end up at
2725 /// position `index`), in-place (i.e. does not allocate), and *O*(*n*) worst-case. This function
2726 /// is also known as "kth element" in other libraries. It returns a triplet of the following from
2727 /// the slice reordered according to the provided comparator function: the subslice prior to
2728 /// `index`, the element at `index`, and the subslice after `index`; accordingly, the values in
2729 /// those two subslices will respectively all be less-than-or-equal-to and greater-than-or-equal-to
2730 /// the value of the element at `index`.
2732 /// # Current implementation
2734 /// The current algorithm is based on the quickselect portion of the same quicksort algorithm
2735 /// used for [`sort_unstable`].
2737 /// [`sort_unstable`]: slice::sort_unstable
2741 /// Panics when `index >= len()`, meaning it always panics on empty slices.
2746 /// let mut v = [-5i32, 4, 1, -3, 2];
2748 /// // Find the median as if the slice were sorted in descending order.
2749 /// v.select_nth_unstable_by(2, |a, b| b.cmp(a));
2751 /// // We are only guaranteed the slice will be one of the following, based on the way we sort
2752 /// // about the specified index.
2753 /// assert!(v == [2, 4, 1, -5, -3] ||
2754 /// v == [2, 4, 1, -3, -5] ||
2755 /// v == [4, 2, 1, -5, -3] ||
2756 /// v == [4, 2, 1, -3, -5]);
2758 #[stable(feature = "slice_select_nth_unstable", since = "1.49.0")]
2760 pub fn select_nth_unstable_by<F>(
2764 ) -> (&mut [T], &mut T, &mut [T])
2766 F: FnMut(&T, &T) -> Ordering,
2768 sort::partition_at_index(self, index, |a: &T, b: &T| compare(a, b) == Less)
2771 /// Reorder the slice with a key extraction function such that the element at `index` is at its
2772 /// final sorted position.
2774 /// This reordering has the additional property that any value at position `i < index` will be
2775 /// less than or equal to any value at a position `j > index` using the key extraction function.
2776 /// Additionally, this reordering is unstable (i.e. any number of equal elements may end up at
2777 /// position `index`), in-place (i.e. does not allocate), and *O*(*n*) worst-case. This function
2778 /// is also known as "kth element" in other libraries. It returns a triplet of the following from
2779 /// the slice reordered according to the provided key extraction function: the subslice prior to
2780 /// `index`, the element at `index`, and the subslice after `index`; accordingly, the values in
2781 /// those two subslices will respectively all be less-than-or-equal-to and greater-than-or-equal-to
2782 /// the value of the element at `index`.
2784 /// # Current implementation
2786 /// The current algorithm is based on the quickselect portion of the same quicksort algorithm
2787 /// used for [`sort_unstable`].
2789 /// [`sort_unstable`]: slice::sort_unstable
2793 /// Panics when `index >= len()`, meaning it always panics on empty slices.
2798 /// let mut v = [-5i32, 4, 1, -3, 2];
2800 /// // Return the median as if the array were sorted according to absolute value.
2801 /// v.select_nth_unstable_by_key(2, |a| a.abs());
2803 /// // We are only guaranteed the slice will be one of the following, based on the way we sort
2804 /// // about the specified index.
2805 /// assert!(v == [1, 2, -3, 4, -5] ||
2806 /// v == [1, 2, -3, -5, 4] ||
2807 /// v == [2, 1, -3, 4, -5] ||
2808 /// v == [2, 1, -3, -5, 4]);
2810 #[stable(feature = "slice_select_nth_unstable", since = "1.49.0")]
2812 pub fn select_nth_unstable_by_key<K, F>(
2816 ) -> (&mut [T], &mut T, &mut [T])
2821 sort::partition_at_index(self, index, |a: &T, b: &T| f(a).lt(&f(b)))
2824 /// Moves all consecutive repeated elements to the end of the slice according to the
2825 /// [`PartialEq`] trait implementation.
2827 /// Returns two slices. The first contains no consecutive repeated elements.
2828 /// The second contains all the duplicates in no specified order.
2830 /// If the slice is sorted, the first returned slice contains no duplicates.
2835 /// #![feature(slice_partition_dedup)]
2837 /// let mut slice = [1, 2, 2, 3, 3, 2, 1, 1];
2839 /// let (dedup, duplicates) = slice.partition_dedup();
2841 /// assert_eq!(dedup, [1, 2, 3, 2, 1]);
2842 /// assert_eq!(duplicates, [2, 3, 1]);
2844 #[unstable(feature = "slice_partition_dedup", issue = "54279")]
2846 pub fn partition_dedup(&mut self) -> (&mut [T], &mut [T])
2850 self.partition_dedup_by(|a, b| a == b)
2853 /// Moves all but the first of consecutive elements to the end of the slice satisfying
2854 /// a given equality relation.
2856 /// Returns two slices. The first contains no consecutive repeated elements.
2857 /// The second contains all the duplicates in no specified order.
2859 /// The `same_bucket` function is passed references to two elements from the slice and
2860 /// must determine if the elements compare equal. The elements are passed in opposite order
2861 /// from their order in the slice, so if `same_bucket(a, b)` returns `true`, `a` is moved
2862 /// at the end of the slice.
2864 /// If the slice is sorted, the first returned slice contains no duplicates.
2869 /// #![feature(slice_partition_dedup)]
2871 /// let mut slice = ["foo", "Foo", "BAZ", "Bar", "bar", "baz", "BAZ"];
2873 /// let (dedup, duplicates) = slice.partition_dedup_by(|a, b| a.eq_ignore_ascii_case(b));
2875 /// assert_eq!(dedup, ["foo", "BAZ", "Bar", "baz"]);
2876 /// assert_eq!(duplicates, ["bar", "Foo", "BAZ"]);
2878 #[unstable(feature = "slice_partition_dedup", issue = "54279")]
2880 pub fn partition_dedup_by<F>(&mut self, mut same_bucket: F) -> (&mut [T], &mut [T])
2882 F: FnMut(&mut T, &mut T) -> bool,
2884 // Although we have a mutable reference to `self`, we cannot make
2885 // *arbitrary* changes. The `same_bucket` calls could panic, so we
2886 // must ensure that the slice is in a valid state at all times.
2888 // The way that we handle this is by using swaps; we iterate
2889 // over all the elements, swapping as we go so that at the end
2890 // the elements we wish to keep are in the front, and those we
2891 // wish to reject are at the back. We can then split the slice.
2892 // This operation is still `O(n)`.
2894 // Example: We start in this state, where `r` represents "next
2895 // read" and `w` represents "next_write`.
2898 // +---+---+---+---+---+---+
2899 // | 0 | 1 | 1 | 2 | 3 | 3 |
2900 // +---+---+---+---+---+---+
2903 // Comparing self[r] against self[w-1], this is not a duplicate, so
2904 // we swap self[r] and self[w] (no effect as r==w) and then increment both
2905 // r and w, leaving us with:
2908 // +---+---+---+---+---+---+
2909 // | 0 | 1 | 1 | 2 | 3 | 3 |
2910 // +---+---+---+---+---+---+
2913 // Comparing self[r] against self[w-1], this value is a duplicate,
2914 // so we increment `r` but leave everything else unchanged:
2917 // +---+---+---+---+---+---+
2918 // | 0 | 1 | 1 | 2 | 3 | 3 |
2919 // +---+---+---+---+---+---+
2922 // Comparing self[r] against self[w-1], this is not a duplicate,
2923 // so swap self[r] and self[w] and advance r and w:
2926 // +---+---+---+---+---+---+
2927 // | 0 | 1 | 2 | 1 | 3 | 3 |
2928 // +---+---+---+---+---+---+
2931 // Not a duplicate, repeat:
2934 // +---+---+---+---+---+---+
2935 // | 0 | 1 | 2 | 3 | 1 | 3 |
2936 // +---+---+---+---+---+---+
2939 // Duplicate, advance r. End of slice. Split at w.
2941 let len = self.len();
2943 return (self, &mut []);
2946 let ptr = self.as_mut_ptr();
2947 let mut next_read: usize = 1;
2948 let mut next_write: usize = 1;
2950 // SAFETY: the `while` condition guarantees `next_read` and `next_write`
2951 // are less than `len`, thus are inside `self`. `prev_ptr_write` points to
2952 // one element before `ptr_write`, but `next_write` starts at 1, so
2953 // `prev_ptr_write` is never less than 0 and is inside the slice.
2954 // This fulfils the requirements for dereferencing `ptr_read`, `prev_ptr_write`
2955 // and `ptr_write`, and for using `ptr.add(next_read)`, `ptr.add(next_write - 1)`
2956 // and `prev_ptr_write.offset(1)`.
2958 // `next_write` is also incremented at most once per loop at most meaning
2959 // no element is skipped when it may need to be swapped.
2961 // `ptr_read` and `prev_ptr_write` never point to the same element. This
2962 // is required for `&mut *ptr_read`, `&mut *prev_ptr_write` to be safe.
2963 // The explanation is simply that `next_read >= next_write` is always true,
2964 // thus `next_read > next_write - 1` is too.
2966 // Avoid bounds checks by using raw pointers.
2967 while next_read < len {
2968 let ptr_read = ptr.add(next_read);
2969 let prev_ptr_write = ptr.add(next_write - 1);
2970 if !same_bucket(&mut *ptr_read, &mut *prev_ptr_write) {
2971 if next_read != next_write {
2972 let ptr_write = prev_ptr_write.add(1);
2973 mem::swap(&mut *ptr_read, &mut *ptr_write);
2981 self.split_at_mut(next_write)
2984 /// Moves all but the first of consecutive elements to the end of the slice that resolve
2985 /// to the same key.
2987 /// Returns two slices. The first contains no consecutive repeated elements.
2988 /// The second contains all the duplicates in no specified order.
2990 /// If the slice is sorted, the first returned slice contains no duplicates.
2995 /// #![feature(slice_partition_dedup)]
2997 /// let mut slice = [10, 20, 21, 30, 30, 20, 11, 13];
2999 /// let (dedup, duplicates) = slice.partition_dedup_by_key(|i| *i / 10);
3001 /// assert_eq!(dedup, [10, 20, 30, 20, 11]);
3002 /// assert_eq!(duplicates, [21, 30, 13]);
3004 #[unstable(feature = "slice_partition_dedup", issue = "54279")]
3006 pub fn partition_dedup_by_key<K, F>(&mut self, mut key: F) -> (&mut [T], &mut [T])
3008 F: FnMut(&mut T) -> K,
3011 self.partition_dedup_by(|a, b| key(a) == key(b))
3014 /// Rotates the slice in-place such that the first `mid` elements of the
3015 /// slice move to the end while the last `self.len() - mid` elements move to
3016 /// the front. After calling `rotate_left`, the element previously at index
3017 /// `mid` will become the first element in the slice.
3021 /// This function will panic if `mid` is greater than the length of the
3022 /// slice. Note that `mid == self.len()` does _not_ panic and is a no-op
3027 /// Takes linear (in `self.len()`) time.
3032 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3033 /// a.rotate_left(2);
3034 /// assert_eq!(a, ['c', 'd', 'e', 'f', 'a', 'b']);
3037 /// Rotating a subslice:
3040 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3041 /// a[1..5].rotate_left(1);
3042 /// assert_eq!(a, ['a', 'c', 'd', 'e', 'b', 'f']);
3044 #[stable(feature = "slice_rotate", since = "1.26.0")]
3045 pub fn rotate_left(&mut self, mid: usize) {
3046 assert!(mid <= self.len());
3047 let k = self.len() - mid;
3048 let p = self.as_mut_ptr();
3050 // SAFETY: The range `[p.add(mid) - mid, p.add(mid) + k)` is trivially
3051 // valid for reading and writing, as required by `ptr_rotate`.
3053 rotate::ptr_rotate(mid, p.add(mid), k);
3057 /// Rotates the slice in-place such that the first `self.len() - k`
3058 /// elements of the slice move to the end while the last `k` elements move
3059 /// to the front. After calling `rotate_right`, the element previously at
3060 /// index `self.len() - k` will become the first element in the slice.
3064 /// This function will panic if `k` is greater than the length of the
3065 /// slice. Note that `k == self.len()` does _not_ panic and is a no-op
3070 /// Takes linear (in `self.len()`) time.
3075 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3076 /// a.rotate_right(2);
3077 /// assert_eq!(a, ['e', 'f', 'a', 'b', 'c', 'd']);
3080 /// Rotate a subslice:
3083 /// let mut a = ['a', 'b', 'c', 'd', 'e', 'f'];
3084 /// a[1..5].rotate_right(1);
3085 /// assert_eq!(a, ['a', 'e', 'b', 'c', 'd', 'f']);
3087 #[stable(feature = "slice_rotate", since = "1.26.0")]
3088 pub fn rotate_right(&mut self, k: usize) {
3089 assert!(k <= self.len());
3090 let mid = self.len() - k;
3091 let p = self.as_mut_ptr();
3093 // SAFETY: The range `[p.add(mid) - mid, p.add(mid) + k)` is trivially
3094 // valid for reading and writing, as required by `ptr_rotate`.
3096 rotate::ptr_rotate(mid, p.add(mid), k);
3100 /// Fills `self` with elements by cloning `value`.
3105 /// let mut buf = vec![0; 10];
3107 /// assert_eq!(buf, vec![1; 10]);
3109 #[doc(alias = "memset")]
3110 #[stable(feature = "slice_fill", since = "1.50.0")]
3111 pub fn fill(&mut self, value: T)
3115 specialize::SpecFill::spec_fill(self, value);
3118 /// Fills `self` with elements returned by calling a closure repeatedly.
3120 /// This method uses a closure to create new values. If you'd rather
3121 /// [`Clone`] a given value, use [`fill`]. If you want to use the [`Default`]
3122 /// trait to generate values, you can pass [`Default::default`] as the
3125 /// [`fill`]: slice::fill
3130 /// let mut buf = vec![1; 10];
3131 /// buf.fill_with(Default::default);
3132 /// assert_eq!(buf, vec![0; 10]);
3134 #[stable(feature = "slice_fill_with", since = "1.51.0")]
3135 pub fn fill_with<F>(&mut self, mut f: F)
3144 /// Copies the elements from `src` into `self`.
3146 /// The length of `src` must be the same as `self`.
3150 /// This function will panic if the two slices have different lengths.
3154 /// Cloning two elements from a slice into another:
3157 /// let src = [1, 2, 3, 4];
3158 /// let mut dst = [0, 0];
3160 /// // Because the slices have to be the same length,
3161 /// // we slice the source slice from four elements
3162 /// // to two. It will panic if we don't do this.
3163 /// dst.clone_from_slice(&src[2..]);
3165 /// assert_eq!(src, [1, 2, 3, 4]);
3166 /// assert_eq!(dst, [3, 4]);
3169 /// Rust enforces that there can only be one mutable reference with no
3170 /// immutable references to a particular piece of data in a particular
3171 /// scope. Because of this, attempting to use `clone_from_slice` on a
3172 /// single slice will result in a compile failure:
3175 /// let mut slice = [1, 2, 3, 4, 5];
3177 /// slice[..2].clone_from_slice(&slice[3..]); // compile fail!
3180 /// To work around this, we can use [`split_at_mut`] to create two distinct
3181 /// sub-slices from a slice:
3184 /// let mut slice = [1, 2, 3, 4, 5];
3187 /// let (left, right) = slice.split_at_mut(2);
3188 /// left.clone_from_slice(&right[1..]);
3191 /// assert_eq!(slice, [4, 5, 3, 4, 5]);
3194 /// [`copy_from_slice`]: slice::copy_from_slice
3195 /// [`split_at_mut`]: slice::split_at_mut
3196 #[stable(feature = "clone_from_slice", since = "1.7.0")]
3198 pub fn clone_from_slice(&mut self, src: &[T])
3202 self.spec_clone_from(src);
3205 /// Copies all elements from `src` into `self`, using a memcpy.
3207 /// The length of `src` must be the same as `self`.
3209 /// If `T` does not implement `Copy`, use [`clone_from_slice`].
3213 /// This function will panic if the two slices have different lengths.
3217 /// Copying two elements from a slice into another:
3220 /// let src = [1, 2, 3, 4];
3221 /// let mut dst = [0, 0];
3223 /// // Because the slices have to be the same length,
3224 /// // we slice the source slice from four elements
3225 /// // to two. It will panic if we don't do this.
3226 /// dst.copy_from_slice(&src[2..]);
3228 /// assert_eq!(src, [1, 2, 3, 4]);
3229 /// assert_eq!(dst, [3, 4]);
3232 /// Rust enforces that there can only be one mutable reference with no
3233 /// immutable references to a particular piece of data in a particular
3234 /// scope. Because of this, attempting to use `copy_from_slice` on a
3235 /// single slice will result in a compile failure:
3238 /// let mut slice = [1, 2, 3, 4, 5];
3240 /// slice[..2].copy_from_slice(&slice[3..]); // compile fail!
3243 /// To work around this, we can use [`split_at_mut`] to create two distinct
3244 /// sub-slices from a slice:
3247 /// let mut slice = [1, 2, 3, 4, 5];
3250 /// let (left, right) = slice.split_at_mut(2);
3251 /// left.copy_from_slice(&right[1..]);
3254 /// assert_eq!(slice, [4, 5, 3, 4, 5]);
3257 /// [`clone_from_slice`]: slice::clone_from_slice
3258 /// [`split_at_mut`]: slice::split_at_mut
3259 #[doc(alias = "memcpy")]
3260 #[stable(feature = "copy_from_slice", since = "1.9.0")]
3262 pub fn copy_from_slice(&mut self, src: &[T])
3266 // The panic code path was put into a cold function to not bloat the
3271 fn len_mismatch_fail(dst_len: usize, src_len: usize) -> ! {
3273 "source slice length ({}) does not match destination slice length ({})",
3278 if self.len() != src.len() {
3279 len_mismatch_fail(self.len(), src.len());
3282 // SAFETY: `self` is valid for `self.len()` elements by definition, and `src` was
3283 // checked to have the same length. The slices cannot overlap because
3284 // mutable references are exclusive.
3286 ptr::copy_nonoverlapping(src.as_ptr(), self.as_mut_ptr(), self.len());
3290 /// Copies elements from one part of the slice to another part of itself,
3291 /// using a memmove.
3293 /// `src` is the range within `self` to copy from. `dest` is the starting
3294 /// index of the range within `self` to copy to, which will have the same
3295 /// length as `src`. The two ranges may overlap. The ends of the two ranges
3296 /// must be less than or equal to `self.len()`.
3300 /// This function will panic if either range exceeds the end of the slice,
3301 /// or if the end of `src` is before the start.
3305 /// Copying four bytes within a slice:
3308 /// let mut bytes = *b"Hello, World!";
3310 /// bytes.copy_within(1..5, 8);
3312 /// assert_eq!(&bytes, b"Hello, Wello!");
3314 #[stable(feature = "copy_within", since = "1.37.0")]
3316 pub fn copy_within<R: RangeBounds<usize>>(&mut self, src: R, dest: usize)
3320 let Range { start: src_start, end: src_end } = slice::range(src, ..self.len());
3321 let count = src_end - src_start;
3322 assert!(dest <= self.len() - count, "dest is out of bounds");
3323 // SAFETY: the conditions for `ptr::copy` have all been checked above,
3324 // as have those for `ptr::add`.
3326 // Derive both `src_ptr` and `dest_ptr` from the same loan
3327 let ptr = self.as_mut_ptr();
3328 let src_ptr = ptr.add(src_start);
3329 let dest_ptr = ptr.add(dest);
3330 ptr::copy(src_ptr, dest_ptr, count);
3334 /// Swaps all elements in `self` with those in `other`.
3336 /// The length of `other` must be the same as `self`.
3340 /// This function will panic if the two slices have different lengths.
3344 /// Swapping two elements across slices:
3347 /// let mut slice1 = [0, 0];
3348 /// let mut slice2 = [1, 2, 3, 4];
3350 /// slice1.swap_with_slice(&mut slice2[2..]);
3352 /// assert_eq!(slice1, [3, 4]);
3353 /// assert_eq!(slice2, [1, 2, 0, 0]);
3356 /// Rust enforces that there can only be one mutable reference to a
3357 /// particular piece of data in a particular scope. Because of this,
3358 /// attempting to use `swap_with_slice` on a single slice will result in
3359 /// a compile failure:
3362 /// let mut slice = [1, 2, 3, 4, 5];
3363 /// slice[..2].swap_with_slice(&mut slice[3..]); // compile fail!
3366 /// To work around this, we can use [`split_at_mut`] to create two distinct
3367 /// mutable sub-slices from a slice:
3370 /// let mut slice = [1, 2, 3, 4, 5];
3373 /// let (left, right) = slice.split_at_mut(2);
3374 /// left.swap_with_slice(&mut right[1..]);
3377 /// assert_eq!(slice, [4, 5, 3, 1, 2]);
3380 /// [`split_at_mut`]: slice::split_at_mut
3381 #[stable(feature = "swap_with_slice", since = "1.27.0")]
3383 pub fn swap_with_slice(&mut self, other: &mut [T]) {
3384 assert!(self.len() == other.len(), "destination and source slices have different lengths");
3385 // SAFETY: `self` is valid for `self.len()` elements by definition, and `src` was
3386 // checked to have the same length. The slices cannot overlap because
3387 // mutable references are exclusive.
3389 ptr::swap_nonoverlapping(self.as_mut_ptr(), other.as_mut_ptr(), self.len());
3393 /// Function to calculate lengths of the middle and trailing slice for `align_to{,_mut}`.
3394 fn align_to_offsets<U>(&self) -> (usize, usize) {
3395 // What we gonna do about `rest` is figure out what multiple of `U`s we can put in a
3396 // lowest number of `T`s. And how many `T`s we need for each such "multiple".
3398 // Consider for example T=u8 U=u16. Then we can put 1 U in 2 Ts. Simple. Now, consider
3399 // for example a case where size_of::<T> = 16, size_of::<U> = 24. We can put 2 Us in
3400 // place of every 3 Ts in the `rest` slice. A bit more complicated.
3402 // Formula to calculate this is:
3404 // Us = lcm(size_of::<T>, size_of::<U>) / size_of::<U>
3405 // Ts = lcm(size_of::<T>, size_of::<U>) / size_of::<T>
3407 // Expanded and simplified:
3409 // Us = size_of::<T> / gcd(size_of::<T>, size_of::<U>)
3410 // Ts = size_of::<U> / gcd(size_of::<T>, size_of::<U>)
3412 // Luckily since all this is constant-evaluated... performance here matters not!
3414 fn gcd(a: usize, b: usize) -> usize {
3415 use crate::intrinsics;
3416 // iterative stein’s algorithm
3417 // We should still make this `const fn` (and revert to recursive algorithm if we do)
3418 // because relying on llvm to consteval all this is… well, it makes me uncomfortable.
3420 // SAFETY: `a` and `b` are checked to be non-zero values.
3421 let (ctz_a, mut ctz_b) = unsafe {
3428 (intrinsics::cttz_nonzero(a), intrinsics::cttz_nonzero(b))
3430 let k = ctz_a.min(ctz_b);
3431 let mut a = a >> ctz_a;
3434 // remove all factors of 2 from b
3437 mem::swap(&mut a, &mut b);
3440 // SAFETY: `b` is checked to be non-zero.
3445 ctz_b = intrinsics::cttz_nonzero(b);
3450 let gcd: usize = gcd(mem::size_of::<T>(), mem::size_of::<U>());
3451 let ts: usize = mem::size_of::<U>() / gcd;
3452 let us: usize = mem::size_of::<T>() / gcd;
3454 // Armed with this knowledge, we can find how many `U`s we can fit!
3455 let us_len = self.len() / ts * us;
3456 // And how many `T`s will be in the trailing slice!
3457 let ts_len = self.len() % ts;
3461 /// Transmute the slice to a slice of another type, ensuring alignment of the types is
3464 /// This method splits the slice into three distinct slices: prefix, correctly aligned middle
3465 /// slice of a new type, and the suffix slice. The method may make the middle slice the greatest
3466 /// length possible for a given type and input slice, but only your algorithm's performance
3467 /// should depend on that, not its correctness. It is permissible for all of the input data to
3468 /// be returned as the prefix or suffix slice.
3470 /// This method has no purpose when either input element `T` or output element `U` are
3471 /// zero-sized and will return the original slice without splitting anything.
3475 /// This method is essentially a `transmute` with respect to the elements in the returned
3476 /// middle slice, so all the usual caveats pertaining to `transmute::<T, U>` also apply here.
3484 /// let bytes: [u8; 7] = [1, 2, 3, 4, 5, 6, 7];
3485 /// let (prefix, shorts, suffix) = bytes.align_to::<u16>();
3486 /// // less_efficient_algorithm_for_bytes(prefix);
3487 /// // more_efficient_algorithm_for_aligned_shorts(shorts);
3488 /// // less_efficient_algorithm_for_bytes(suffix);
3491 #[stable(feature = "slice_align_to", since = "1.30.0")]
3493 pub unsafe fn align_to<U>(&self) -> (&[T], &[U], &[T]) {
3494 // Note that most of this function will be constant-evaluated,
3495 if U::IS_ZST || T::IS_ZST {
3496 // handle ZSTs specially, which is – don't handle them at all.
3497 return (self, &[], &[]);
3500 // First, find at what point do we split between the first and 2nd slice. Easy with
3501 // ptr.align_offset.
3502 let ptr = self.as_ptr();
3503 // SAFETY: See the `align_to_mut` method for the detailed safety comment.
3504 let offset = unsafe { crate::ptr::align_offset(ptr, mem::align_of::<U>()) };
3505 if offset > self.len() {
3508 let (left, rest) = self.split_at(offset);
3509 let (us_len, ts_len) = rest.align_to_offsets::<U>();
3510 // SAFETY: now `rest` is definitely aligned, so `from_raw_parts` below is okay,
3511 // since the caller guarantees that we can transmute `T` to `U` safely.
3515 from_raw_parts(rest.as_ptr() as *const U, us_len),
3516 from_raw_parts(rest.as_ptr().add(rest.len() - ts_len), ts_len),
3522 /// Transmute the slice to a slice of another type, ensuring alignment of the types is
3525 /// This method splits the slice into three distinct slices: prefix, correctly aligned middle
3526 /// slice of a new type, and the suffix slice. The method may make the middle slice the greatest
3527 /// length possible for a given type and input slice, but only your algorithm's performance
3528 /// should depend on that, not its correctness. It is permissible for all of the input data to
3529 /// be returned as the prefix or suffix slice.
3531 /// This method has no purpose when either input element `T` or output element `U` are
3532 /// zero-sized and will return the original slice without splitting anything.
3536 /// This method is essentially a `transmute` with respect to the elements in the returned
3537 /// middle slice, so all the usual caveats pertaining to `transmute::<T, U>` also apply here.
3545 /// let mut bytes: [u8; 7] = [1, 2, 3, 4, 5, 6, 7];
3546 /// let (prefix, shorts, suffix) = bytes.align_to_mut::<u16>();
3547 /// // less_efficient_algorithm_for_bytes(prefix);
3548 /// // more_efficient_algorithm_for_aligned_shorts(shorts);
3549 /// // less_efficient_algorithm_for_bytes(suffix);
3552 #[stable(feature = "slice_align_to", since = "1.30.0")]
3554 pub unsafe fn align_to_mut<U>(&mut self) -> (&mut [T], &mut [U], &mut [T]) {
3555 // Note that most of this function will be constant-evaluated,
3556 if U::IS_ZST || T::IS_ZST {
3557 // handle ZSTs specially, which is – don't handle them at all.
3558 return (self, &mut [], &mut []);
3561 // First, find at what point do we split between the first and 2nd slice. Easy with
3562 // ptr.align_offset.
3563 let ptr = self.as_ptr();
3564 // SAFETY: Here we are ensuring we will use aligned pointers for U for the
3565 // rest of the method. This is done by passing a pointer to &[T] with an
3566 // alignment targeted for U.
3567 // `crate::ptr::align_offset` is called with a correctly aligned and
3568 // valid pointer `ptr` (it comes from a reference to `self`) and with
3569 // a size that is a power of two (since it comes from the alignment for U),
3570 // satisfying its safety constraints.
3571 let offset = unsafe { crate::ptr::align_offset(ptr, mem::align_of::<U>()) };
3572 if offset > self.len() {
3573 (self, &mut [], &mut [])
3575 let (left, rest) = self.split_at_mut(offset);
3576 let (us_len, ts_len) = rest.align_to_offsets::<U>();
3577 let rest_len = rest.len();
3578 let mut_ptr = rest.as_mut_ptr();
3579 // We can't use `rest` again after this, that would invalidate its alias `mut_ptr`!
3580 // SAFETY: see comments for `align_to`.
3584 from_raw_parts_mut(mut_ptr as *mut U, us_len),
3585 from_raw_parts_mut(mut_ptr.add(rest_len - ts_len), ts_len),
3591 /// Split a slice into a prefix, a middle of aligned SIMD types, and a suffix.
3593 /// This is a safe wrapper around [`slice::align_to`], so has the same weak
3594 /// postconditions as that method. You're only assured that
3595 /// `self.len() == prefix.len() + middle.len() * LANES + suffix.len()`.
3597 /// Notably, all of the following are possible:
3598 /// - `prefix.len() >= LANES`.
3599 /// - `middle.is_empty()` despite `self.len() >= 3 * LANES`.
3600 /// - `suffix.len() >= LANES`.
3602 /// That said, this is a safe method, so if you're only writing safe code,
3603 /// then this can at most cause incorrect logic, not unsoundness.
3607 /// This will panic if the size of the SIMD type is different from
3608 /// `LANES` times that of the scalar.
3610 /// At the time of writing, the trait restrictions on `Simd<T, LANES>` keeps
3611 /// that from ever happening, as only power-of-two numbers of lanes are
3612 /// supported. It's possible that, in the future, those restrictions might
3613 /// be lifted in a way that would make it possible to see panics from this
3614 /// method for something like `LANES == 3`.
3619 /// #![feature(portable_simd)]
3620 /// use core::simd::SimdFloat;
3622 /// let short = &[1, 2, 3];
3623 /// let (prefix, middle, suffix) = short.as_simd::<4>();
3624 /// assert_eq!(middle, []); // Not enough elements for anything in the middle
3626 /// // They might be split in any possible way between prefix and suffix
3627 /// let it = prefix.iter().chain(suffix).copied();
3628 /// assert_eq!(it.collect::<Vec<_>>(), vec![1, 2, 3]);
3630 /// fn basic_simd_sum(x: &[f32]) -> f32 {
3631 /// use std::ops::Add;
3632 /// use std::simd::f32x4;
3633 /// let (prefix, middle, suffix) = x.as_simd();
3634 /// let sums = f32x4::from_array([
3635 /// prefix.iter().copied().sum(),
3638 /// suffix.iter().copied().sum(),
3640 /// let sums = middle.iter().copied().fold(sums, f32x4::add);
3641 /// sums.reduce_sum()
3644 /// let numbers: Vec<f32> = (1..101).map(|x| x as _).collect();
3645 /// assert_eq!(basic_simd_sum(&numbers[1..99]), 4949.0);
3647 #[unstable(feature = "portable_simd", issue = "86656")]
3649 pub fn as_simd<const LANES: usize>(&self) -> (&[T], &[Simd<T, LANES>], &[T])
3651 Simd<T, LANES>: AsRef<[T; LANES]>,
3652 T: simd::SimdElement,
3653 simd::LaneCount<LANES>: simd::SupportedLaneCount,
3655 // These are expected to always match, as vector types are laid out like
3656 // arrays per <https://llvm.org/docs/LangRef.html#vector-type>, but we
3657 // might as well double-check since it'll optimize away anyhow.
3658 assert_eq!(mem::size_of::<Simd<T, LANES>>(), mem::size_of::<[T; LANES]>());
3660 // SAFETY: The simd types have the same layout as arrays, just with
3661 // potentially-higher alignment, so the de-facto transmutes are sound.
3662 unsafe { self.align_to() }
3665 /// Split a slice into a prefix, a middle of aligned SIMD types, and a suffix.
3667 /// This is a safe wrapper around [`slice::align_to_mut`], so has the same weak
3668 /// postconditions as that method. You're only assured that
3669 /// `self.len() == prefix.len() + middle.len() * LANES + suffix.len()`.
3671 /// Notably, all of the following are possible:
3672 /// - `prefix.len() >= LANES`.
3673 /// - `middle.is_empty()` despite `self.len() >= 3 * LANES`.
3674 /// - `suffix.len() >= LANES`.
3676 /// That said, this is a safe method, so if you're only writing safe code,
3677 /// then this can at most cause incorrect logic, not unsoundness.
3679 /// This is the mutable version of [`slice::as_simd`]; see that for examples.
3683 /// This will panic if the size of the SIMD type is different from
3684 /// `LANES` times that of the scalar.
3686 /// At the time of writing, the trait restrictions on `Simd<T, LANES>` keeps
3687 /// that from ever happening, as only power-of-two numbers of lanes are
3688 /// supported. It's possible that, in the future, those restrictions might
3689 /// be lifted in a way that would make it possible to see panics from this
3690 /// method for something like `LANES == 3`.
3691 #[unstable(feature = "portable_simd", issue = "86656")]
3693 pub fn as_simd_mut<const LANES: usize>(&mut self) -> (&mut [T], &mut [Simd<T, LANES>], &mut [T])
3695 Simd<T, LANES>: AsMut<[T; LANES]>,
3696 T: simd::SimdElement,
3697 simd::LaneCount<LANES>: simd::SupportedLaneCount,
3699 // These are expected to always match, as vector types are laid out like
3700 // arrays per <https://llvm.org/docs/LangRef.html#vector-type>, but we
3701 // might as well double-check since it'll optimize away anyhow.
3702 assert_eq!(mem::size_of::<Simd<T, LANES>>(), mem::size_of::<[T; LANES]>());
3704 // SAFETY: The simd types have the same layout as arrays, just with
3705 // potentially-higher alignment, so the de-facto transmutes are sound.
3706 unsafe { self.align_to_mut() }
3709 /// Checks if the elements of this slice are sorted.
3711 /// That is, for each element `a` and its following element `b`, `a <= b` must hold. If the
3712 /// slice yields exactly zero or one element, `true` is returned.
3714 /// Note that if `Self::Item` is only `PartialOrd`, but not `Ord`, the above definition
3715 /// implies that this function returns `false` if any two consecutive items are not
3721 /// #![feature(is_sorted)]
3722 /// let empty: [i32; 0] = [];
3724 /// assert!([1, 2, 2, 9].is_sorted());
3725 /// assert!(![1, 3, 2, 4].is_sorted());
3726 /// assert!([0].is_sorted());
3727 /// assert!(empty.is_sorted());
3728 /// assert!(![0.0, 1.0, f32::NAN].is_sorted());
3731 #[unstable(feature = "is_sorted", reason = "new API", issue = "53485")]
3733 pub fn is_sorted(&self) -> bool
3737 self.is_sorted_by(|a, b| a.partial_cmp(b))
3740 /// Checks if the elements of this slice are sorted using the given comparator function.
3742 /// Instead of using `PartialOrd::partial_cmp`, this function uses the given `compare`
3743 /// function to determine the ordering of two elements. Apart from that, it's equivalent to
3744 /// [`is_sorted`]; see its documentation for more information.
3746 /// [`is_sorted`]: slice::is_sorted
3747 #[unstable(feature = "is_sorted", reason = "new API", issue = "53485")]
3749 pub fn is_sorted_by<F>(&self, mut compare: F) -> bool
3751 F: FnMut(&T, &T) -> Option<Ordering>,
3753 self.iter().is_sorted_by(|a, b| compare(*a, *b))
3756 /// Checks if the elements of this slice are sorted using the given key extraction function.
3758 /// Instead of comparing the slice's elements directly, this function compares the keys of the
3759 /// elements, as determined by `f`. Apart from that, it's equivalent to [`is_sorted`]; see its
3760 /// documentation for more information.
3762 /// [`is_sorted`]: slice::is_sorted
3767 /// #![feature(is_sorted)]
3769 /// assert!(["c", "bb", "aaa"].is_sorted_by_key(|s| s.len()));
3770 /// assert!(![-2i32, -1, 0, 3].is_sorted_by_key(|n| n.abs()));
3773 #[unstable(feature = "is_sorted", reason = "new API", issue = "53485")]
3775 pub fn is_sorted_by_key<F, K>(&self, f: F) -> bool
3780 self.iter().is_sorted_by_key(f)
3783 /// Returns the index of the partition point according to the given predicate
3784 /// (the index of the first element of the second partition).
3786 /// The slice is assumed to be partitioned according to the given predicate.
3787 /// This means that all elements for which the predicate returns true are at the start of the slice
3788 /// and all elements for which the predicate returns false are at the end.
3789 /// For example, [7, 15, 3, 5, 4, 12, 6] is a partitioned under the predicate x % 2 != 0
3790 /// (all odd numbers are at the start, all even at the end).
3792 /// If this slice is not partitioned, the returned result is unspecified and meaningless,
3793 /// as this method performs a kind of binary search.
3795 /// See also [`binary_search`], [`binary_search_by`], and [`binary_search_by_key`].
3797 /// [`binary_search`]: slice::binary_search
3798 /// [`binary_search_by`]: slice::binary_search_by
3799 /// [`binary_search_by_key`]: slice::binary_search_by_key
3804 /// let v = [1, 2, 3, 3, 5, 6, 7];
3805 /// let i = v.partition_point(|&x| x < 5);
3807 /// assert_eq!(i, 4);
3808 /// assert!(v[..i].iter().all(|&x| x < 5));
3809 /// assert!(v[i..].iter().all(|&x| !(x < 5)));
3812 /// If all elements of the slice match the predicate, including if the slice
3813 /// is empty, then the length of the slice will be returned:
3816 /// let a = [2, 4, 8];
3817 /// assert_eq!(a.partition_point(|x| x < &100), a.len());
3818 /// let a: [i32; 0] = [];
3819 /// assert_eq!(a.partition_point(|x| x < &100), 0);
3822 /// If you want to insert an item to a sorted vector, while maintaining
3826 /// let mut s = vec![0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 55];
3828 /// let idx = s.partition_point(|&x| x < num);
3829 /// s.insert(idx, num);
3830 /// assert_eq!(s, [0, 1, 1, 1, 1, 2, 3, 5, 8, 13, 21, 34, 42, 55]);
3832 #[stable(feature = "partition_point", since = "1.52.0")]
3834 pub fn partition_point<P>(&self, mut pred: P) -> usize
3836 P: FnMut(&T) -> bool,
3838 self.binary_search_by(|x| if pred(x) { Less } else { Greater }).unwrap_or_else(|i| i)
3841 /// Removes the subslice corresponding to the given range
3842 /// and returns a reference to it.
3844 /// Returns `None` and does not modify the slice if the given
3845 /// range is out of bounds.
3847 /// Note that this method only accepts one-sided ranges such as
3848 /// `2..` or `..6`, but not `2..6`.
3852 /// Taking the first three elements of a slice:
3855 /// #![feature(slice_take)]
3857 /// let mut slice: &[_] = &['a', 'b', 'c', 'd'];
3858 /// let mut first_three = slice.take(..3).unwrap();
3860 /// assert_eq!(slice, &['d']);
3861 /// assert_eq!(first_three, &['a', 'b', 'c']);
3864 /// Taking the last two elements of a slice:
3867 /// #![feature(slice_take)]
3869 /// let mut slice: &[_] = &['a', 'b', 'c', 'd'];
3870 /// let mut tail = slice.take(2..).unwrap();
3872 /// assert_eq!(slice, &['a', 'b']);
3873 /// assert_eq!(tail, &['c', 'd']);
3876 /// Getting `None` when `range` is out of bounds:
3879 /// #![feature(slice_take)]
3881 /// let mut slice: &[_] = &['a', 'b', 'c', 'd'];
3883 /// assert_eq!(None, slice.take(5..));
3884 /// assert_eq!(None, slice.take(..5));
3885 /// assert_eq!(None, slice.take(..=4));
3886 /// let expected: &[char] = &['a', 'b', 'c', 'd'];
3887 /// assert_eq!(Some(expected), slice.take(..4));
3890 #[must_use = "method does not modify the slice if the range is out of bounds"]
3891 #[unstable(feature = "slice_take", issue = "62280")]
3892 pub fn take<'a, R: OneSidedRange<usize>>(self: &mut &'a Self, range: R) -> Option<&'a Self> {
3893 let (direction, split_index) = split_point_of(range)?;
3894 if split_index > self.len() {
3897 let (front, back) = self.split_at(split_index);
3899 Direction::Front => {
3903 Direction::Back => {
3910 /// Removes the subslice corresponding to the given range
3911 /// and returns a mutable reference to it.
3913 /// Returns `None` and does not modify the slice if the given
3914 /// range is out of bounds.
3916 /// Note that this method only accepts one-sided ranges such as
3917 /// `2..` or `..6`, but not `2..6`.
3921 /// Taking the first three elements of a slice:
3924 /// #![feature(slice_take)]
3926 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c', 'd'];
3927 /// let mut first_three = slice.take_mut(..3).unwrap();
3929 /// assert_eq!(slice, &mut ['d']);
3930 /// assert_eq!(first_three, &mut ['a', 'b', 'c']);
3933 /// Taking the last two elements of a slice:
3936 /// #![feature(slice_take)]
3938 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c', 'd'];
3939 /// let mut tail = slice.take_mut(2..).unwrap();
3941 /// assert_eq!(slice, &mut ['a', 'b']);
3942 /// assert_eq!(tail, &mut ['c', 'd']);
3945 /// Getting `None` when `range` is out of bounds:
3948 /// #![feature(slice_take)]
3950 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c', 'd'];
3952 /// assert_eq!(None, slice.take_mut(5..));
3953 /// assert_eq!(None, slice.take_mut(..5));
3954 /// assert_eq!(None, slice.take_mut(..=4));
3955 /// let expected: &mut [_] = &mut ['a', 'b', 'c', 'd'];
3956 /// assert_eq!(Some(expected), slice.take_mut(..4));
3959 #[must_use = "method does not modify the slice if the range is out of bounds"]
3960 #[unstable(feature = "slice_take", issue = "62280")]
3961 pub fn take_mut<'a, R: OneSidedRange<usize>>(
3962 self: &mut &'a mut Self,
3964 ) -> Option<&'a mut Self> {
3965 let (direction, split_index) = split_point_of(range)?;
3966 if split_index > self.len() {
3969 let (front, back) = mem::take(self).split_at_mut(split_index);
3971 Direction::Front => {
3975 Direction::Back => {
3982 /// Removes the first element of the slice and returns a reference
3985 /// Returns `None` if the slice is empty.
3990 /// #![feature(slice_take)]
3992 /// let mut slice: &[_] = &['a', 'b', 'c'];
3993 /// let first = slice.take_first().unwrap();
3995 /// assert_eq!(slice, &['b', 'c']);
3996 /// assert_eq!(first, &'a');
3999 #[unstable(feature = "slice_take", issue = "62280")]
4000 pub fn take_first<'a>(self: &mut &'a Self) -> Option<&'a T> {
4001 let (first, rem) = self.split_first()?;
4006 /// Removes the first element of the slice and returns a mutable
4007 /// reference to it.
4009 /// Returns `None` if the slice is empty.
4014 /// #![feature(slice_take)]
4016 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c'];
4017 /// let first = slice.take_first_mut().unwrap();
4020 /// assert_eq!(slice, &['b', 'c']);
4021 /// assert_eq!(first, &'d');
4024 #[unstable(feature = "slice_take", issue = "62280")]
4025 pub fn take_first_mut<'a>(self: &mut &'a mut Self) -> Option<&'a mut T> {
4026 let (first, rem) = mem::take(self).split_first_mut()?;
4031 /// Removes the last element of the slice and returns a reference
4034 /// Returns `None` if the slice is empty.
4039 /// #![feature(slice_take)]
4041 /// let mut slice: &[_] = &['a', 'b', 'c'];
4042 /// let last = slice.take_last().unwrap();
4044 /// assert_eq!(slice, &['a', 'b']);
4045 /// assert_eq!(last, &'c');
4048 #[unstable(feature = "slice_take", issue = "62280")]
4049 pub fn take_last<'a>(self: &mut &'a Self) -> Option<&'a T> {
4050 let (last, rem) = self.split_last()?;
4055 /// Removes the last element of the slice and returns a mutable
4056 /// reference to it.
4058 /// Returns `None` if the slice is empty.
4063 /// #![feature(slice_take)]
4065 /// let mut slice: &mut [_] = &mut ['a', 'b', 'c'];
4066 /// let last = slice.take_last_mut().unwrap();
4069 /// assert_eq!(slice, &['a', 'b']);
4070 /// assert_eq!(last, &'d');
4073 #[unstable(feature = "slice_take", issue = "62280")]
4074 pub fn take_last_mut<'a>(self: &mut &'a mut Self) -> Option<&'a mut T> {
4075 let (last, rem) = mem::take(self).split_last_mut()?;
4081 impl<T, const N: usize> [[T; N]] {
4082 /// Takes a `&[[T; N]]`, and flattens it to a `&[T]`.
4086 /// This panics if the length of the resulting slice would overflow a `usize`.
4088 /// This is only possible when flattening a slice of arrays of zero-sized
4089 /// types, and thus tends to be irrelevant in practice. If
4090 /// `size_of::<T>() > 0`, this will never panic.
4095 /// #![feature(slice_flatten)]
4097 /// assert_eq!([[1, 2, 3], [4, 5, 6]].flatten(), &[1, 2, 3, 4, 5, 6]);
4100 /// [[1, 2, 3], [4, 5, 6]].flatten(),
4101 /// [[1, 2], [3, 4], [5, 6]].flatten(),
4104 /// let slice_of_empty_arrays: &[[i32; 0]] = &[[], [], [], [], []];
4105 /// assert!(slice_of_empty_arrays.flatten().is_empty());
4107 /// let empty_slice_of_arrays: &[[u32; 10]] = &[];
4108 /// assert!(empty_slice_of_arrays.flatten().is_empty());
4110 #[unstable(feature = "slice_flatten", issue = "95629")]
4111 pub fn flatten(&self) -> &[T] {
4112 let len = if T::IS_ZST {
4113 self.len().checked_mul(N).expect("slice len overflow")
4115 // SAFETY: `self.len() * N` cannot overflow because `self` is
4116 // already in the address space.
4117 unsafe { self.len().unchecked_mul(N) }
4119 // SAFETY: `[T]` is layout-identical to `[T; N]`
4120 unsafe { from_raw_parts(self.as_ptr().cast(), len) }
4123 /// Takes a `&mut [[T; N]]`, and flattens it to a `&mut [T]`.
4127 /// This panics if the length of the resulting slice would overflow a `usize`.
4129 /// This is only possible when flattening a slice of arrays of zero-sized
4130 /// types, and thus tends to be irrelevant in practice. If
4131 /// `size_of::<T>() > 0`, this will never panic.
4136 /// #![feature(slice_flatten)]
4138 /// fn add_5_to_all(slice: &mut [i32]) {
4139 /// for i in slice {
4144 /// let mut array = [[1, 2, 3], [4, 5, 6], [7, 8, 9]];
4145 /// add_5_to_all(array.flatten_mut());
4146 /// assert_eq!(array, [[6, 7, 8], [9, 10, 11], [12, 13, 14]]);
4148 #[unstable(feature = "slice_flatten", issue = "95629")]
4149 pub fn flatten_mut(&mut self) -> &mut [T] {
4150 let len = if T::IS_ZST {
4151 self.len().checked_mul(N).expect("slice len overflow")
4153 // SAFETY: `self.len() * N` cannot overflow because `self` is
4154 // already in the address space.
4155 unsafe { self.len().unchecked_mul(N) }
4157 // SAFETY: `[T]` is layout-identical to `[T; N]`
4158 unsafe { from_raw_parts_mut(self.as_mut_ptr().cast(), len) }
4164 /// Sorts the slice of floats.
4166 /// This sort is in-place (i.e. does not allocate), *O*(*n* \* log(*n*)) worst-case, and uses
4167 /// the ordering defined by [`f32::total_cmp`].
4169 /// # Current implementation
4171 /// This uses the same sorting algorithm as [`sort_unstable_by`](slice::sort_unstable_by).
4176 /// #![feature(sort_floats)]
4177 /// let mut v = [2.6, -5e-8, f32::NAN, 8.29, f32::INFINITY, -1.0, 0.0, -f32::INFINITY, -0.0];
4179 /// v.sort_floats();
4180 /// let sorted = [-f32::INFINITY, -1.0, -5e-8, -0.0, 0.0, 2.6, 8.29, f32::INFINITY, f32::NAN];
4181 /// assert_eq!(&v[..8], &sorted[..8]);
4182 /// assert!(v[8].is_nan());
4184 #[unstable(feature = "sort_floats", issue = "93396")]
4186 pub fn sort_floats(&mut self) {
4187 self.sort_unstable_by(f32::total_cmp);
4193 /// Sorts the slice of floats.
4195 /// This sort is in-place (i.e. does not allocate), *O*(*n* \* log(*n*)) worst-case, and uses
4196 /// the ordering defined by [`f64::total_cmp`].
4198 /// # Current implementation
4200 /// This uses the same sorting algorithm as [`sort_unstable_by`](slice::sort_unstable_by).
4205 /// #![feature(sort_floats)]
4206 /// let mut v = [2.6, -5e-8, f64::NAN, 8.29, f64::INFINITY, -1.0, 0.0, -f64::INFINITY, -0.0];
4208 /// v.sort_floats();
4209 /// let sorted = [-f64::INFINITY, -1.0, -5e-8, -0.0, 0.0, 2.6, 8.29, f64::INFINITY, f64::NAN];
4210 /// assert_eq!(&v[..8], &sorted[..8]);
4211 /// assert!(v[8].is_nan());
4213 #[unstable(feature = "sort_floats", issue = "93396")]
4215 pub fn sort_floats(&mut self) {
4216 self.sort_unstable_by(f64::total_cmp);
4220 trait CloneFromSpec<T> {
4221 fn spec_clone_from(&mut self, src: &[T]);
4224 impl<T> CloneFromSpec<T> for [T]
4229 default fn spec_clone_from(&mut self, src: &[T]) {
4230 assert!(self.len() == src.len(), "destination and source slices have different lengths");
4231 // NOTE: We need to explicitly slice them to the same length
4232 // to make it easier for the optimizer to elide bounds checking.
4233 // But since it can't be relied on we also have an explicit specialization for T: Copy.
4234 let len = self.len();
4235 let src = &src[..len];
4237 self[i].clone_from(&src[i]);
4242 impl<T> CloneFromSpec<T> for [T]
4247 fn spec_clone_from(&mut self, src: &[T]) {
4248 self.copy_from_slice(src);
4252 #[stable(feature = "rust1", since = "1.0.0")]
4253 #[rustc_const_unstable(feature = "const_default_impls", issue = "87864")]
4254 impl<T> const Default for &[T] {
4255 /// Creates an empty slice.
4256 fn default() -> Self {
4261 #[stable(feature = "mut_slice_default", since = "1.5.0")]
4262 #[rustc_const_unstable(feature = "const_default_impls", issue = "87864")]
4263 impl<T> const Default for &mut [T] {
4264 /// Creates a mutable empty slice.
4265 fn default() -> Self {
4270 #[unstable(feature = "slice_pattern", reason = "stopgap trait for slice patterns", issue = "56345")]
4271 /// Patterns in slices - currently, only used by `strip_prefix` and `strip_suffix`. At a future
4272 /// point, we hope to generalise `core::str::Pattern` (which at the time of writing is limited to
4273 /// `str`) to slices, and then this trait will be replaced or abolished.
4274 pub trait SlicePattern {
4275 /// The element type of the slice being matched on.
4278 /// Currently, the consumers of `SlicePattern` need a slice.
4279 fn as_slice(&self) -> &[Self::Item];
4282 #[stable(feature = "slice_strip", since = "1.51.0")]
4283 impl<T> SlicePattern for [T] {
4287 fn as_slice(&self) -> &[Self::Item] {
4292 #[stable(feature = "slice_strip", since = "1.51.0")]
4293 impl<T, const N: usize> SlicePattern for [T; N] {
4297 fn as_slice(&self) -> &[Self::Item] {