1 //! Unicode string slices.
3 //! *[See also the `str` primitive type](../../std/primitive.str.html).*
5 //! The `&str` type is one of the two main string types, the other being `String`.
6 //! Unlike its `String` counterpart, its contents are borrowed.
10 //! A basic string declaration of `&str` type:
13 //! let hello_world = "Hello, World!";
16 //! Here we have declared a string literal, also known as a string slice.
17 //! String literals have a static lifetime, which means the string `hello_world`
18 //! is guaranteed to be valid for the duration of the entire program.
19 //! We can explicitly specify `hello_world`'s lifetime as well:
22 //! let hello_world: &'static str = "Hello, world!";
25 #![stable(feature = "rust1", since = "1.0.0")]
26 // Many of the usings in this module are only used in the test configuration.
27 // It's cleaner to just turn off the unused_imports warning than to fix them.
28 #![allow(unused_imports)]
30 use core::borrow::{Borrow, BorrowMut};
31 use core::iter::FusedIterator;
34 use core::str::pattern::{DoubleEndedSearcher, Pattern, ReverseSearcher, Searcher};
35 use core::unicode::conversions;
37 use crate::borrow::ToOwned;
38 use crate::boxed::Box;
39 use crate::slice::{Concat, Join, SliceIndex};
40 use crate::string::String;
43 #[stable(feature = "rust1", since = "1.0.0")]
44 pub use core::str::pattern;
45 #[stable(feature = "encode_utf16", since = "1.8.0")]
46 pub use core::str::EncodeUtf16;
47 #[stable(feature = "split_ascii_whitespace", since = "1.34.0")]
48 pub use core::str::SplitAsciiWhitespace;
49 #[stable(feature = "rust1", since = "1.0.0")]
50 pub use core::str::SplitWhitespace;
51 #[stable(feature = "rust1", since = "1.0.0")]
52 pub use core::str::{from_utf8, from_utf8_mut, Bytes, CharIndices, Chars};
53 #[stable(feature = "rust1", since = "1.0.0")]
54 pub use core::str::{from_utf8_unchecked, from_utf8_unchecked_mut, ParseBoolError};
55 #[stable(feature = "str_escape", since = "1.34.0")]
56 pub use core::str::{EscapeDebug, EscapeDefault, EscapeUnicode};
57 #[stable(feature = "rust1", since = "1.0.0")]
58 pub use core::str::{FromStr, Utf8Error};
60 #[stable(feature = "rust1", since = "1.0.0")]
61 pub use core::str::{Lines, LinesAny};
62 #[stable(feature = "rust1", since = "1.0.0")]
63 pub use core::str::{MatchIndices, RMatchIndices};
64 #[stable(feature = "rust1", since = "1.0.0")]
65 pub use core::str::{Matches, RMatches};
66 #[stable(feature = "rust1", since = "1.0.0")]
67 pub use core::str::{RSplit, Split};
68 #[stable(feature = "rust1", since = "1.0.0")]
69 pub use core::str::{RSplitN, SplitN};
70 #[stable(feature = "rust1", since = "1.0.0")]
71 pub use core::str::{RSplitTerminator, SplitTerminator};
73 /// Note: `str` in `Concat<str>` is not meaningful here.
74 /// This type parameter of the trait only exists to enable another impl.
75 #[unstable(feature = "slice_concat_ext", issue = "27747")]
76 impl<S: Borrow<str>> Concat<str> for [S] {
79 fn concat(slice: &Self) -> String {
84 #[unstable(feature = "slice_concat_ext", issue = "27747")]
85 impl<S: Borrow<str>> Join<&str> for [S] {
88 fn join(slice: &Self, sep: &str) -> String {
89 unsafe { String::from_utf8_unchecked(join_generic_copy(slice, sep.as_bytes())) }
93 macro_rules! spezialize_for_lengths {
94 ($separator:expr, $target:expr, $iter:expr; $($num:expr),*) => {
95 let mut target = $target;
97 let sep_bytes = $separator;
98 match $separator.len() {
100 // loops with hardcoded sizes run much faster
101 // specialize the cases with small separator lengths
104 copy_slice_and_advance!(target, sep_bytes);
105 copy_slice_and_advance!(target, s.borrow().as_ref());
110 // arbitrary non-zero size fallback
112 copy_slice_and_advance!(target, sep_bytes);
113 copy_slice_and_advance!(target, s.borrow().as_ref());
120 macro_rules! copy_slice_and_advance {
121 ($target:expr, $bytes:expr) => {
122 let len = $bytes.len();
123 let (head, tail) = { $target }.split_at_mut(len);
124 head.copy_from_slice($bytes);
129 // Optimized join implementation that works for both Vec<T> (T: Copy) and String's inner vec
130 // Currently (2018-05-13) there is a bug with type inference and specialization (see issue #36262)
131 // For this reason SliceConcat<T> is not specialized for T: Copy and SliceConcat<str> is the
132 // only user of this function. It is left in place for the time when that is fixed.
134 // the bounds for String-join are S: Borrow<str> and for Vec-join Borrow<[T]>
135 // [T] and str both impl AsRef<[T]> for some T
136 // => s.borrow().as_ref() and we always have slices
137 fn join_generic_copy<B, T, S>(slice: &[S], sep: &[T]) -> Vec<T>
140 B: AsRef<[T]> + ?Sized,
143 let sep_len = sep.len();
144 let mut iter = slice.iter();
146 // the first slice is the only one without a separator preceding it
147 let first = match iter.next() {
148 Some(first) => first,
149 None => return vec![],
152 // compute the exact total length of the joined Vec
153 // if the `len` calculation overflows, we'll panic
154 // we would have run out of memory anyway and the rest of the function requires
155 // the entire Vec pre-allocated for safety
157 .checked_mul(iter.len())
159 slice.iter().map(|s| s.borrow().as_ref().len()).try_fold(n, usize::checked_add)
161 .expect("attempt to join into collection with len > usize::MAX");
163 // crucial for safety
164 let mut result = Vec::with_capacity(len);
165 assert!(result.capacity() >= len);
167 result.extend_from_slice(first.borrow().as_ref());
171 let pos = result.len();
172 let target = result.get_unchecked_mut(pos..len);
174 // copy separator and slices over without bounds checks
175 // generate loops with hardcoded offsets for small separators
176 // massive improvements possible (~ x2)
177 spezialize_for_lengths!(sep, target, iter; 0, 1, 2, 3, 4);
184 #[stable(feature = "rust1", since = "1.0.0")]
185 impl Borrow<str> for String {
187 fn borrow(&self) -> &str {
192 #[stable(feature = "string_borrow_mut", since = "1.36.0")]
193 impl BorrowMut<str> for String {
195 fn borrow_mut(&mut self) -> &mut str {
200 #[stable(feature = "rust1", since = "1.0.0")]
201 impl ToOwned for str {
204 fn to_owned(&self) -> String {
205 unsafe { String::from_utf8_unchecked(self.as_bytes().to_owned()) }
208 fn clone_into(&self, target: &mut String) {
209 let mut b = mem::take(target).into_bytes();
210 self.as_bytes().clone_into(&mut b);
211 *target = unsafe { String::from_utf8_unchecked(b) }
215 /// Methods for string slices.
216 #[lang = "str_alloc"]
219 /// Converts a `Box<str>` into a `Box<[u8]>` without copying or allocating.
226 /// let s = "this is a string";
227 /// let boxed_str = s.to_owned().into_boxed_str();
228 /// let boxed_bytes = boxed_str.into_boxed_bytes();
229 /// assert_eq!(*boxed_bytes, *s.as_bytes());
231 #[stable(feature = "str_box_extras", since = "1.20.0")]
233 pub fn into_boxed_bytes(self: Box<str>) -> Box<[u8]> {
237 /// Replaces all matches of a pattern with another string.
239 /// `replace` creates a new [`String`], and copies the data from this string slice into it.
240 /// While doing so, it attempts to find matches of a pattern. If it finds any, it
241 /// replaces them with the replacement string slice.
243 /// [`String`]: string/struct.String.html
250 /// let s = "this is old";
252 /// assert_eq!("this is new", s.replace("old", "new"));
255 /// When the pattern doesn't match:
258 /// let s = "this is old";
259 /// assert_eq!(s, s.replace("cookie monster", "little lamb"));
261 #[must_use = "this returns the replaced string as a new allocation, \
262 without modifying the original"]
263 #[stable(feature = "rust1", since = "1.0.0")]
265 pub fn replace<'a, P: Pattern<'a>>(&'a self, from: P, to: &str) -> String {
266 let mut result = String::new();
267 let mut last_end = 0;
268 for (start, part) in self.match_indices(from) {
269 result.push_str(unsafe { self.get_unchecked(last_end..start) });
271 last_end = start + part.len();
273 result.push_str(unsafe { self.get_unchecked(last_end..self.len()) });
277 /// Replaces first N matches of a pattern with another string.
279 /// `replacen` creates a new [`String`], and copies the data from this string slice into it.
280 /// While doing so, it attempts to find matches of a pattern. If it finds any, it
281 /// replaces them with the replacement string slice at most `count` times.
283 /// [`String`]: string/struct.String.html
290 /// let s = "foo foo 123 foo";
291 /// assert_eq!("new new 123 foo", s.replacen("foo", "new", 2));
292 /// assert_eq!("faa fao 123 foo", s.replacen('o', "a", 3));
293 /// assert_eq!("foo foo new23 foo", s.replacen(char::is_numeric, "new", 1));
296 /// When the pattern doesn't match:
299 /// let s = "this is old";
300 /// assert_eq!(s, s.replacen("cookie monster", "little lamb", 10));
302 #[must_use = "this returns the replaced string as a new allocation, \
303 without modifying the original"]
304 #[stable(feature = "str_replacen", since = "1.16.0")]
305 pub fn replacen<'a, P: Pattern<'a>>(&'a self, pat: P, to: &str, count: usize) -> String {
306 // Hope to reduce the times of re-allocation
307 let mut result = String::with_capacity(32);
308 let mut last_end = 0;
309 for (start, part) in self.match_indices(pat).take(count) {
310 result.push_str(unsafe { self.get_unchecked(last_end..start) });
312 last_end = start + part.len();
314 result.push_str(unsafe { self.get_unchecked(last_end..self.len()) });
318 /// Returns the lowercase equivalent of this string slice, as a new [`String`].
320 /// 'Lowercase' is defined according to the terms of the Unicode Derived Core Property
323 /// Since some characters can expand into multiple characters when changing
324 /// the case, this function returns a [`String`] instead of modifying the
325 /// parameter in-place.
327 /// [`String`]: string/struct.String.html
336 /// assert_eq!("hello", s.to_lowercase());
339 /// A tricky example, with sigma:
344 /// assert_eq!("σ", sigma.to_lowercase());
346 /// // but at the end of a word, it's ς, not σ:
347 /// let odysseus = "ὈΔΥΣΣΕΎΣ";
349 /// assert_eq!("ὀδυσσεύς", odysseus.to_lowercase());
352 /// Languages without case are not changed:
355 /// let new_year = "农历新年";
357 /// assert_eq!(new_year, new_year.to_lowercase());
359 #[stable(feature = "unicode_case_mapping", since = "1.2.0")]
360 pub fn to_lowercase(&self) -> String {
361 let mut s = String::with_capacity(self.len());
362 for (i, c) in self[..].char_indices() {
364 // Σ maps to σ, except at the end of a word where it maps to ς.
365 // This is the only conditional (contextual) but language-independent mapping
366 // in `SpecialCasing.txt`,
367 // so hard-code it rather than have a generic "condition" mechanism.
368 // See https://github.com/rust-lang/rust/issues/26035
369 map_uppercase_sigma(self, i, &mut s)
371 match conversions::to_lower(c) {
372 [a, '\0', _] => s.push(a),
387 fn map_uppercase_sigma(from: &str, i: usize, to: &mut String) {
388 // See http://www.unicode.org/versions/Unicode7.0.0/ch03.pdf#G33992
389 // for the definition of `Final_Sigma`.
390 debug_assert!('Σ'.len_utf8() == 2);
391 let is_word_final = case_ignoreable_then_cased(from[..i].chars().rev())
392 && !case_ignoreable_then_cased(from[i + 2..].chars());
393 to.push_str(if is_word_final { "ς" } else { "σ" });
396 fn case_ignoreable_then_cased<I: Iterator<Item = char>>(iter: I) -> bool {
397 use core::unicode::derived_property::{Case_Ignorable, Cased};
398 match iter.skip_while(|&c| Case_Ignorable(c)).next() {
405 /// Returns the uppercase equivalent of this string slice, as a new [`String`].
407 /// 'Uppercase' is defined according to the terms of the Unicode Derived Core Property
410 /// Since some characters can expand into multiple characters when changing
411 /// the case, this function returns a [`String`] instead of modifying the
412 /// parameter in-place.
414 /// [`String`]: string/struct.String.html
423 /// assert_eq!("HELLO", s.to_uppercase());
426 /// Scripts without case are not changed:
429 /// let new_year = "农历新年";
431 /// assert_eq!(new_year, new_year.to_uppercase());
434 /// One character can become multiple:
436 /// let s = "tschüß";
438 /// assert_eq!("TSCHÜSS", s.to_uppercase());
440 #[stable(feature = "unicode_case_mapping", since = "1.2.0")]
441 pub fn to_uppercase(&self) -> String {
442 let mut s = String::with_capacity(self.len());
443 for c in self[..].chars() {
444 match conversions::to_upper(c) {
445 [a, '\0', _] => s.push(a),
460 /// Converts a [`Box<str>`] into a [`String`] without copying or allocating.
462 /// [`String`]: string/struct.String.html
463 /// [`Box<str>`]: boxed/struct.Box.html
470 /// let string = String::from("birthday gift");
471 /// let boxed_str = string.clone().into_boxed_str();
473 /// assert_eq!(boxed_str.into_string(), string);
475 #[stable(feature = "box_str", since = "1.4.0")]
477 pub fn into_string(self: Box<str>) -> String {
478 let slice = Box::<[u8]>::from(self);
479 unsafe { String::from_utf8_unchecked(slice.into_vec()) }
482 /// Creates a new [`String`] by repeating a string `n` times.
486 /// This function will panic if the capacity would overflow.
488 /// [`String`]: string/struct.String.html
495 /// assert_eq!("abc".repeat(4), String::from("abcabcabcabc"));
498 /// A panic upon overflow:
501 /// // this will panic at runtime
502 /// "0123456789abcdef".repeat(usize::MAX);
504 #[stable(feature = "repeat_str", since = "1.16.0")]
505 pub fn repeat(&self, n: usize) -> String {
506 unsafe { String::from_utf8_unchecked(self.as_bytes().repeat(n)) }
509 /// Returns a copy of this string where each character is mapped to its
510 /// ASCII upper case equivalent.
512 /// ASCII letters 'a' to 'z' are mapped to 'A' to 'Z',
513 /// but non-ASCII letters are unchanged.
515 /// To uppercase the value in-place, use [`make_ascii_uppercase`].
517 /// To uppercase ASCII characters in addition to non-ASCII characters, use
518 /// [`to_uppercase`].
523 /// let s = "Grüße, Jürgen ❤";
525 /// assert_eq!("GRüßE, JüRGEN ❤", s.to_ascii_uppercase());
528 /// [`make_ascii_uppercase`]: #method.make_ascii_uppercase
529 /// [`to_uppercase`]: #method.to_uppercase
530 #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")]
532 pub fn to_ascii_uppercase(&self) -> String {
533 let mut bytes = self.as_bytes().to_vec();
534 bytes.make_ascii_uppercase();
535 // make_ascii_uppercase() preserves the UTF-8 invariant.
536 unsafe { String::from_utf8_unchecked(bytes) }
539 /// Returns a copy of this string where each character is mapped to its
540 /// ASCII lower case equivalent.
542 /// ASCII letters 'A' to 'Z' are mapped to 'a' to 'z',
543 /// but non-ASCII letters are unchanged.
545 /// To lowercase the value in-place, use [`make_ascii_lowercase`].
547 /// To lowercase ASCII characters in addition to non-ASCII characters, use
548 /// [`to_lowercase`].
553 /// let s = "Grüße, Jürgen ❤";
555 /// assert_eq!("grüße, jürgen ❤", s.to_ascii_lowercase());
558 /// [`make_ascii_lowercase`]: #method.make_ascii_lowercase
559 /// [`to_lowercase`]: #method.to_lowercase
560 #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")]
562 pub fn to_ascii_lowercase(&self) -> String {
563 let mut bytes = self.as_bytes().to_vec();
564 bytes.make_ascii_lowercase();
565 // make_ascii_lowercase() preserves the UTF-8 invariant.
566 unsafe { String::from_utf8_unchecked(bytes) }
570 /// Converts a boxed slice of bytes to a boxed string slice without checking
571 /// that the string contains valid UTF-8.
578 /// let smile_utf8 = Box::new([226, 152, 186]);
579 /// let smile = unsafe { std::str::from_boxed_utf8_unchecked(smile_utf8) };
581 /// assert_eq!("☺", &*smile);
583 #[stable(feature = "str_box_extras", since = "1.20.0")]
585 pub unsafe fn from_boxed_utf8_unchecked(v: Box<[u8]>) -> Box<str> {
586 unsafe { Box::from_raw(Box::into_raw(v) as *mut str) }