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")]
27 // Many of the usings in this module are only used in the test configuration.
28 // It's cleaner to just turn off the unused_imports warning than to fix them.
29 #![allow(unused_imports)]
31 use core::borrow::Borrow;
32 use core::str::pattern::{Pattern, Searcher, ReverseSearcher, DoubleEndedSearcher};
35 use core::iter::FusedIterator;
36 use core::unicode::conversions;
38 use crate::borrow::ToOwned;
39 use crate::boxed::Box;
40 use crate::slice::{SliceConcatExt, SliceIndex};
41 use crate::string::String;
44 #[stable(feature = "rust1", since = "1.0.0")]
45 pub use core::str::{FromStr, Utf8Error};
47 #[stable(feature = "rust1", since = "1.0.0")]
48 pub use core::str::{Lines, LinesAny};
49 #[stable(feature = "rust1", since = "1.0.0")]
50 pub use core::str::{Split, RSplit};
51 #[stable(feature = "rust1", since = "1.0.0")]
52 pub use core::str::{SplitN, RSplitN};
53 #[stable(feature = "rust1", since = "1.0.0")]
54 pub use core::str::{SplitTerminator, RSplitTerminator};
55 #[stable(feature = "rust1", since = "1.0.0")]
56 pub use core::str::{Matches, RMatches};
57 #[stable(feature = "rust1", since = "1.0.0")]
58 pub use core::str::{MatchIndices, RMatchIndices};
59 #[stable(feature = "rust1", since = "1.0.0")]
60 pub use core::str::{from_utf8, from_utf8_mut, Chars, CharIndices, Bytes};
61 #[stable(feature = "rust1", since = "1.0.0")]
62 pub use core::str::{from_utf8_unchecked, from_utf8_unchecked_mut, ParseBoolError};
63 #[stable(feature = "rust1", since = "1.0.0")]
64 pub use core::str::SplitWhitespace;
65 #[stable(feature = "rust1", since = "1.0.0")]
66 pub use core::str::pattern;
67 #[stable(feature = "encode_utf16", since = "1.8.0")]
68 pub use core::str::EncodeUtf16;
69 #[stable(feature = "split_ascii_whitespace", since = "1.34.0")]
70 pub use core::str::SplitAsciiWhitespace;
72 #[unstable(feature = "slice_concat_ext",
73 reason = "trait should not have to exist",
75 impl<S: Borrow<str>> SliceConcatExt<str> for [S] {
78 fn concat(&self) -> String {
82 fn join(&self, sep: &str) -> String {
84 String::from_utf8_unchecked( join_generic_copy(self, sep.as_bytes()) )
88 fn connect(&self, sep: &str) -> String {
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 SliceConcatExt<T> is not specialized for T: Copy and SliceConcatExt<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
156 let len = sep_len.checked_mul(iter.len()).and_then(|n| {
158 .map(|s| s.borrow().as_ref().len())
159 .try_fold(n, usize::checked_add)
160 }).expect("attempt to join into collection with len > usize::MAX");
162 // crucial for safety
163 let mut result = Vec::with_capacity(len);
164 assert!(result.capacity() >= len);
166 result.extend_from_slice(first.borrow().as_ref());
170 let pos = result.len();
171 let target = result.get_unchecked_mut(pos..len);
173 // copy separator and slices over without bounds checks
174 // generate loops with hardcoded offsets for small separators
175 // massive improvements possible (~ x2)
176 spezialize_for_lengths!(sep, target, iter; 0, 1, 2, 3, 4);
183 #[stable(feature = "rust1", since = "1.0.0")]
184 impl Borrow<str> for String {
186 fn borrow(&self) -> &str {
191 #[stable(feature = "rust1", since = "1.0.0")]
192 impl ToOwned for str {
195 fn to_owned(&self) -> String {
196 unsafe { String::from_utf8_unchecked(self.as_bytes().to_owned()) }
199 fn clone_into(&self, target: &mut String) {
200 let mut b = mem::replace(target, String::new()).into_bytes();
201 self.as_bytes().clone_into(&mut b);
202 *target = unsafe { String::from_utf8_unchecked(b) }
206 /// Methods for string slices.
207 #[lang = "str_alloc"]
210 /// Converts a `Box<str>` into a `Box<[u8]>` without copying or allocating.
217 /// let s = "this is a string";
218 /// let boxed_str = s.to_owned().into_boxed_str();
219 /// let boxed_bytes = boxed_str.into_boxed_bytes();
220 /// assert_eq!(*boxed_bytes, *s.as_bytes());
222 #[stable(feature = "str_box_extras", since = "1.20.0")]
224 pub fn into_boxed_bytes(self: Box<str>) -> Box<[u8]> {
228 /// Replaces all matches of a pattern with another string.
230 /// `replace` creates a new [`String`], and copies the data from this string slice into it.
231 /// While doing so, it attempts to find matches of a pattern. If it finds any, it
232 /// replaces them with the replacement string slice.
234 /// [`String`]: string/struct.String.html
241 /// let s = "this is old";
243 /// assert_eq!("this is new", s.replace("old", "new"));
246 /// When the pattern doesn't match:
249 /// let s = "this is old";
250 /// assert_eq!(s, s.replace("cookie monster", "little lamb"));
252 #[must_use = "this returns the replaced string as a new allocation, \
253 without modifying the original"]
254 #[stable(feature = "rust1", since = "1.0.0")]
256 pub fn replace<'a, P: Pattern<'a>>(&'a self, from: P, to: &str) -> String {
257 let mut result = String::new();
258 let mut last_end = 0;
259 for (start, part) in self.match_indices(from) {
260 result.push_str(unsafe { self.get_unchecked(last_end..start) });
262 last_end = start + part.len();
264 result.push_str(unsafe { self.get_unchecked(last_end..self.len()) });
268 /// Replaces first N matches of a pattern with another string.
270 /// `replacen` creates a new [`String`], and copies the data from this string slice into it.
271 /// While doing so, it attempts to find matches of a pattern. If it finds any, it
272 /// replaces them with the replacement string slice at most `count` times.
274 /// [`String`]: string/struct.String.html
281 /// let s = "foo foo 123 foo";
282 /// assert_eq!("new new 123 foo", s.replacen("foo", "new", 2));
283 /// assert_eq!("faa fao 123 foo", s.replacen('o', "a", 3));
284 /// assert_eq!("foo foo new23 foo", s.replacen(char::is_numeric, "new", 1));
287 /// When the pattern doesn't match:
290 /// let s = "this is old";
291 /// assert_eq!(s, s.replacen("cookie monster", "little lamb", 10));
293 #[must_use = "this returns the replaced string as a new allocation, \
294 without modifying the original"]
295 #[stable(feature = "str_replacen", since = "1.16.0")]
296 pub fn replacen<'a, P: Pattern<'a>>(&'a self, pat: P, to: &str, count: usize) -> String {
297 // Hope to reduce the times of re-allocation
298 let mut result = String::with_capacity(32);
299 let mut last_end = 0;
300 for (start, part) in self.match_indices(pat).take(count) {
301 result.push_str(unsafe { self.get_unchecked(last_end..start) });
303 last_end = start + part.len();
305 result.push_str(unsafe { self.get_unchecked(last_end..self.len()) });
309 /// Returns the lowercase equivalent of this string slice, as a new [`String`].
311 /// 'Lowercase' is defined according to the terms of the Unicode Derived Core Property
314 /// Since some characters can expand into multiple characters when changing
315 /// the case, this function returns a [`String`] instead of modifying the
316 /// parameter in-place.
318 /// [`String`]: string/struct.String.html
327 /// assert_eq!("hello", s.to_lowercase());
330 /// A tricky example, with sigma:
335 /// assert_eq!("σ", sigma.to_lowercase());
337 /// // but at the end of a word, it's ς, not σ:
338 /// let odysseus = "ὈΔΥΣΣΕΎΣ";
340 /// assert_eq!("ὀδυσσεύς", odysseus.to_lowercase());
343 /// Languages without case are not changed:
346 /// let new_year = "农历新年";
348 /// assert_eq!(new_year, new_year.to_lowercase());
350 #[stable(feature = "unicode_case_mapping", since = "1.2.0")]
351 pub fn to_lowercase(&self) -> String {
352 let mut s = String::with_capacity(self.len());
353 for (i, c) in self[..].char_indices() {
355 // Σ maps to σ, except at the end of a word where it maps to ς.
356 // This is the only conditional (contextual) but language-independent mapping
357 // in `SpecialCasing.txt`,
358 // so hard-code it rather than have a generic "condition" mechanism.
359 // See https://github.com/rust-lang/rust/issues/26035
360 map_uppercase_sigma(self, i, &mut s)
362 match conversions::to_lower(c) {
363 [a, '\0', _] => s.push(a),
378 fn map_uppercase_sigma(from: &str, i: usize, to: &mut String) {
379 // See http://www.unicode.org/versions/Unicode7.0.0/ch03.pdf#G33992
380 // for the definition of `Final_Sigma`.
381 debug_assert!('Σ'.len_utf8() == 2);
382 let is_word_final = case_ignoreable_then_cased(from[..i].chars().rev()) &&
383 !case_ignoreable_then_cased(from[i + 2..].chars());
384 to.push_str(if is_word_final { "ς" } else { "σ" });
387 fn case_ignoreable_then_cased<I: Iterator<Item = char>>(iter: I) -> bool {
388 use core::unicode::derived_property::{Cased, Case_Ignorable};
389 match iter.skip_while(|&c| Case_Ignorable(c)).next() {
396 /// Returns the uppercase equivalent of this string slice, as a new [`String`].
398 /// 'Uppercase' is defined according to the terms of the Unicode Derived Core Property
401 /// Since some characters can expand into multiple characters when changing
402 /// the case, this function returns a [`String`] instead of modifying the
403 /// parameter in-place.
405 /// [`String`]: string/struct.String.html
414 /// assert_eq!("HELLO", s.to_uppercase());
417 /// Scripts without case are not changed:
420 /// let new_year = "农历新年";
422 /// assert_eq!(new_year, new_year.to_uppercase());
424 #[stable(feature = "unicode_case_mapping", since = "1.2.0")]
425 pub fn to_uppercase(&self) -> String {
426 let mut s = String::with_capacity(self.len());
427 for c in self[..].chars() {
428 match conversions::to_upper(c) {
429 [a, '\0', _] => s.push(a),
444 /// Converts a [`Box<str>`] into a [`String`] without copying or allocating.
446 /// [`String`]: string/struct.String.html
447 /// [`Box<str>`]: boxed/struct.Box.html
454 /// let string = String::from("birthday gift");
455 /// let boxed_str = string.clone().into_boxed_str();
457 /// assert_eq!(boxed_str.into_string(), string);
459 #[stable(feature = "box_str", since = "1.4.0")]
461 pub fn into_string(self: Box<str>) -> String {
462 let slice = Box::<[u8]>::from(self);
463 unsafe { String::from_utf8_unchecked(slice.into_vec()) }
466 /// Creates a new [`String`] by repeating a string `n` times.
470 /// This function will panic if the capacity would overflow.
472 /// [`String`]: string/struct.String.html
479 /// assert_eq!("abc".repeat(4), String::from("abcabcabcabc"));
482 /// A panic upon overflow:
486 /// // this will panic at runtime
487 /// "0123456789abcdef".repeat(usize::max_value());
490 #[stable(feature = "repeat_str", since = "1.16.0")]
491 pub fn repeat(&self, n: usize) -> String {
492 unsafe { String::from_utf8_unchecked(self.as_bytes().repeat(n)) }
495 /// Returns a copy of this string where each character is mapped to its
496 /// ASCII upper case equivalent.
498 /// ASCII letters 'a' to 'z' are mapped to 'A' to 'Z',
499 /// but non-ASCII letters are unchanged.
501 /// To uppercase the value in-place, use [`make_ascii_uppercase`].
503 /// To uppercase ASCII characters in addition to non-ASCII characters, use
504 /// [`to_uppercase`].
509 /// let s = "Grüße, Jürgen ❤";
511 /// assert_eq!("GRüßE, JüRGEN ❤", s.to_ascii_uppercase());
514 /// [`make_ascii_uppercase`]: #method.make_ascii_uppercase
515 /// [`to_uppercase`]: #method.to_uppercase
516 #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")]
518 pub fn to_ascii_uppercase(&self) -> String {
519 let mut bytes = self.as_bytes().to_vec();
520 bytes.make_ascii_uppercase();
521 // make_ascii_uppercase() preserves the UTF-8 invariant.
522 unsafe { String::from_utf8_unchecked(bytes) }
525 /// Returns a copy of this string where each character is mapped to its
526 /// ASCII lower case equivalent.
528 /// ASCII letters 'A' to 'Z' are mapped to 'a' to 'z',
529 /// but non-ASCII letters are unchanged.
531 /// To lowercase the value in-place, use [`make_ascii_lowercase`].
533 /// To lowercase ASCII characters in addition to non-ASCII characters, use
534 /// [`to_lowercase`].
539 /// let s = "Grüße, Jürgen ❤";
541 /// assert_eq!("grüße, jürgen ❤", s.to_ascii_lowercase());
544 /// [`make_ascii_lowercase`]: #method.make_ascii_lowercase
545 /// [`to_lowercase`]: #method.to_lowercase
546 #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")]
548 pub fn to_ascii_lowercase(&self) -> String {
549 let mut bytes = self.as_bytes().to_vec();
550 bytes.make_ascii_lowercase();
551 // make_ascii_lowercase() preserves the UTF-8 invariant.
552 unsafe { String::from_utf8_unchecked(bytes) }
556 /// Converts a boxed slice of bytes to a boxed string slice without checking
557 /// that the string contains valid UTF-8.
564 /// let smile_utf8 = Box::new([226, 152, 186]);
565 /// let smile = unsafe { std::str::from_boxed_utf8_unchecked(smile_utf8) };
567 /// assert_eq!("☺", &*smile);
569 #[stable(feature = "str_box_extras", since = "1.20.0")]
571 pub unsafe fn from_boxed_utf8_unchecked(v: Box<[u8]>) -> Box<str> {
572 Box::from_raw(Box::into_raw(v) as *mut str)