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, BorrowMut};
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;
71 #[stable(feature = "str_escape", since = "1.34.0")]
72 pub use core::str::{EscapeDebug, EscapeDefault, EscapeUnicode};
74 #[unstable(feature = "slice_concat_ext",
75 reason = "trait should not have to exist",
77 impl<S: Borrow<str>> SliceConcatExt<str> for [S] {
80 fn concat(&self) -> String {
84 fn join(&self, sep: &str) -> String {
86 String::from_utf8_unchecked( join_generic_copy(self, sep.as_bytes()) )
90 fn connect(&self, sep: &str) -> String {
95 macro_rules! spezialize_for_lengths {
96 ($separator:expr, $target:expr, $iter:expr; $($num:expr),*) => {
97 let mut target = $target;
99 let sep_bytes = $separator;
100 match $separator.len() {
102 // loops with hardcoded sizes run much faster
103 // specialize the cases with small separator lengths
106 copy_slice_and_advance!(target, sep_bytes);
107 copy_slice_and_advance!(target, s.borrow().as_ref());
112 // arbitrary non-zero size fallback
114 copy_slice_and_advance!(target, sep_bytes);
115 copy_slice_and_advance!(target, s.borrow().as_ref());
122 macro_rules! copy_slice_and_advance {
123 ($target:expr, $bytes:expr) => {
124 let len = $bytes.len();
125 let (head, tail) = {$target}.split_at_mut(len);
126 head.copy_from_slice($bytes);
131 // Optimized join implementation that works for both Vec<T> (T: Copy) and String's inner vec
132 // Currently (2018-05-13) there is a bug with type inference and specialization (see issue #36262)
133 // For this reason SliceConcatExt<T> is not specialized for T: Copy and SliceConcatExt<str> is the
134 // only user of this function. It is left in place for the time when that is fixed.
136 // the bounds for String-join are S: Borrow<str> and for Vec-join Borrow<[T]>
137 // [T] and str both impl AsRef<[T]> for some T
138 // => s.borrow().as_ref() and we always have slices
139 fn join_generic_copy<B, T, S>(slice: &[S], sep: &[T]) -> Vec<T>
142 B: AsRef<[T]> + ?Sized,
145 let sep_len = sep.len();
146 let mut iter = slice.iter();
148 // the first slice is the only one without a separator preceding it
149 let first = match iter.next() {
150 Some(first) => first,
151 None => return vec![],
154 // compute the exact total length of the joined Vec
155 // if the `len` calculation overflows, we'll panic
156 // we would have run out of memory anyway and the rest of the function requires
157 // the entire Vec pre-allocated for safety
158 let len = sep_len.checked_mul(iter.len()).and_then(|n| {
160 .map(|s| s.borrow().as_ref().len())
161 .try_fold(n, usize::checked_add)
162 }).expect("attempt to join into collection with len > usize::MAX");
164 // crucial for safety
165 let mut result = Vec::with_capacity(len);
166 assert!(result.capacity() >= len);
168 result.extend_from_slice(first.borrow().as_ref());
172 let pos = result.len();
173 let target = result.get_unchecked_mut(pos..len);
175 // copy separator and slices over without bounds checks
176 // generate loops with hardcoded offsets for small separators
177 // massive improvements possible (~ x2)
178 spezialize_for_lengths!(sep, target, iter; 0, 1, 2, 3, 4);
185 #[stable(feature = "rust1", since = "1.0.0")]
186 impl Borrow<str> for String {
188 fn borrow(&self) -> &str {
193 #[stable(feature = "string_borrow_mut", since = "1.36.0")]
194 impl BorrowMut<str> for String {
196 fn borrow_mut(&mut self) -> &mut str {
201 #[stable(feature = "rust1", since = "1.0.0")]
202 impl ToOwned for str {
205 fn to_owned(&self) -> String {
206 unsafe { String::from_utf8_unchecked(self.as_bytes().to_owned()) }
209 fn clone_into(&self, target: &mut String) {
210 let mut b = mem::replace(target, String::new()).into_bytes();
211 self.as_bytes().clone_into(&mut b);
212 *target = unsafe { String::from_utf8_unchecked(b) }
216 /// Methods for string slices.
217 #[lang = "str_alloc"]
220 /// Converts a `Box<str>` into a `Box<[u8]>` without copying or allocating.
227 /// let s = "this is a string";
228 /// let boxed_str = s.to_owned().into_boxed_str();
229 /// let boxed_bytes = boxed_str.into_boxed_bytes();
230 /// assert_eq!(*boxed_bytes, *s.as_bytes());
232 #[stable(feature = "str_box_extras", since = "1.20.0")]
234 pub fn into_boxed_bytes(self: Box<str>) -> Box<[u8]> {
238 /// Replaces all matches of a pattern with another string.
240 /// `replace` creates a new [`String`], and copies the data from this string slice into it.
241 /// While doing so, it attempts to find matches of a pattern. If it finds any, it
242 /// replaces them with the replacement string slice.
244 /// [`String`]: string/struct.String.html
251 /// let s = "this is old";
253 /// assert_eq!("this is new", s.replace("old", "new"));
256 /// When the pattern doesn't match:
259 /// let s = "this is old";
260 /// assert_eq!(s, s.replace("cookie monster", "little lamb"));
262 #[must_use = "this returns the replaced string as a new allocation, \
263 without modifying the original"]
264 #[stable(feature = "rust1", since = "1.0.0")]
266 pub fn replace<'a, P: Pattern<'a>>(&'a self, from: P, to: &str) -> String {
267 let mut result = String::new();
268 let mut last_end = 0;
269 for (start, part) in self.match_indices(from) {
270 result.push_str(unsafe { self.get_unchecked(last_end..start) });
272 last_end = start + part.len();
274 result.push_str(unsafe { self.get_unchecked(last_end..self.len()) });
278 /// Replaces first N matches of a pattern with another string.
280 /// `replacen` creates a new [`String`], and copies the data from this string slice into it.
281 /// While doing so, it attempts to find matches of a pattern. If it finds any, it
282 /// replaces them with the replacement string slice at most `count` times.
284 /// [`String`]: string/struct.String.html
291 /// let s = "foo foo 123 foo";
292 /// assert_eq!("new new 123 foo", s.replacen("foo", "new", 2));
293 /// assert_eq!("faa fao 123 foo", s.replacen('o', "a", 3));
294 /// assert_eq!("foo foo new23 foo", s.replacen(char::is_numeric, "new", 1));
297 /// When the pattern doesn't match:
300 /// let s = "this is old";
301 /// assert_eq!(s, s.replacen("cookie monster", "little lamb", 10));
303 #[must_use = "this returns the replaced string as a new allocation, \
304 without modifying the original"]
305 #[stable(feature = "str_replacen", since = "1.16.0")]
306 pub fn replacen<'a, P: Pattern<'a>>(&'a self, pat: P, to: &str, count: usize) -> String {
307 // Hope to reduce the times of re-allocation
308 let mut result = String::with_capacity(32);
309 let mut last_end = 0;
310 for (start, part) in self.match_indices(pat).take(count) {
311 result.push_str(unsafe { self.get_unchecked(last_end..start) });
313 last_end = start + part.len();
315 result.push_str(unsafe { self.get_unchecked(last_end..self.len()) });
319 /// Returns the lowercase equivalent of this string slice, as a new [`String`].
321 /// 'Lowercase' is defined according to the terms of the Unicode Derived Core Property
324 /// Since some characters can expand into multiple characters when changing
325 /// the case, this function returns a [`String`] instead of modifying the
326 /// parameter in-place.
328 /// [`String`]: string/struct.String.html
337 /// assert_eq!("hello", s.to_lowercase());
340 /// A tricky example, with sigma:
345 /// assert_eq!("σ", sigma.to_lowercase());
347 /// // but at the end of a word, it's ς, not σ:
348 /// let odysseus = "ὈΔΥΣΣΕΎΣ";
350 /// assert_eq!("ὀδυσσεύς", odysseus.to_lowercase());
353 /// Languages without case are not changed:
356 /// let new_year = "农历新年";
358 /// assert_eq!(new_year, new_year.to_lowercase());
360 #[stable(feature = "unicode_case_mapping", since = "1.2.0")]
361 pub fn to_lowercase(&self) -> String {
362 let mut s = String::with_capacity(self.len());
363 for (i, c) in self[..].char_indices() {
365 // Σ maps to σ, except at the end of a word where it maps to ς.
366 // This is the only conditional (contextual) but language-independent mapping
367 // in `SpecialCasing.txt`,
368 // so hard-code it rather than have a generic "condition" mechanism.
369 // See https://github.com/rust-lang/rust/issues/26035
370 map_uppercase_sigma(self, i, &mut s)
372 match conversions::to_lower(c) {
373 [a, '\0', _] => s.push(a),
388 fn map_uppercase_sigma(from: &str, i: usize, to: &mut String) {
389 // See http://www.unicode.org/versions/Unicode7.0.0/ch03.pdf#G33992
390 // for the definition of `Final_Sigma`.
391 debug_assert!('Σ'.len_utf8() == 2);
392 let is_word_final = case_ignoreable_then_cased(from[..i].chars().rev()) &&
393 !case_ignoreable_then_cased(from[i + 2..].chars());
394 to.push_str(if is_word_final { "ς" } else { "σ" });
397 fn case_ignoreable_then_cased<I: Iterator<Item = char>>(iter: I) -> bool {
398 use core::unicode::derived_property::{Cased, Case_Ignorable};
399 match iter.skip_while(|&c| Case_Ignorable(c)).next() {
406 /// Returns the uppercase equivalent of this string slice, as a new [`String`].
408 /// 'Uppercase' is defined according to the terms of the Unicode Derived Core Property
411 /// Since some characters can expand into multiple characters when changing
412 /// the case, this function returns a [`String`] instead of modifying the
413 /// parameter in-place.
415 /// [`String`]: string/struct.String.html
424 /// assert_eq!("HELLO", s.to_uppercase());
427 /// Scripts without case are not changed:
430 /// let new_year = "农历新年";
432 /// assert_eq!(new_year, new_year.to_uppercase());
435 /// One character can become multiple:
437 /// let s = "tschüß";
439 /// assert_eq!("TSCHÜSS", s.to_uppercase());
441 #[stable(feature = "unicode_case_mapping", since = "1.2.0")]
442 pub fn to_uppercase(&self) -> String {
443 let mut s = String::with_capacity(self.len());
444 for c in self[..].chars() {
445 match conversions::to_upper(c) {
446 [a, '\0', _] => s.push(a),
461 /// Converts a [`Box<str>`] into a [`String`] without copying or allocating.
463 /// [`String`]: string/struct.String.html
464 /// [`Box<str>`]: boxed/struct.Box.html
471 /// let string = String::from("birthday gift");
472 /// let boxed_str = string.clone().into_boxed_str();
474 /// assert_eq!(boxed_str.into_string(), string);
476 #[stable(feature = "box_str", since = "1.4.0")]
478 pub fn into_string(self: Box<str>) -> String {
479 let slice = Box::<[u8]>::from(self);
480 unsafe { String::from_utf8_unchecked(slice.into_vec()) }
483 /// Creates a new [`String`] by repeating a string `n` times.
487 /// This function will panic if the capacity would overflow.
489 /// [`String`]: string/struct.String.html
496 /// assert_eq!("abc".repeat(4), String::from("abcabcabcabc"));
499 /// A panic upon overflow:
503 /// // this will panic at runtime
504 /// "0123456789abcdef".repeat(usize::max_value());
507 #[stable(feature = "repeat_str", since = "1.16.0")]
508 pub fn repeat(&self, n: usize) -> String {
509 unsafe { String::from_utf8_unchecked(self.as_bytes().repeat(n)) }
512 /// Returns a copy of this string where each character is mapped to its
513 /// ASCII upper case equivalent.
515 /// ASCII letters 'a' to 'z' are mapped to 'A' to 'Z',
516 /// but non-ASCII letters are unchanged.
518 /// To uppercase the value in-place, use [`make_ascii_uppercase`].
520 /// To uppercase ASCII characters in addition to non-ASCII characters, use
521 /// [`to_uppercase`].
526 /// let s = "Grüße, Jürgen ❤";
528 /// assert_eq!("GRüßE, JüRGEN ❤", s.to_ascii_uppercase());
531 /// [`make_ascii_uppercase`]: #method.make_ascii_uppercase
532 /// [`to_uppercase`]: #method.to_uppercase
533 #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")]
535 pub fn to_ascii_uppercase(&self) -> String {
536 let mut bytes = self.as_bytes().to_vec();
537 bytes.make_ascii_uppercase();
538 // make_ascii_uppercase() preserves the UTF-8 invariant.
539 unsafe { String::from_utf8_unchecked(bytes) }
542 /// Returns a copy of this string where each character is mapped to its
543 /// ASCII lower case equivalent.
545 /// ASCII letters 'A' to 'Z' are mapped to 'a' to 'z',
546 /// but non-ASCII letters are unchanged.
548 /// To lowercase the value in-place, use [`make_ascii_lowercase`].
550 /// To lowercase ASCII characters in addition to non-ASCII characters, use
551 /// [`to_lowercase`].
556 /// let s = "Grüße, Jürgen ❤";
558 /// assert_eq!("grüße, jürgen ❤", s.to_ascii_lowercase());
561 /// [`make_ascii_lowercase`]: #method.make_ascii_lowercase
562 /// [`to_lowercase`]: #method.to_lowercase
563 #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")]
565 pub fn to_ascii_lowercase(&self) -> String {
566 let mut bytes = self.as_bytes().to_vec();
567 bytes.make_ascii_lowercase();
568 // make_ascii_lowercase() preserves the UTF-8 invariant.
569 unsafe { String::from_utf8_unchecked(bytes) }
573 /// Converts a boxed slice of bytes to a boxed string slice without checking
574 /// that the string contains valid UTF-8.
581 /// let smile_utf8 = Box::new([226, 152, 186]);
582 /// let smile = unsafe { std::str::from_boxed_utf8_unchecked(smile_utf8) };
584 /// assert_eq!("☺", &*smile);
586 #[stable(feature = "str_box_extras", since = "1.20.0")]
588 pub unsafe fn from_boxed_utf8_unchecked(v: Box<[u8]>) -> Box<str> {
589 Box::from_raw(Box::into_raw(v) as *mut str)