library/alloc/src/str.rs

   1 //! Unicode string slices.
   2 //!
   3 //! *[See also the `str` primitive type](str).*
   4 //!
   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.
   7 //!
   8 //! # Basic Usage
   9 //!
  10 //! A basic string declaration of `&str` type:
  11 //!
  12 //! ```
  13 //! let hello_world = "Hello, World!";
  14 //! ```
  15 //!
  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:
  20 //!
  21 //! ```
  22 //! let hello_world: &'static str = "Hello, world!";
  23 //! ```
  24
  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)]
  29
  30 use core::borrow::{Borrow, BorrowMut};
  31 use core::iter::FusedIterator;
  32 use core::mem;
  33 use core::ptr;
  34 use core::str::pattern::{DoubleEndedSearcher, Pattern, ReverseSearcher, Searcher};
  35 use core::unicode::conversions;
  36
  37 use crate::borrow::ToOwned;
  38 use crate::boxed::Box;
  39 use crate::slice::{Concat, Join, SliceIndex};
  40 use crate::string::String;
  41 use crate::vec::Vec;
  42
  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};
  59 #[allow(deprecated)]
  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};
  72
  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] {
  77     type Output = String;
  78
  79     fn concat(slice: &Self) -> String {
  80         Join::join(slice, "")
  81     }
  82 }
  83
  84 #[unstable(feature = "slice_concat_ext", issue = "27747")]
  85 impl<S: Borrow<str>> Join<&str> for [S] {
  86     type Output = String;
  87
  88     fn join(slice: &Self, sep: &str) -> String {
  89         unsafe { String::from_utf8_unchecked(join_generic_copy(slice, sep.as_bytes())) }
  90     }
  91 }
  92
  93 macro_rules! spezialize_for_lengths {
  94     ($separator:expr, $target:expr, $iter:expr; $($num:expr),*) => {
  95         let mut target = $target;
  96         let iter = $iter;
  97         let sep_bytes = $separator;
  98         match $separator.len() {
  99             $(
 100                 // loops with hardcoded sizes run much faster
 101                 // specialize the cases with small separator lengths
 102                 $num => {
 103                     for s in iter {
 104                         copy_slice_and_advance!(target, sep_bytes);
 105                         copy_slice_and_advance!(target, s.borrow().as_ref());
 106                     }
 107                 },
 108             )*
 109             _ => {
 110                 // arbitrary non-zero size fallback
 111                 for s in iter {
 112                     copy_slice_and_advance!(target, sep_bytes);
 113                     copy_slice_and_advance!(target, s.borrow().as_ref());
 114                 }
 115             }
 116         }
 117     };
 118 }
 119
 120 macro_rules! copy_slice_and_advance {
 121     ($target:expr, $bytes:expr) => {
 122         let len = $bytes.len();
 123         let (head, tail) = { $target }.split_at_mut(len);
 124         head.copy_from_slice($bytes);
 125         $target = tail;
 126     };
 127 }
 128
 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.
 133 //
 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>
 138 where
 139     T: Copy,
 140     B: AsRef<[T]> + ?Sized,
 141     S: Borrow<B>,
 142 {
 143     let sep_len = sep.len();
 144     let mut iter = slice.iter();
 145
 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![],
 150     };
 151
 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
 157         .checked_mul(iter.len())
 158         .and_then(|n| {
 159             slice.iter().map(|s| s.borrow().as_ref().len()).try_fold(n, usize::checked_add)
 160         })
 161         .expect("attempt to join into collection with len > usize::MAX");
 162
 163     // crucial for safety
 164     let mut result = Vec::with_capacity(len);
 165     assert!(result.capacity() >= len);
 166
 167     result.extend_from_slice(first.borrow().as_ref());
 168
 169     unsafe {
 170         {
 171             let pos = result.len();
 172             let target = result.get_unchecked_mut(pos..len);
 173
 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);
 178         }
 179         result.set_len(len);
 180     }
 181     result
 182 }
 183
 184 #[stable(feature = "rust1", since = "1.0.0")]
 185 impl Borrow<str> for String {
 186     #[inline]
 187     fn borrow(&self) -> &str {
 188         &self[..]
 189     }
 190 }
 191
 192 #[stable(feature = "string_borrow_mut", since = "1.36.0")]
 193 impl BorrowMut<str> for String {
 194     #[inline]
 195     fn borrow_mut(&mut self) -> &mut str {
 196         &mut self[..]
 197     }
 198 }
 199
 200 #[stable(feature = "rust1", since = "1.0.0")]
 201 impl ToOwned for str {
 202     type Owned = String;
 203     #[inline]
 204     fn to_owned(&self) -> String {
 205         unsafe { String::from_utf8_unchecked(self.as_bytes().to_owned()) }
 206     }
 207
 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) }
 212     }
 213 }
 214
 215 /// Methods for string slices.
 216 #[lang = "str_alloc"]
 217 #[cfg(not(test))]
 218 impl str {
 219     /// Converts a `Box<str>` into a `Box<[u8]>` without copying or allocating.
 220     ///
 221     /// # Examples
 222     ///
 223     /// Basic usage:
 224     ///
 225     /// ```
 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());
 230     /// ```
 231     #[stable(feature = "str_box_extras", since = "1.20.0")]
 232     #[inline]
 233     pub fn into_boxed_bytes(self: Box<str>) -> Box<[u8]> {
 234         self.into()
 235     }
 236
 237     /// Replaces all matches of a pattern with another string.
 238     ///
 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.
 242     ///
 243     /// # Examples
 244     ///
 245     /// Basic usage:
 246     ///
 247     /// ```
 248     /// let s = "this is old";
 249     ///
 250     /// assert_eq!("this is new", s.replace("old", "new"));
 251     /// ```
 252     ///
 253     /// When the pattern doesn't match:
 254     ///
 255     /// ```
 256     /// let s = "this is old";
 257     /// assert_eq!(s, s.replace("cookie monster", "little lamb"));
 258     /// ```
 259     #[must_use = "this returns the replaced string as a new allocation, \
 260                   without modifying the original"]
 261     #[stable(feature = "rust1", since = "1.0.0")]
 262     #[inline]
 263     pub fn replace<'a, P: Pattern<'a>>(&'a self, from: P, to: &str) -> String {
 264         let mut result = String::new();
 265         let mut last_end = 0;
 266         for (start, part) in self.match_indices(from) {
 267             result.push_str(unsafe { self.get_unchecked(last_end..start) });
 268             result.push_str(to);
 269             last_end = start + part.len();
 270         }
 271         result.push_str(unsafe { self.get_unchecked(last_end..self.len()) });
 272         result
 273     }
 274
 275     /// Replaces first N matches of a pattern with another string.
 276     ///
 277     /// `replacen` creates a new [`String`], and copies the data from this string slice into it.
 278     /// While doing so, it attempts to find matches of a pattern. If it finds any, it
 279     /// replaces them with the replacement string slice at most `count` times.
 280     ///
 281     /// # Examples
 282     ///
 283     /// Basic usage:
 284     ///
 285     /// ```
 286     /// let s = "foo foo 123 foo";
 287     /// assert_eq!("new new 123 foo", s.replacen("foo", "new", 2));
 288     /// assert_eq!("faa fao 123 foo", s.replacen('o', "a", 3));
 289     /// assert_eq!("foo foo new23 foo", s.replacen(char::is_numeric, "new", 1));
 290     /// ```
 291     ///
 292     /// When the pattern doesn't match:
 293     ///
 294     /// ```
 295     /// let s = "this is old";
 296     /// assert_eq!(s, s.replacen("cookie monster", "little lamb", 10));
 297     /// ```
 298     #[must_use = "this returns the replaced string as a new allocation, \
 299                   without modifying the original"]
 300     #[stable(feature = "str_replacen", since = "1.16.0")]
 301     pub fn replacen<'a, P: Pattern<'a>>(&'a self, pat: P, to: &str, count: usize) -> String {
 302         // Hope to reduce the times of re-allocation
 303         let mut result = String::with_capacity(32);
 304         let mut last_end = 0;
 305         for (start, part) in self.match_indices(pat).take(count) {
 306             result.push_str(unsafe { self.get_unchecked(last_end..start) });
 307             result.push_str(to);
 308             last_end = start + part.len();
 309         }
 310         result.push_str(unsafe { self.get_unchecked(last_end..self.len()) });
 311         result
 312     }
 313
 314     /// Returns the lowercase equivalent of this string slice, as a new [`String`].
 315     ///
 316     /// 'Lowercase' is defined according to the terms of the Unicode Derived Core Property
 317     /// `Lowercase`.
 318     ///
 319     /// Since some characters can expand into multiple characters when changing
 320     /// the case, this function returns a [`String`] instead of modifying the
 321     /// parameter in-place.
 322     ///
 323     /// # Examples
 324     ///
 325     /// Basic usage:
 326     ///
 327     /// ```
 328     /// let s = "HELLO";
 329     ///
 330     /// assert_eq!("hello", s.to_lowercase());
 331     /// ```
 332     ///
 333     /// A tricky example, with sigma:
 334     ///
 335     /// ```
 336     /// let sigma = "Σ";
 337     ///
 338     /// assert_eq!("σ", sigma.to_lowercase());
 339     ///
 340     /// // but at the end of a word, it's ς, not σ:
 341     /// let odysseus = "ὈΔΥΣΣΕΎΣ";
 342     ///
 343     /// assert_eq!("ὀδυσσεύς", odysseus.to_lowercase());
 344     /// ```
 345     ///
 346     /// Languages without case are not changed:
 347     ///
 348     /// ```
 349     /// let new_year = "农历新年";
 350     ///
 351     /// assert_eq!(new_year, new_year.to_lowercase());
 352     /// ```
 353     #[stable(feature = "unicode_case_mapping", since = "1.2.0")]
 354     pub fn to_lowercase(&self) -> String {
 355         let mut s = String::with_capacity(self.len());
 356         for (i, c) in self[..].char_indices() {
 357             if c == 'Σ' {
 358                 // Σ maps to σ, except at the end of a word where it maps to ς.
 359                 // This is the only conditional (contextual) but language-independent mapping
 360                 // in `SpecialCasing.txt`,
 361                 // so hard-code it rather than have a generic "condition" mechanism.
 362                 // See https://github.com/rust-lang/rust/issues/26035
 363                 map_uppercase_sigma(self, i, &mut s)
 364             } else {
 365                 match conversions::to_lower(c) {
 366                     [a, '\0', _] => s.push(a),
 367                     [a, b, '\0'] => {
 368                         s.push(a);
 369                         s.push(b);
 370                     }
 371                     [a, b, c] => {
 372                         s.push(a);
 373                         s.push(b);
 374                         s.push(c);
 375                     }
 376                 }
 377             }
 378         }
 379         return s;
 380
 381         fn map_uppercase_sigma(from: &str, i: usize, to: &mut String) {
 382             // See http://www.unicode.org/versions/Unicode7.0.0/ch03.pdf#G33992
 383             // for the definition of `Final_Sigma`.
 384             debug_assert!('Σ'.len_utf8() == 2);
 385             let is_word_final = case_ignoreable_then_cased(from[..i].chars().rev())
 386                 && !case_ignoreable_then_cased(from[i + 2..].chars());
 387             to.push_str(if is_word_final { "ς" } else { "σ" });
 388         }
 389
 390         fn case_ignoreable_then_cased<I: Iterator<Item = char>>(iter: I) -> bool {
 391             use core::unicode::{Case_Ignorable, Cased};
 392             match iter.skip_while(|&c| Case_Ignorable(c)).next() {
 393                 Some(c) => Cased(c),
 394                 None => false,
 395             }
 396         }
 397     }
 398
 399     /// Returns the uppercase equivalent of this string slice, as a new [`String`].
 400     ///
 401     /// 'Uppercase' is defined according to the terms of the Unicode Derived Core Property
 402     /// `Uppercase`.
 403     ///
 404     /// Since some characters can expand into multiple characters when changing
 405     /// the case, this function returns a [`String`] instead of modifying the
 406     /// parameter in-place.
 407     ///
 408     /// # Examples
 409     ///
 410     /// Basic usage:
 411     ///
 412     /// ```
 413     /// let s = "hello";
 414     ///
 415     /// assert_eq!("HELLO", s.to_uppercase());
 416     /// ```
 417     ///
 418     /// Scripts without case are not changed:
 419     ///
 420     /// ```
 421     /// let new_year = "农历新年";
 422     ///
 423     /// assert_eq!(new_year, new_year.to_uppercase());
 424     /// ```
 425     ///
 426     /// One character can become multiple:
 427     /// ```
 428     /// let s = "tschüß";
 429     ///
 430     /// assert_eq!("TSCHÜSS", s.to_uppercase());
 431     /// ```
 432     #[stable(feature = "unicode_case_mapping", since = "1.2.0")]
 433     pub fn to_uppercase(&self) -> String {
 434         let mut s = String::with_capacity(self.len());
 435         for c in self[..].chars() {
 436             match conversions::to_upper(c) {
 437                 [a, '\0', _] => s.push(a),
 438                 [a, b, '\0'] => {
 439                     s.push(a);
 440                     s.push(b);
 441                 }
 442                 [a, b, c] => {
 443                     s.push(a);
 444                     s.push(b);
 445                     s.push(c);
 446                 }
 447             }
 448         }
 449         s
 450     }
 451
 452     /// Converts a [`Box<str>`] into a [`String`] without copying or allocating.
 453     ///
 454     /// # Examples
 455     ///
 456     /// Basic usage:
 457     ///
 458     /// ```
 459     /// let string = String::from("birthday gift");
 460     /// let boxed_str = string.clone().into_boxed_str();
 461     ///
 462     /// assert_eq!(boxed_str.into_string(), string);
 463     /// ```
 464     #[stable(feature = "box_str", since = "1.4.0")]
 465     #[inline]
 466     pub fn into_string(self: Box<str>) -> String {
 467         let slice = Box::<[u8]>::from(self);
 468         unsafe { String::from_utf8_unchecked(slice.into_vec()) }
 469     }
 470
 471     /// Creates a new [`String`] by repeating a string `n` times.
 472     ///
 473     /// # Panics
 474     ///
 475     /// This function will panic if the capacity would overflow.
 476     ///
 477     /// # Examples
 478     ///
 479     /// Basic usage:
 480     ///
 481     /// ```
 482     /// assert_eq!("abc".repeat(4), String::from("abcabcabcabc"));
 483     /// ```
 484     ///
 485     /// A panic upon overflow:
 486     ///
 487     /// ```should_panic
 488     /// // this will panic at runtime
 489     /// "0123456789abcdef".repeat(usize::MAX);
 490     /// ```
 491     #[stable(feature = "repeat_str", since = "1.16.0")]
 492     pub fn repeat(&self, n: usize) -> String {
 493         unsafe { String::from_utf8_unchecked(self.as_bytes().repeat(n)) }
 494     }
 495
 496     /// Returns a copy of this string where each character is mapped to its
 497     /// ASCII upper case equivalent.
 498     ///
 499     /// ASCII letters 'a' to 'z' are mapped to 'A' to 'Z',
 500     /// but non-ASCII letters are unchanged.
 501     ///
 502     /// To uppercase the value in-place, use [`make_ascii_uppercase`].
 503     ///
 504     /// To uppercase ASCII characters in addition to non-ASCII characters, use
 505     /// [`to_uppercase`].
 506     ///
 507     /// # Examples
 508     ///
 509     /// ```
 510     /// let s = "Grüße, Jürgen ❤";
 511     ///
 512     /// assert_eq!("GRüßE, JüRGEN ❤", s.to_ascii_uppercase());
 513     /// ```
 514     ///
 515     /// [`make_ascii_uppercase`]: str::make_ascii_uppercase
 516     /// [`to_uppercase`]: #method.to_uppercase
 517     #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")]
 518     #[inline]
 519     pub fn to_ascii_uppercase(&self) -> String {
 520         let mut bytes = self.as_bytes().to_vec();
 521         bytes.make_ascii_uppercase();
 522         // make_ascii_uppercase() preserves the UTF-8 invariant.
 523         unsafe { String::from_utf8_unchecked(bytes) }
 524     }
 525
 526     /// Returns a copy of this string where each character is mapped to its
 527     /// ASCII lower case equivalent.
 528     ///
 529     /// ASCII letters 'A' to 'Z' are mapped to 'a' to 'z',
 530     /// but non-ASCII letters are unchanged.
 531     ///
 532     /// To lowercase the value in-place, use [`make_ascii_lowercase`].
 533     ///
 534     /// To lowercase ASCII characters in addition to non-ASCII characters, use
 535     /// [`to_lowercase`].
 536     ///
 537     /// # Examples
 538     ///
 539     /// ```
 540     /// let s = "Grüße, Jürgen ❤";
 541     ///
 542     /// assert_eq!("grüße, jürgen ❤", s.to_ascii_lowercase());
 543     /// ```
 544     ///
 545     /// [`make_ascii_lowercase`]: str::make_ascii_lowercase
 546     /// [`to_lowercase`]: #method.to_lowercase
 547     #[stable(feature = "ascii_methods_on_intrinsics", since = "1.23.0")]
 548     #[inline]
 549     pub fn to_ascii_lowercase(&self) -> String {
 550         let mut bytes = self.as_bytes().to_vec();
 551         bytes.make_ascii_lowercase();
 552         // make_ascii_lowercase() preserves the UTF-8 invariant.
 553         unsafe { String::from_utf8_unchecked(bytes) }
 554     }
 555 }
 556
 557 /// Converts a boxed slice of bytes to a boxed string slice without checking
 558 /// that the string contains valid UTF-8.
 559 ///
 560 /// # Examples
 561 ///
 562 /// Basic usage:
 563 ///
 564 /// ```
 565 /// let smile_utf8 = Box::new([226, 152, 186]);
 566 /// let smile = unsafe { std::str::from_boxed_utf8_unchecked(smile_utf8) };
 567 ///
 568 /// assert_eq!("☺", &*smile);
 569 /// ```
 570 #[stable(feature = "str_box_extras", since = "1.20.0")]
 571 #[inline]
 572 pub unsafe fn from_boxed_utf8_unchecked(v: Box<[u8]>) -> Box<str> {
 573     unsafe { Box::from_raw(Box::into_raw(v) as *mut str) }
 574 }