blob: 4c64019de097ef8a7c2a5295df19c60681f8c2ef [file] [log] [blame]
// Copyright 2012-2014 The Rust Project Developers. See the COPYRIGHT
// file at the top-level directory of this distribution and at
// http://rust-lang.org/COPYRIGHT.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// http://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or http://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
//! Unicode string slices.
//!
//! *[See also the `str` primitive type](../../std/primitive.str.html).*
#![stable(feature = "rust1", since = "1.0.0")]
// Many of the usings in this module are only used in the test configuration.
// It's cleaner to just turn off the unused_imports warning than to fix them.
#![allow(unused_imports)]
use core::str as core_str;
use core::str::pattern::Pattern;
use core::str::pattern::{Searcher, ReverseSearcher, DoubleEndedSearcher};
use core::mem;
use rustc_unicode::str::{UnicodeStr, Utf16Encoder};
use vec_deque::VecDeque;
use borrow::{Borrow, ToOwned};
use string::String;
use rustc_unicode;
use vec::Vec;
use slice::SliceConcatExt;
use boxed::Box;
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{FromStr, Utf8Error};
#[allow(deprecated)]
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{Lines, LinesAny};
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{Split, RSplit};
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{SplitN, RSplitN};
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{SplitTerminator, RSplitTerminator};
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{Matches, RMatches};
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{MatchIndices, RMatchIndices};
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{from_utf8, Chars, CharIndices, Bytes};
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::{from_utf8_unchecked, ParseBoolError};
#[stable(feature = "rust1", since = "1.0.0")]
pub use rustc_unicode::str::SplitWhitespace;
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::str::pattern;
#[unstable(feature = "slice_concat_ext",
reason = "trait should not have to exist",
issue = "27747")]
impl<S: Borrow<str>> SliceConcatExt<str> for [S] {
type Output = String;
fn concat(&self) -> String {
if self.is_empty() {
return String::new();
}
// `len` calculation may overflow but push_str will check boundaries
let len = self.iter().map(|s| s.borrow().len()).sum();
let mut result = String::with_capacity(len);
for s in self {
result.push_str(s.borrow())
}
result
}
fn join(&self, sep: &str) -> String {
if self.is_empty() {
return String::new();
}
// concat is faster
if sep.is_empty() {
return self.concat();
}
// this is wrong without the guarantee that `self` is non-empty
// `len` calculation may overflow but push_str but will check boundaries
let len = sep.len() * (self.len() - 1) +
self.iter().map(|s| s.borrow().len()).sum::<usize>();
let mut result = String::with_capacity(len);
let mut first = true;
for s in self {
if first {
first = false;
} else {
result.push_str(sep);
}
result.push_str(s.borrow());
}
result
}
fn connect(&self, sep: &str) -> String {
self.join(sep)
}
}
/// External iterator for a string's UTF-16 code units.
///
/// For use with the `std::iter` module.
#[derive(Clone)]
#[stable(feature = "encode_utf16", since = "1.8.0")]
pub struct EncodeUtf16<'a> {
encoder: Utf16Encoder<Chars<'a>>,
}
#[stable(feature = "rust1", since = "1.0.0")]
impl<'a> Iterator for EncodeUtf16<'a> {
type Item = u16;
#[inline]
fn next(&mut self) -> Option<u16> {
self.encoder.next()
}
#[inline]
fn size_hint(&self) -> (usize, Option<usize>) {
self.encoder.size_hint()
}
}
// Return the initial codepoint accumulator for the first byte.
// The first byte is special, only want bottom 5 bits for width 2, 4 bits
// for width 3, and 3 bits for width 4
macro_rules! utf8_first_byte {
($byte:expr, $width:expr) => (($byte & (0x7F >> $width)) as u32)
}
// return the value of $ch updated with continuation byte $byte
macro_rules! utf8_acc_cont_byte {
($ch:expr, $byte:expr) => (($ch << 6) | ($byte & 63) as u32)
}
#[stable(feature = "rust1", since = "1.0.0")]
impl Borrow<str> for String {
#[inline]
fn borrow(&self) -> &str {
&self[..]
}
}
#[stable(feature = "rust1", since = "1.0.0")]
impl ToOwned for str {
type Owned = String;
fn to_owned(&self) -> String {
unsafe { String::from_utf8_unchecked(self.as_bytes().to_owned()) }
}
}
/// Methods for string slices.
#[lang = "str"]
#[cfg(not(test))]
impl str {
/// Returns the length of `self`.
///
/// This length is in bytes, not [`char`]s or graphemes. In other words,
/// it may not be what a human considers the length of the string.
///
/// [`char`]: primitive.char.html
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let len = "foo".len();
/// assert_eq!(3, len);
///
/// let len = "ƒoo".len(); // fancy f!
/// assert_eq!(4, len);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn len(&self) -> usize {
core_str::StrExt::len(self)
}
/// Returns true if this slice has a length of zero bytes.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = "";
/// assert!(s.is_empty());
///
/// let s = "not empty";
/// assert!(!s.is_empty());
/// ```
#[inline]
#[stable(feature = "rust1", since = "1.0.0")]
pub fn is_empty(&self) -> bool {
core_str::StrExt::is_empty(self)
}
/// Checks that `index`-th byte lies at the start and/or end of a
/// UTF-8 code point sequence.
///
/// The start and end of the string (when `index == self.len()`) are
/// considered to be
/// boundaries.
///
/// Returns `false` if `index` is greater than `self.len()`.
///
/// # Examples
///
/// ```
/// let s = "Löwe 老虎 Léopard";
/// assert!(s.is_char_boundary(0));
/// // start of `老`
/// assert!(s.is_char_boundary(6));
/// assert!(s.is_char_boundary(s.len()));
///
/// // second byte of `ö`
/// assert!(!s.is_char_boundary(2));
///
/// // third byte of `老`
/// assert!(!s.is_char_boundary(8));
/// ```
#[stable(feature = "is_char_boundary", since = "1.9.0")]
#[inline]
pub fn is_char_boundary(&self, index: usize) -> bool {
core_str::StrExt::is_char_boundary(self, index)
}
/// Converts a string slice to a byte slice.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let bytes = "bors".as_bytes();
/// assert_eq!(b"bors", bytes);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline(always)]
pub fn as_bytes(&self) -> &[u8] {
core_str::StrExt::as_bytes(self)
}
/// Converts a string slice to a raw pointer.
///
/// As string slices are a slice of bytes, the raw pointer points to a
/// [`u8`]. This pointer will be pointing to the first byte of the string
/// slice.
///
/// [`u8`]: primitive.u8.html
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = "Hello";
/// let ptr = s.as_ptr();
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn as_ptr(&self) -> *const u8 {
core_str::StrExt::as_ptr(self)
}
/// Creates a string slice from another string slice, bypassing safety
/// checks.
///
/// This new slice goes from `begin` to `end`, including `begin` but
/// excluding `end`.
///
/// To get a mutable string slice instead, see the
/// [`slice_mut_unchecked()`] method.
///
/// [`slice_mut_unchecked()`]: #method.slice_mut_unchecked
///
/// # Safety
///
/// Callers of this function are responsible that three preconditions are
/// satisfied:
///
/// * `begin` must come before `end`.
/// * `begin` and `end` must be byte positions within the string slice.
/// * `begin` and `end` must lie on UTF-8 sequence boundaries.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = "Löwe 老虎 Léopard";
///
/// unsafe {
/// assert_eq!("Löwe 老虎 Léopard", s.slice_unchecked(0, 21));
/// }
///
/// let s = "Hello, world!";
///
/// unsafe {
/// assert_eq!("world", s.slice_unchecked(7, 12));
/// }
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub unsafe fn slice_unchecked(&self, begin: usize, end: usize) -> &str {
core_str::StrExt::slice_unchecked(self, begin, end)
}
/// Creates a string slice from another string slice, bypassing safety
/// checks.
///
/// This new slice goes from `begin` to `end`, including `begin` but
/// excluding `end`.
///
/// To get an immutable string slice instead, see the
/// [`slice_unchecked()`] method.
///
/// [`slice_unchecked()`]: #method.slice_unchecked
///
/// # Safety
///
/// Callers of this function are responsible that three preconditions are
/// satisfied:
///
/// * `begin` must come before `end`.
/// * `begin` and `end` must be byte positions within the string slice.
/// * `begin` and `end` must lie on UTF-8 sequence boundaries.
#[stable(feature = "str_slice_mut", since = "1.5.0")]
#[inline]
pub unsafe fn slice_mut_unchecked(&mut self, begin: usize, end: usize) -> &mut str {
core_str::StrExt::slice_mut_unchecked(self, begin, end)
}
/// Divide one string slice into two at an index.
///
/// The argument, `mid`, should be a byte offset from the start of the
/// string. It must also be on the boundary of a UTF-8 code point.
///
/// The two slices returned go from the start of the string slice to `mid`,
/// and from `mid` to the end of the string slice.
///
/// To get mutable string slices instead, see the [`split_at_mut()`]
/// method.
///
/// [`split_at_mut()`]: #method.split_at_mut
///
/// # Panics
///
/// Panics if `mid` is not on a UTF-8 code point boundary, or if it is
/// beyond the last code point of the string slice.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = "Per Martin-Löf";
///
/// let (first, last) = s.split_at(3);
///
/// assert_eq!("Per", first);
/// assert_eq!(" Martin-Löf", last);
/// ```
#[inline]
#[stable(feature = "str_split_at", since = "1.4.0")]
pub fn split_at(&self, mid: usize) -> (&str, &str) {
core_str::StrExt::split_at(self, mid)
}
/// Divide one mutable string slice into two at an index.
///
/// The argument, `mid`, should be a byte offset from the start of the
/// string. It must also be on the boundary of a UTF-8 code point.
///
/// The two slices returned go from the start of the string slice to `mid`,
/// and from `mid` to the end of the string slice.
///
/// To get immutable string slices instead, see the [`split_at()`] method.
///
/// [`split_at()`]: #method.split_at
///
/// # Panics
///
/// Panics if `mid` is not on a UTF-8 code point boundary, or if it is
/// beyond the last code point of the string slice.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let mut s = "Per Martin-Löf".to_string();
///
/// let (first, last) = s.split_at_mut(3);
///
/// assert_eq!("Per", first);
/// assert_eq!(" Martin-Löf", last);
/// ```
#[inline]
#[stable(feature = "str_split_at", since = "1.4.0")]
pub fn split_at_mut(&mut self, mid: usize) -> (&mut str, &mut str) {
core_str::StrExt::split_at_mut(self, mid)
}
/// Returns an iterator over the `char`s of a string slice.
///
/// As a string slice consists of valid UTF-8, we can iterate through a
/// string slice by [`char`]. This method returns such an iterator.
///
/// It's important to remember that [`char`] represents a Unicode Scalar
/// Value, and may not match your idea of what a 'character' is. Iteration
/// over grapheme clusters may be what you actually want.
///
/// [`char`]: primitive.char.html
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let word = "goodbye";
///
/// let count = word.chars().count();
/// assert_eq!(7, count);
///
/// let mut chars = word.chars();
///
/// assert_eq!(Some('g'), chars.next());
/// assert_eq!(Some('o'), chars.next());
/// assert_eq!(Some('o'), chars.next());
/// assert_eq!(Some('d'), chars.next());
/// assert_eq!(Some('b'), chars.next());
/// assert_eq!(Some('y'), chars.next());
/// assert_eq!(Some('e'), chars.next());
///
/// assert_eq!(None, chars.next());
/// ```
///
/// Remember, [`char`]s may not match your human intuition about characters:
///
/// ```
/// let y = "y̆";
///
/// let mut chars = y.chars();
///
/// assert_eq!(Some('y'), chars.next()); // not 'y̆'
/// assert_eq!(Some('\u{0306}'), chars.next());
///
/// assert_eq!(None, chars.next());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn chars(&self) -> Chars {
core_str::StrExt::chars(self)
}
/// Returns an iterator over the [`char`]s of a string slice, and their
/// positions.
///
/// As a string slice consists of valid UTF-8, we can iterate through a
/// string slice by [`char`]. This method returns an iterator of both
/// these [`char`]s, as well as their byte positions.
///
/// The iterator yields tuples. The position is first, the [`char`] is
/// second.
///
/// [`char`]: primitive.char.html
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let word = "goodbye";
///
/// let count = word.char_indices().count();
/// assert_eq!(7, count);
///
/// let mut char_indices = word.char_indices();
///
/// assert_eq!(Some((0, 'g')), char_indices.next());
/// assert_eq!(Some((1, 'o')), char_indices.next());
/// assert_eq!(Some((2, 'o')), char_indices.next());
/// assert_eq!(Some((3, 'd')), char_indices.next());
/// assert_eq!(Some((4, 'b')), char_indices.next());
/// assert_eq!(Some((5, 'y')), char_indices.next());
/// assert_eq!(Some((6, 'e')), char_indices.next());
///
/// assert_eq!(None, char_indices.next());
/// ```
///
/// Remember, [`char`]s may not match your human intuition about characters:
///
/// ```
/// let y = "y̆";
///
/// let mut char_indices = y.char_indices();
///
/// assert_eq!(Some((0, 'y')), char_indices.next()); // not (0, 'y̆')
/// assert_eq!(Some((1, '\u{0306}')), char_indices.next());
///
/// assert_eq!(None, char_indices.next());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn char_indices(&self) -> CharIndices {
core_str::StrExt::char_indices(self)
}
/// An iterator over the bytes of a string slice.
///
/// As a string slice consists of a sequence of bytes, we can iterate
/// through a string slice by byte. This method returns such an iterator.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let mut bytes = "bors".bytes();
///
/// assert_eq!(Some(b'b'), bytes.next());
/// assert_eq!(Some(b'o'), bytes.next());
/// assert_eq!(Some(b'r'), bytes.next());
/// assert_eq!(Some(b's'), bytes.next());
///
/// assert_eq!(None, bytes.next());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn bytes(&self) -> Bytes {
core_str::StrExt::bytes(self)
}
/// Split a string slice by whitespace.
///
/// The iterator returned will return string slices that are sub-slices of
/// the original string slice, separated by any amount of whitespace.
///
/// 'Whitespace' is defined according to the terms of the Unicode Derived
/// Core Property `White_Space`.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let mut iter = "A few words".split_whitespace();
///
/// assert_eq!(Some("A"), iter.next());
/// assert_eq!(Some("few"), iter.next());
/// assert_eq!(Some("words"), iter.next());
///
/// assert_eq!(None, iter.next());
/// ```
///
/// All kinds of whitespace are considered:
///
/// ```
/// let mut iter = " Mary had\ta\u{2009}little \n\t lamb".split_whitespace();
/// assert_eq!(Some("Mary"), iter.next());
/// assert_eq!(Some("had"), iter.next());
/// assert_eq!(Some("a"), iter.next());
/// assert_eq!(Some("little"), iter.next());
/// assert_eq!(Some("lamb"), iter.next());
///
/// assert_eq!(None, iter.next());
/// ```
#[stable(feature = "split_whitespace", since = "1.1.0")]
#[inline]
pub fn split_whitespace(&self) -> SplitWhitespace {
UnicodeStr::split_whitespace(self)
}
/// An iterator over the lines of a string, as string slices.
///
/// Lines are ended with either a newline (`\n`) or a carriage return with
/// a line feed (`\r\n`).
///
/// The final line ending is optional.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let text = "foo\r\nbar\n\nbaz\n";
/// let mut lines = text.lines();
///
/// assert_eq!(Some("foo"), lines.next());
/// assert_eq!(Some("bar"), lines.next());
/// assert_eq!(Some(""), lines.next());
/// assert_eq!(Some("baz"), lines.next());
///
/// assert_eq!(None, lines.next());
/// ```
///
/// The final line ending isn't required:
///
/// ```
/// let text = "foo\nbar\n\r\nbaz";
/// let mut lines = text.lines();
///
/// assert_eq!(Some("foo"), lines.next());
/// assert_eq!(Some("bar"), lines.next());
/// assert_eq!(Some(""), lines.next());
/// assert_eq!(Some("baz"), lines.next());
///
/// assert_eq!(None, lines.next());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn lines(&self) -> Lines {
core_str::StrExt::lines(self)
}
/// An iterator over the lines of a string.
#[stable(feature = "rust1", since = "1.0.0")]
#[rustc_deprecated(since = "1.4.0", reason = "use lines() instead now")]
#[inline]
#[allow(deprecated)]
pub fn lines_any(&self) -> LinesAny {
core_str::StrExt::lines_any(self)
}
/// Returns an iterator of `u16` over the string encoded as UTF-16.
#[stable(feature = "encode_utf16", since = "1.8.0")]
pub fn encode_utf16(&self) -> EncodeUtf16 {
EncodeUtf16 { encoder: Utf16Encoder::new(self[..].chars()) }
}
/// Returns `true` if the given pattern matches a sub-slice of
/// this string slice.
///
/// Returns `false` if it does not.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let bananas = "bananas";
///
/// assert!(bananas.contains("nana"));
/// assert!(!bananas.contains("apples"));
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn contains<'a, P: Pattern<'a>>(&'a self, pat: P) -> bool {
core_str::StrExt::contains(self, pat)
}
/// Returns `true` if the given pattern matches a prefix of this
/// string slice.
///
/// Returns `false` if it does not.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let bananas = "bananas";
///
/// assert!(bananas.starts_with("bana"));
/// assert!(!bananas.starts_with("nana"));
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn starts_with<'a, P: Pattern<'a>>(&'a self, pat: P) -> bool {
core_str::StrExt::starts_with(self, pat)
}
/// Returns `true` if the given pattern matches a suffix of this
/// string slice.
///
/// Returns `false` if it does not.
///
/// # Examples
///
/// Basic usage:
///
/// ```rust
/// let bananas = "bananas";
///
/// assert!(bananas.ends_with("anas"));
/// assert!(!bananas.ends_with("nana"));
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn ends_with<'a, P: Pattern<'a>>(&'a self, pat: P) -> bool
where P::Searcher: ReverseSearcher<'a>
{
core_str::StrExt::ends_with(self, pat)
}
/// Returns the byte index of the first character of this string slice that
/// matches the pattern.
///
/// Returns [`None`] if the pattern doesn't match.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines if
/// a character matches.
///
/// [`char`]: primitive.char.html
/// [`None`]: option/enum.Option.html#variant.None
///
/// # Examples
///
/// Simple patterns:
///
/// ```
/// let s = "Löwe 老虎 Léopard";
///
/// assert_eq!(s.find('L'), Some(0));
/// assert_eq!(s.find('é'), Some(14));
/// assert_eq!(s.find("Léopard"), Some(13));
/// ```
///
/// More complex patterns with closures:
///
/// ```
/// let s = "Löwe 老虎 Léopard";
///
/// assert_eq!(s.find(char::is_whitespace), Some(5));
/// assert_eq!(s.find(char::is_lowercase), Some(1));
/// ```
///
/// Not finding the pattern:
///
/// ```
/// let s = "Löwe 老虎 Léopard";
/// let x: &[_] = &['1', '2'];
///
/// assert_eq!(s.find(x), None);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn find<'a, P: Pattern<'a>>(&'a self, pat: P) -> Option<usize> {
core_str::StrExt::find(self, pat)
}
/// Returns the byte index of the last character of this string slice that
/// matches the pattern.
///
/// Returns [`None`] if the pattern doesn't match.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines if
/// a character matches.
///
/// [`char`]: primitive.char.html
/// [`None`]: option/enum.Option.html#variant.None
///
/// # Examples
///
/// Simple patterns:
///
/// ```
/// let s = "Löwe 老虎 Léopard";
///
/// assert_eq!(s.rfind('L'), Some(13));
/// assert_eq!(s.rfind('é'), Some(14));
/// ```
///
/// More complex patterns with closures:
///
/// ```
/// let s = "Löwe 老虎 Léopard";
///
/// assert_eq!(s.rfind(char::is_whitespace), Some(12));
/// assert_eq!(s.rfind(char::is_lowercase), Some(20));
/// ```
///
/// Not finding the pattern:
///
/// ```
/// let s = "Löwe 老虎 Léopard";
/// let x: &[_] = &['1', '2'];
///
/// assert_eq!(s.rfind(x), None);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn rfind<'a, P: Pattern<'a>>(&'a self, pat: P) -> Option<usize>
where P::Searcher: ReverseSearcher<'a>
{
core_str::StrExt::rfind(self, pat)
}
/// An iterator over substrings of this string slice, separated by
/// characters matched by a pattern.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines the
/// split.
///
/// # Iterator behavior
///
/// The returned iterator will be a [`DoubleEndedIterator`] if the pattern
/// allows a reverse search and forward/reverse search yields the same
/// elements. This is true for, eg, [`char`] but not for `&str`.
///
/// [`DoubleEndedIterator`]: iter/trait.DoubleEndedIterator.html
///
/// If the pattern allows a reverse search but its results might differ
/// from a forward search, the [`rsplit()`] method can be used.
///
/// [`char`]: primitive.char.html
/// [`rsplit()`]: #method.rsplit
///
/// # Examples
///
/// Simple patterns:
///
/// ```
/// let v: Vec<&str> = "Mary had a little lamb".split(' ').collect();
/// assert_eq!(v, ["Mary", "had", "a", "little", "lamb"]);
///
/// let v: Vec<&str> = "".split('X').collect();
/// assert_eq!(v, [""]);
///
/// let v: Vec<&str> = "lionXXtigerXleopard".split('X').collect();
/// assert_eq!(v, ["lion", "", "tiger", "leopard"]);
///
/// let v: Vec<&str> = "lion::tiger::leopard".split("::").collect();
/// assert_eq!(v, ["lion", "tiger", "leopard"]);
///
/// let v: Vec<&str> = "abc1def2ghi".split(char::is_numeric).collect();
/// assert_eq!(v, ["abc", "def", "ghi"]);
///
/// let v: Vec<&str> = "lionXtigerXleopard".split(char::is_uppercase).collect();
/// assert_eq!(v, ["lion", "tiger", "leopard"]);
/// ```
///
/// A more complex pattern, using a closure:
///
/// ```
/// let v: Vec<&str> = "abc1defXghi".split(|c| c == '1' || c == 'X').collect();
/// assert_eq!(v, ["abc", "def", "ghi"]);
/// ```
///
/// If a string contains multiple contiguous separators, you will end up
/// with empty strings in the output:
///
/// ```
/// let x = "||||a||b|c".to_string();
/// let d: Vec<_> = x.split('|').collect();
///
/// assert_eq!(d, &["", "", "", "", "a", "", "b", "c"]);
/// ```
///
/// Contiguous separators are separated by the empty string.
///
/// ```
/// let x = "(///)".to_string();
/// let d: Vec<_> = x.split('/').collect();;
///
/// assert_eq!(d, &["(", "", "", ")"]);
/// ```
///
/// Separators at the start or end of a string are neighbored
/// by empty strings.
///
/// ```
/// let d: Vec<_> = "010".split("0").collect();
/// assert_eq!(d, &["", "1", ""]);
/// ```
///
/// When the empty string is used as a separator, it separates
/// every character in the string, along with the beginning
/// and end of the string.
///
/// ```
/// let f: Vec<_> = "rust".split("").collect();
/// assert_eq!(f, &["", "r", "u", "s", "t", ""]);
/// ```
///
/// Contiguous separators can lead to possibly surprising behavior
/// when whitespace is used as the separator. This code is correct:
///
/// ```
/// let x = " a b c".to_string();
/// let d: Vec<_> = x.split(' ').collect();
///
/// assert_eq!(d, &["", "", "", "", "a", "", "b", "c"]);
/// ```
///
/// It does _not_ give you:
///
/// ```rust,ignore
/// assert_eq!(d, &["a", "b", "c"]);
/// ```
///
/// Use [`split_whitespace()`] for this behavior.
///
/// [`split_whitespace()`]: #method.split_whitespace
#[stable(feature = "rust1", since = "1.0.0")]
pub fn split<'a, P: Pattern<'a>>(&'a self, pat: P) -> Split<'a, P> {
core_str::StrExt::split(self, pat)
}
/// An iterator over substrings of the given string slice, separated by
/// characters matched by a pattern and yielded in reverse order.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines the
/// split.
///
/// [`char`]: primitive.char.html
///
/// # Iterator behavior
///
/// The returned iterator requires that the pattern supports a reverse
/// search, and it will be a [`DoubleEndedIterator`] if a forward/reverse
/// search yields the same elements.
///
/// [`DoubleEndedIterator`]: iter/trait.DoubleEndedIterator.html
///
/// For iterating from the front, the [`split()`] method can be used.
///
/// [`split()`]: #method.split
///
/// # Examples
///
/// Simple patterns:
///
/// ```
/// let v: Vec<&str> = "Mary had a little lamb".rsplit(' ').collect();
/// assert_eq!(v, ["lamb", "little", "a", "had", "Mary"]);
///
/// let v: Vec<&str> = "".rsplit('X').collect();
/// assert_eq!(v, [""]);
///
/// let v: Vec<&str> = "lionXXtigerXleopard".rsplit('X').collect();
/// assert_eq!(v, ["leopard", "tiger", "", "lion"]);
///
/// let v: Vec<&str> = "lion::tiger::leopard".rsplit("::").collect();
/// assert_eq!(v, ["leopard", "tiger", "lion"]);
/// ```
///
/// A more complex pattern, using a closure:
///
/// ```
/// let v: Vec<&str> = "abc1defXghi".rsplit(|c| c == '1' || c == 'X').collect();
/// assert_eq!(v, ["ghi", "def", "abc"]);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn rsplit<'a, P: Pattern<'a>>(&'a self, pat: P) -> RSplit<'a, P>
where P::Searcher: ReverseSearcher<'a>
{
core_str::StrExt::rsplit(self, pat)
}
/// An iterator over substrings of the given string slice, separated by
/// characters matched by a pattern.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines the
/// split.
///
/// Equivalent to [`split()`], except that the trailing substring
/// is skipped if empty.
///
/// [`split()`]: #method.split
///
/// This method can be used for string data that is _terminated_,
/// rather than _separated_ by a pattern.
///
/// # Iterator behavior
///
/// The returned iterator will be a [`DoubleEndedIterator`] if the pattern
/// allows a reverse search and forward/reverse search yields the same
/// elements. This is true for, eg, [`char`] but not for `&str`.
///
/// [`DoubleEndedIterator`]: iter/trait.DoubleEndedIterator.html
/// [`char`]: primitive.char.html
///
/// If the pattern allows a reverse search but its results might differ
/// from a forward search, the [`rsplit_terminator()`] method can be used.
///
/// [`rsplit_terminator()`]: #method.rsplit_terminator
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let v: Vec<&str> = "A.B.".split_terminator('.').collect();
/// assert_eq!(v, ["A", "B"]);
///
/// let v: Vec<&str> = "A..B..".split_terminator(".").collect();
/// assert_eq!(v, ["A", "", "B", ""]);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn split_terminator<'a, P: Pattern<'a>>(&'a self, pat: P) -> SplitTerminator<'a, P> {
core_str::StrExt::split_terminator(self, pat)
}
/// An iterator over substrings of `self`, separated by characters
/// matched by a pattern and yielded in reverse order.
///
/// The pattern can be a simple `&str`, [`char`], or a closure that
/// determines the split.
/// Additional libraries might provide more complex patterns like
/// regular expressions.
///
/// [`char`]: primitive.char.html
///
/// Equivalent to [`split()`], except that the trailing substring is
/// skipped if empty.
///
/// [`split()`]: #method.split
///
/// This method can be used for string data that is _terminated_,
/// rather than _separated_ by a pattern.
///
/// # Iterator behavior
///
/// The returned iterator requires that the pattern supports a
/// reverse search, and it will be double ended if a forward/reverse
/// search yields the same elements.
///
/// For iterating from the front, the [`split_terminator()`] method can be
/// used.
///
/// [`split_terminator()`]: #method.split_terminator
///
/// # Examples
///
/// ```
/// let v: Vec<&str> = "A.B.".rsplit_terminator('.').collect();
/// assert_eq!(v, ["B", "A"]);
///
/// let v: Vec<&str> = "A..B..".rsplit_terminator(".").collect();
/// assert_eq!(v, ["", "B", "", "A"]);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn rsplit_terminator<'a, P: Pattern<'a>>(&'a self, pat: P) -> RSplitTerminator<'a, P>
where P::Searcher: ReverseSearcher<'a>
{
core_str::StrExt::rsplit_terminator(self, pat)
}
/// An iterator over substrings of the given string slice, separated by a
/// pattern, restricted to returning at most `count` items.
///
/// The last element returned, if any, will contain the remainder of the
/// string slice.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines the
/// split.
///
/// [`char`]: primitive.char.html
///
/// # Iterator behavior
///
/// The returned iterator will not be double ended, because it is
/// not efficient to support.
///
/// If the pattern allows a reverse search, the [`rsplitn()`] method can be
/// used.
///
/// [`rsplitn()`]: #method.rsplitn
///
/// # Examples
///
/// Simple patterns:
///
/// ```
/// let v: Vec<&str> = "Mary had a little lambda".splitn(3, ' ').collect();
/// assert_eq!(v, ["Mary", "had", "a little lambda"]);
///
/// let v: Vec<&str> = "lionXXtigerXleopard".splitn(3, "X").collect();
/// assert_eq!(v, ["lion", "", "tigerXleopard"]);
///
/// let v: Vec<&str> = "abcXdef".splitn(1, 'X').collect();
/// assert_eq!(v, ["abcXdef"]);
///
/// let v: Vec<&str> = "".splitn(1, 'X').collect();
/// assert_eq!(v, [""]);
/// ```
///
/// A more complex pattern, using a closure:
///
/// ```
/// let v: Vec<&str> = "abc1defXghi".splitn(2, |c| c == '1' || c == 'X').collect();
/// assert_eq!(v, ["abc", "defXghi"]);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn splitn<'a, P: Pattern<'a>>(&'a self, count: usize, pat: P) -> SplitN<'a, P> {
core_str::StrExt::splitn(self, count, pat)
}
/// An iterator over substrings of this string slice, separated by a
/// pattern, starting from the end of the string, restricted to returning
/// at most `count` items.
///
/// The last element returned, if any, will contain the remainder of the
/// string slice.
///
/// The pattern can be a `&str`, [`char`], or a closure that
/// determines the split.
///
/// [`char`]: primitive.char.html
///
/// # Iterator behavior
///
/// The returned iterator will not be double ended, because it is not
/// efficient to support.
///
/// For splitting from the front, the [`splitn()`] method can be used.
///
/// [`splitn()`]: #method.splitn
///
/// # Examples
///
/// Simple patterns:
///
/// ```
/// let v: Vec<&str> = "Mary had a little lamb".rsplitn(3, ' ').collect();
/// assert_eq!(v, ["lamb", "little", "Mary had a"]);
///
/// let v: Vec<&str> = "lionXXtigerXleopard".rsplitn(3, 'X').collect();
/// assert_eq!(v, ["leopard", "tiger", "lionX"]);
///
/// let v: Vec<&str> = "lion::tiger::leopard".rsplitn(2, "::").collect();
/// assert_eq!(v, ["leopard", "lion::tiger"]);
/// ```
///
/// A more complex pattern, using a closure:
///
/// ```
/// let v: Vec<&str> = "abc1defXghi".rsplitn(2, |c| c == '1' || c == 'X').collect();
/// assert_eq!(v, ["ghi", "abc1def"]);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn rsplitn<'a, P: Pattern<'a>>(&'a self, count: usize, pat: P) -> RSplitN<'a, P>
where P::Searcher: ReverseSearcher<'a>
{
core_str::StrExt::rsplitn(self, count, pat)
}
/// An iterator over the matches of a pattern within the given string
/// slice.
///
/// The pattern can be a `&str`, [`char`], or a closure that
/// determines if a character matches.
///
/// [`char`]: primitive.char.html
///
/// # Iterator behavior
///
/// The returned iterator will be a [`DoubleEndedIterator`] if the pattern
/// allows a reverse search and forward/reverse search yields the same
/// elements. This is true for, eg, [`char`] but not for `&str`.
///
/// [`DoubleEndedIterator`]: iter/trait.DoubleEndedIterator.html
/// [`char`]: primitive.char.html
///
/// If the pattern allows a reverse search but its results might differ
/// from a forward search, the [`rmatches()`] method can be used.
///
/// [`rmatches()`]: #method.rmatches
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let v: Vec<&str> = "abcXXXabcYYYabc".matches("abc").collect();
/// assert_eq!(v, ["abc", "abc", "abc"]);
///
/// let v: Vec<&str> = "1abc2abc3".matches(char::is_numeric).collect();
/// assert_eq!(v, ["1", "2", "3"]);
/// ```
#[stable(feature = "str_matches", since = "1.2.0")]
pub fn matches<'a, P: Pattern<'a>>(&'a self, pat: P) -> Matches<'a, P> {
core_str::StrExt::matches(self, pat)
}
/// An iterator over the matches of a pattern within this string slice,
/// yielded in reverse order.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines if
/// a character matches.
///
/// [`char`]: primitive.char.html
///
/// # Iterator behavior
///
/// The returned iterator requires that the pattern supports a reverse
/// search, and it will be a [`DoubleEndedIterator`] if a forward/reverse
/// search yields the same elements.
///
/// [`DoubleEndedIterator`]: iter/trait.DoubleEndedIterator.html
///
/// For iterating from the front, the [`matches()`] method can be used.
///
/// [`matches()`]: #method.matches
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let v: Vec<&str> = "abcXXXabcYYYabc".rmatches("abc").collect();
/// assert_eq!(v, ["abc", "abc", "abc"]);
///
/// let v: Vec<&str> = "1abc2abc3".rmatches(char::is_numeric).collect();
/// assert_eq!(v, ["3", "2", "1"]);
/// ```
#[stable(feature = "str_matches", since = "1.2.0")]
pub fn rmatches<'a, P: Pattern<'a>>(&'a self, pat: P) -> RMatches<'a, P>
where P::Searcher: ReverseSearcher<'a>
{
core_str::StrExt::rmatches(self, pat)
}
/// An iterator over the disjoint matches of a pattern within this string
/// slice as well as the index that the match starts at.
///
/// For matches of `pat` within `self` that overlap, only the indices
/// corresponding to the first match are returned.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines
/// if a character matches.
///
/// [`char`]: primitive.char.html
///
/// # Iterator behavior
///
/// The returned iterator will be a [`DoubleEndedIterator`] if the pattern
/// allows a reverse search and forward/reverse search yields the same
/// elements. This is true for, eg, [`char`] but not for `&str`.
///
/// [`DoubleEndedIterator`]: iter/trait.DoubleEndedIterator.html
///
/// If the pattern allows a reverse search but its results might differ
/// from a forward search, the [`rmatch_indices()`] method can be used.
///
/// [`rmatch_indices()`]: #method.rmatch_indices
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let v: Vec<_> = "abcXXXabcYYYabc".match_indices("abc").collect();
/// assert_eq!(v, [(0, "abc"), (6, "abc"), (12, "abc")]);
///
/// let v: Vec<_> = "1abcabc2".match_indices("abc").collect();
/// assert_eq!(v, [(1, "abc"), (4, "abc")]);
///
/// let v: Vec<_> = "ababa".match_indices("aba").collect();
/// assert_eq!(v, [(0, "aba")]); // only the first `aba`
/// ```
#[stable(feature = "str_match_indices", since = "1.5.0")]
pub fn match_indices<'a, P: Pattern<'a>>(&'a self, pat: P) -> MatchIndices<'a, P> {
core_str::StrExt::match_indices(self, pat)
}
/// An iterator over the disjoint matches of a pattern within `self`,
/// yielded in reverse order along with the index of the match.
///
/// For matches of `pat` within `self` that overlap, only the indices
/// corresponding to the last match are returned.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines if a
/// character matches.
///
/// [`char`]: primitive.char.html
///
/// # Iterator behavior
///
/// The returned iterator requires that the pattern supports a reverse
/// search, and it will be a [`DoubleEndedIterator`] if a forward/reverse
/// search yields the same elements.
///
/// [`DoubleEndedIterator`]: iter/trait.DoubleEndedIterator.html
///
/// For iterating from the front, the [`match_indices()`] method can be used.
///
/// [`match_indices()`]: #method.match_indices
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let v: Vec<_> = "abcXXXabcYYYabc".rmatch_indices("abc").collect();
/// assert_eq!(v, [(12, "abc"), (6, "abc"), (0, "abc")]);
///
/// let v: Vec<_> = "1abcabc2".rmatch_indices("abc").collect();
/// assert_eq!(v, [(4, "abc"), (1, "abc")]);
///
/// let v: Vec<_> = "ababa".rmatch_indices("aba").collect();
/// assert_eq!(v, [(2, "aba")]); // only the last `aba`
/// ```
#[stable(feature = "str_match_indices", since = "1.5.0")]
pub fn rmatch_indices<'a, P: Pattern<'a>>(&'a self, pat: P) -> RMatchIndices<'a, P>
where P::Searcher: ReverseSearcher<'a>
{
core_str::StrExt::rmatch_indices(self, pat)
}
/// Returns a string slice with leading and trailing whitespace removed.
///
/// 'Whitespace' is defined according to the terms of the Unicode Derived
/// Core Property `White_Space`.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = " Hello\tworld\t";
///
/// assert_eq!("Hello\tworld", s.trim());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn trim(&self) -> &str {
UnicodeStr::trim(self)
}
/// Returns a string slice with leading whitespace removed.
///
/// 'Whitespace' is defined according to the terms of the Unicode Derived
/// Core Property `White_Space`.
///
/// # Text directionality
///
/// A string is a sequence of bytes. 'Left' in this context means the first
/// position of that byte string; for a language like Arabic or Hebrew
/// which are 'right to left' rather than 'left to right', this will be
/// the _right_ side, not the left.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = " Hello\tworld\t";
///
/// assert_eq!("Hello\tworld\t", s.trim_left());
/// ```
///
/// Directionality:
///
/// ```
/// let s = " English";
/// assert!(Some('E') == s.trim_left().chars().next());
///
/// let s = " עברית";
/// assert!(Some('ע') == s.trim_left().chars().next());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn trim_left(&self) -> &str {
UnicodeStr::trim_left(self)
}
/// Returns a string slice with trailing whitespace removed.
///
/// 'Whitespace' is defined according to the terms of the Unicode Derived
/// Core Property `White_Space`.
///
/// # Text directionality
///
/// A string is a sequence of bytes. 'Right' in this context means the last
/// position of that byte string; for a language like Arabic or Hebrew
/// which are 'right to left' rather than 'left to right', this will be
/// the _left_ side, not the right.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = " Hello\tworld\t";
///
/// assert_eq!(" Hello\tworld", s.trim_right());
/// ```
///
/// Directionality:
///
/// ```
/// let s = "English ";
/// assert!(Some('h') == s.trim_right().chars().rev().next());
///
/// let s = "עברית ";
/// assert!(Some('ת') == s.trim_right().chars().rev().next());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn trim_right(&self) -> &str {
UnicodeStr::trim_right(self)
}
/// Returns a string slice with all prefixes and suffixes that match a
/// pattern repeatedly removed.
///
/// The pattern can be a [`char`] or a closure that determines if a
/// character matches.
///
/// [`char`]: primitive.char.html
///
/// # Examples
///
/// Simple patterns:
///
/// ```
/// assert_eq!("11foo1bar11".trim_matches('1'), "foo1bar");
/// assert_eq!("123foo1bar123".trim_matches(char::is_numeric), "foo1bar");
///
/// let x: &[_] = &['1', '2'];
/// assert_eq!("12foo1bar12".trim_matches(x), "foo1bar");
/// ```
///
/// A more complex pattern, using a closure:
///
/// ```
/// assert_eq!("1foo1barXX".trim_matches(|c| c == '1' || c == 'X'), "foo1bar");
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn trim_matches<'a, P: Pattern<'a>>(&'a self, pat: P) -> &'a str
where P::Searcher: DoubleEndedSearcher<'a>
{
core_str::StrExt::trim_matches(self, pat)
}
/// Returns a string slice with all prefixes that match a pattern
/// repeatedly removed.
///
/// The pattern can be a `&str`, [`char`], or a closure that determines if
/// a character matches.
///
/// [`char`]: primitive.char.html
///
/// # Text directionality
///
/// A string is a sequence of bytes. 'Left' in this context means the first
/// position of that byte string; for a language like Arabic or Hebrew
/// which are 'right to left' rather than 'left to right', this will be
/// the _right_ side, not the left.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert_eq!("11foo1bar11".trim_left_matches('1'), "foo1bar11");
/// assert_eq!("123foo1bar123".trim_left_matches(char::is_numeric), "foo1bar123");
///
/// let x: &[_] = &['1', '2'];
/// assert_eq!("12foo1bar12".trim_left_matches(x), "foo1bar12");
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn trim_left_matches<'a, P: Pattern<'a>>(&'a self, pat: P) -> &'a str {
core_str::StrExt::trim_left_matches(self, pat)
}
/// Returns a string slice with all suffixes that match a pattern
/// repeatedly removed.
///
/// The pattern can be a `&str`, [`char`], or a closure that
/// determines if a character matches.
///
/// [`char`]: primitive.char.html
///
/// # Text directionality
///
/// A string is a sequence of bytes. 'Right' in this context means the last
/// position of that byte string; for a language like Arabic or Hebrew
/// which are 'right to left' rather than 'left to right', this will be
/// the _left_ side, not the right.
///
/// # Examples
///
/// Simple patterns:
///
/// ```
/// assert_eq!("11foo1bar11".trim_right_matches('1'), "11foo1bar");
/// assert_eq!("123foo1bar123".trim_right_matches(char::is_numeric), "123foo1bar");
///
/// let x: &[_] = &['1', '2'];
/// assert_eq!("12foo1bar12".trim_right_matches(x), "12foo1bar");
/// ```
///
/// A more complex pattern, using a closure:
///
/// ```
/// assert_eq!("1fooX".trim_left_matches(|c| c == '1' || c == 'X'), "fooX");
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn trim_right_matches<'a, P: Pattern<'a>>(&'a self, pat: P) -> &'a str
where P::Searcher: ReverseSearcher<'a>
{
core_str::StrExt::trim_right_matches(self, pat)
}
/// Parses this string slice into another type.
///
/// Because `parse()` is so general, it can cause problems with type
/// inference. As such, `parse()` is one of the few times you'll see
/// the syntax affectionately known as the 'turbofish': `::<>`. This
/// helps the inference algorithm understand specifically which type
/// you're trying to parse into.
///
/// `parse()` can parse any type that implements the [`FromStr`] trait.
///
/// [`FromStr`]: str/trait.FromStr.html
///
/// # Errors
///
/// Will return [`Err`] if it's not possible to parse this string slice into
/// the desired type.
///
/// [`Err`]: str/trait.FromStr.html#associatedtype.Err
///
/// # Example
///
/// Basic usage
///
/// ```
/// let four: u32 = "4".parse().unwrap();
///
/// assert_eq!(4, four);
/// ```
///
/// Using the 'turbofish' instead of annotating `four`:
///
/// ```
/// let four = "4".parse::<u32>();
///
/// assert_eq!(Ok(4), four);
/// ```
///
/// Failing to parse:
///
/// ```
/// let nope = "j".parse::<u32>();
///
/// assert!(nope.is_err());
/// ```
#[inline]
#[stable(feature = "rust1", since = "1.0.0")]
pub fn parse<F: FromStr>(&self) -> Result<F, F::Err> {
core_str::StrExt::parse(self)
}
/// Replaces all matches of a pattern with another string.
///
/// `replace` creates a new [`String`], and copies the data from this string slice into it.
/// While doing so, it attempts to find matches of a pattern. If it finds any, it
/// replaces them with the replacement string slice.
///
/// [`String`]: string/struct.String.html
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = "this is old";
///
/// assert_eq!("this is new", s.replace("old", "new"));
/// ```
///
/// When the pattern doesn't match:
///
/// ```
/// let s = "this is old";
/// assert_eq!(s, s.replace("cookie monster", "little lamb"));
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
pub fn replace<'a, P: Pattern<'a>>(&'a self, from: P, to: &str) -> String {
let mut result = String::new();
let mut last_end = 0;
for (start, part) in self.match_indices(from) {
result.push_str(unsafe { self.slice_unchecked(last_end, start) });
result.push_str(to);
last_end = start + part.len();
}
result.push_str(unsafe { self.slice_unchecked(last_end, self.len()) });
result
}
/// Returns the lowercase equivalent of this string slice, as a new [`String`].
///
/// 'Lowercase' is defined according to the terms of the Unicode Derived Core Property
/// `Lowercase`.
///
/// [`String`]: string/struct.String.html
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = "HELLO";
///
/// assert_eq!("hello", s.to_lowercase());
/// ```
///
/// A tricky example, with sigma:
///
/// ```
/// let sigma = "Σ";
///
/// assert_eq!("σ", sigma.to_lowercase());
///
/// // but at the end of a word, it's ς, not σ:
/// let odysseus = "ὈΔΥΣΣΕΎΣ";
///
/// assert_eq!("ὀδυσσεύς", odysseus.to_lowercase());
/// ```
///
/// Languages without case are not changed:
///
/// ```
/// let new_year = "农历新年";
///
/// assert_eq!(new_year, new_year.to_lowercase());
/// ```
#[stable(feature = "unicode_case_mapping", since = "1.2.0")]
pub fn to_lowercase(&self) -> String {
let mut s = String::with_capacity(self.len());
for (i, c) in self[..].char_indices() {
if c == 'Σ' {
// Σ maps to σ, except at the end of a word where it maps to ς.
// This is the only conditional (contextual) but language-independent mapping
// in `SpecialCasing.txt`,
// so hard-code it rather than have a generic "condition" mechanism.
// See https://github.com/rust-lang/rust/issues/26035
map_uppercase_sigma(self, i, &mut s)
} else {
s.extend(c.to_lowercase());
}
}
return s;
fn map_uppercase_sigma(from: &str, i: usize, to: &mut String) {
// See http://www.unicode.org/versions/Unicode7.0.0/ch03.pdf#G33992
// for the definition of `Final_Sigma`.
debug_assert!('Σ'.len_utf8() == 2);
let is_word_final = case_ignoreable_then_cased(from[..i].chars().rev()) &&
!case_ignoreable_then_cased(from[i + 2..].chars());
to.push_str(if is_word_final {
"ς"
} else {
"σ"
});
}
fn case_ignoreable_then_cased<I: Iterator<Item = char>>(iter: I) -> bool {
use rustc_unicode::derived_property::{Cased, Case_Ignorable};
match iter.skip_while(|&c| Case_Ignorable(c)).next() {
Some(c) => Cased(c),
None => false,
}
}
}
/// Returns the uppercase equivalent of this string slice, as a new [`String`].
///
/// 'Uppercase' is defined according to the terms of the Unicode Derived Core Property
/// `Uppercase`.
///
/// [`String`]: string/struct.String.html
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let s = "hello";
///
/// assert_eq!("HELLO", s.to_uppercase());
/// ```
///
/// Scripts without case are not changed:
///
/// ```
/// let new_year = "农历新年";
///
/// assert_eq!(new_year, new_year.to_uppercase());
/// ```
#[stable(feature = "unicode_case_mapping", since = "1.2.0")]
pub fn to_uppercase(&self) -> String {
let mut s = String::with_capacity(self.len());
s.extend(self.chars().flat_map(|c| c.to_uppercase()));
return s;
}
/// Escapes each char in `s` with `char::escape_debug`.
#[unstable(feature = "str_escape",
reason = "return type may change to be an iterator",
issue = "27791")]
pub fn escape_debug(&self) -> String {
self.chars().flat_map(|c| c.escape_debug()).collect()
}
/// Escapes each char in `s` with `char::escape_default`.
#[unstable(feature = "str_escape",
reason = "return type may change to be an iterator",
issue = "27791")]
pub fn escape_default(&self) -> String {
self.chars().flat_map(|c| c.escape_default()).collect()
}
/// Escapes each char in `s` with `char::escape_unicode`.
#[unstable(feature = "str_escape",
reason = "return type may change to be an iterator",
issue = "27791")]
pub fn escape_unicode(&self) -> String {
self.chars().flat_map(|c| c.escape_unicode()).collect()
}
/// Converts a `Box<str>` into a [`String`] without copying or allocating.
///
/// [`String`]: string/struct.String.html
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let string = String::from("birthday gift");
/// let boxed_str = string.clone().into_boxed_str();
///
/// assert_eq!(boxed_str.into_string(), string);
/// ```
#[stable(feature = "box_str", since = "1.4.0")]
pub fn into_string(self: Box<str>) -> String {
unsafe {
let slice = mem::transmute::<Box<str>, Box<[u8]>>(self);
String::from_utf8_unchecked(slice.into_vec())
}
}
}