2017-03-25 15:58:35 -07:00

1070 lines
31 KiB
Rust
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

// 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.
//! A character type.
//!
//! The `char` type represents a single character. More specifically, since
//! 'character' isn't a well-defined concept in Unicode, `char` is a '[Unicode
//! scalar value]', which is similar to, but not the same as, a '[Unicode code
//! point]'.
//!
//! [Unicode scalar value]: http://www.unicode.org/glossary/#unicode_scalar_value
//! [Unicode code point]: http://www.unicode.org/glossary/#code_point
//!
//! This module exists for technical reasons, the primary documentation for
//! `char` is directly on [the `char` primitive type](../../std/primitive.char.html)
//! itself.
//!
//! This module is the home of the iterator implementations for the iterators
//! implemented on `char`, as well as some useful constants and conversion
//! functions that convert various types to `char`.
#![stable(feature = "rust1", since = "1.0.0")]
use core::char::CharExt as C;
use core::iter::FusedIterator;
use core::fmt::{self, Write};
use tables::{conversions, derived_property, general_category, property};
// stable reexports
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::char::{MAX, from_digit, from_u32, from_u32_unchecked};
#[stable(feature = "rust1", since = "1.0.0")]
pub use core::char::{EscapeDebug, EscapeDefault, EscapeUnicode};
// unstable reexports
#[unstable(feature = "try_from", issue = "33417")]
pub use core::char::CharTryFromError;
#[unstable(feature = "decode_utf8", issue = "33906")]
pub use core::char::{DecodeUtf8, decode_utf8};
#[unstable(feature = "unicode", issue = "27783")]
pub use tables::UNICODE_VERSION;
/// Returns an iterator that yields the lowercase equivalent of a `char`.
///
/// This `struct` is created by the [`to_lowercase`] method on [`char`]. See
/// its documentation for more.
///
/// [`to_lowercase`]: ../../std/primitive.char.html#method.to_lowercase
/// [`char`]: ../../std/primitive.char.html
#[stable(feature = "rust1", since = "1.0.0")]
pub struct ToLowercase(CaseMappingIter);
#[stable(feature = "rust1", since = "1.0.0")]
impl Iterator for ToLowercase {
type Item = char;
fn next(&mut self) -> Option<char> {
self.0.next()
}
}
#[unstable(feature = "fused", issue = "35602")]
impl FusedIterator for ToLowercase {}
/// Returns an iterator that yields the uppercase equivalent of a `char`.
///
/// This `struct` is created by the [`to_uppercase`] method on [`char`]. See
/// its documentation for more.
///
/// [`to_uppercase`]: ../../std/primitive.char.html#method.to_uppercase
/// [`char`]: ../../std/primitive.char.html
#[stable(feature = "rust1", since = "1.0.0")]
pub struct ToUppercase(CaseMappingIter);
#[stable(feature = "rust1", since = "1.0.0")]
impl Iterator for ToUppercase {
type Item = char;
fn next(&mut self) -> Option<char> {
self.0.next()
}
}
#[unstable(feature = "fused", issue = "35602")]
impl FusedIterator for ToUppercase {}
enum CaseMappingIter {
Three(char, char, char),
Two(char, char),
One(char),
Zero,
}
impl CaseMappingIter {
fn new(chars: [char; 3]) -> CaseMappingIter {
if chars[2] == '\0' {
if chars[1] == '\0' {
CaseMappingIter::One(chars[0]) // Including if chars[0] == '\0'
} else {
CaseMappingIter::Two(chars[0], chars[1])
}
} else {
CaseMappingIter::Three(chars[0], chars[1], chars[2])
}
}
}
impl Iterator for CaseMappingIter {
type Item = char;
fn next(&mut self) -> Option<char> {
match *self {
CaseMappingIter::Three(a, b, c) => {
*self = CaseMappingIter::Two(b, c);
Some(a)
}
CaseMappingIter::Two(b, c) => {
*self = CaseMappingIter::One(c);
Some(b)
}
CaseMappingIter::One(c) => {
*self = CaseMappingIter::Zero;
Some(c)
}
CaseMappingIter::Zero => None,
}
}
}
impl fmt::Display for CaseMappingIter {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match *self {
CaseMappingIter::Three(a, b, c) => {
f.write_char(a)?;
f.write_char(b)?;
f.write_char(c)
}
CaseMappingIter::Two(b, c) => {
f.write_char(b)?;
f.write_char(c)
}
CaseMappingIter::One(c) => {
f.write_char(c)
}
CaseMappingIter::Zero => Ok(()),
}
}
}
#[stable(feature = "char_struct_display", since = "1.16.0")]
impl fmt::Display for ToLowercase {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
fmt::Display::fmt(&self.0, f)
}
}
#[stable(feature = "char_struct_display", since = "1.16.0")]
impl fmt::Display for ToUppercase {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
fmt::Display::fmt(&self.0, f)
}
}
#[lang = "char"]
impl char {
/// Checks if a `char` is a digit in the given radix.
///
/// A 'radix' here is sometimes also called a 'base'. A radix of two
/// indicates a binary number, a radix of ten, decimal, and a radix of
/// sixteen, hexadecimal, to give some common values. Arbitrary
/// radices are supported.
///
/// Compared to `is_numeric()`, this function only recognizes the characters
/// `0-9`, `a-z` and `A-Z`.
///
/// 'Digit' is defined to be only the following characters:
///
/// * `0-9`
/// * `a-z`
/// * `A-Z`
///
/// For a more comprehensive understanding of 'digit', see [`is_numeric`][is_numeric].
///
/// [is_numeric]: #method.is_numeric
///
/// # Panics
///
/// Panics if given a radix larger than 36.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert!('1'.is_digit(10));
/// assert!('f'.is_digit(16));
/// assert!(!'f'.is_digit(10));
/// ```
///
/// Passing a large radix, causing a panic:
///
/// ```
/// use std::thread;
///
/// let result = thread::spawn(|| {
/// // this panics
/// '1'.is_digit(37);
/// }).join();
///
/// assert!(result.is_err());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn is_digit(self, radix: u32) -> bool {
C::is_digit(self, radix)
}
/// Converts a `char` to a digit in the given radix.
///
/// A 'radix' here is sometimes also called a 'base'. A radix of two
/// indicates a binary number, a radix of ten, decimal, and a radix of
/// sixteen, hexadecimal, to give some common values. Arbitrary
/// radices are supported.
///
/// 'Digit' is defined to be only the following characters:
///
/// * `0-9`
/// * `a-z`
/// * `A-Z`
///
/// # Errors
///
/// Returns `None` if the `char` does not refer to a digit in the given radix.
///
/// # Panics
///
/// Panics if given a radix larger than 36.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert_eq!('1'.to_digit(10), Some(1));
/// assert_eq!('f'.to_digit(16), Some(15));
/// ```
///
/// Passing a non-digit results in failure:
///
/// ```
/// assert_eq!('f'.to_digit(10), None);
/// assert_eq!('z'.to_digit(16), None);
/// ```
///
/// Passing a large radix, causing a panic:
///
/// ```
/// use std::thread;
///
/// let result = thread::spawn(|| {
/// '1'.to_digit(37);
/// }).join();
///
/// assert!(result.is_err());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn to_digit(self, radix: u32) -> Option<u32> {
C::to_digit(self, radix)
}
/// Returns an iterator that yields the hexadecimal Unicode escape of a
/// character as `char`s.
///
/// This will escape characters with the Rust syntax of the form
/// `\u{NNNNNN}` where `NNNNNN` is a hexadecimal representation.
///
/// # Examples
///
/// As an iterator:
///
/// ```
/// for c in '❤'.escape_unicode() {
/// print!("{}", c);
/// }
/// println!();
/// ```
///
/// Using `println!` directly:
///
/// ```
/// println!("{}", '❤'.escape_unicode());
/// ```
///
/// Both are equivalent to:
///
/// ```
/// println!("\\u{{2764}}");
/// ```
///
/// Using `to_string`:
///
/// ```
/// assert_eq!('❤'.escape_unicode().to_string(), "\\u{2764}");
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn escape_unicode(self) -> EscapeUnicode {
C::escape_unicode(self)
}
/// Returns an iterator that yields the literal escape code of a character
/// as `char`s.
///
/// This will escape the characters similar to the `Debug` implementations
/// of `str` or `char`.
///
/// # Examples
///
/// As an iterator:
///
/// ```
/// # #![feature(char_escape_debug)]
/// for c in '\n'.escape_debug() {
/// print!("{}", c);
/// }
/// println!();
/// ```
///
/// Using `println!` directly:
///
/// ```
/// # #![feature(char_escape_debug)]
/// println!("{}", '\n'.escape_debug());
/// ```
///
/// Both are equivalent to:
///
/// ```
/// println!("\\n");
/// ```
///
/// Using `to_string`:
///
/// ```
/// # #![feature(char_escape_debug)]
/// assert_eq!('\n'.escape_debug().to_string(), "\\n");
/// ```
#[unstable(feature = "char_escape_debug", issue = "35068")]
#[inline]
pub fn escape_debug(self) -> EscapeDebug {
C::escape_debug(self)
}
/// Returns an iterator that yields the literal escape code of a character
/// as `char`s.
///
/// The default is chosen with a bias toward producing literals that are
/// legal in a variety of languages, including C++11 and similar C-family
/// languages. The exact rules are:
///
/// * Tab is escaped as `\t`.
/// * Carriage return is escaped as `\r`.
/// * Line feed is escaped as `\n`.
/// * Single quote is escaped as `\'`.
/// * Double quote is escaped as `\"`.
/// * Backslash is escaped as `\\`.
/// * Any character in the 'printable ASCII' range `0x20` .. `0x7e`
/// inclusive is not escaped.
/// * All other characters are given hexadecimal Unicode escapes; see
/// [`escape_unicode`][escape_unicode].
///
/// [escape_unicode]: #method.escape_unicode
///
/// # Examples
///
/// As an iterator:
///
/// ```
/// for c in '"'.escape_default() {
/// print!("{}", c);
/// }
/// println!();
/// ```
///
/// Using `println!` directly:
///
/// ```
/// println!("{}", '"'.escape_default());
/// ```
///
///
/// Both are equivalent to:
///
/// ```
/// println!("\\\"");
/// ```
///
/// Using `to_string`:
///
/// ```
/// assert_eq!('"'.escape_default().to_string(), "\\\"");
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn escape_default(self) -> EscapeDefault {
C::escape_default(self)
}
/// Returns the number of bytes this `char` would need if encoded in UTF-8.
///
/// That number of bytes is always between 1 and 4, inclusive.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let len = 'A'.len_utf8();
/// assert_eq!(len, 1);
///
/// let len = 'ß'.len_utf8();
/// assert_eq!(len, 2);
///
/// let len = ''.len_utf8();
/// assert_eq!(len, 3);
///
/// let len = '💣'.len_utf8();
/// assert_eq!(len, 4);
/// ```
///
/// The `&str` type guarantees that its contents are UTF-8, and so we can compare the length it
/// would take if each code point was represented as a `char` vs in the `&str` itself:
///
/// ```
/// // as chars
/// let eastern = '東';
/// let capitol = '京';
///
/// // both can be represented as three bytes
/// assert_eq!(3, eastern.len_utf8());
/// assert_eq!(3, capitol.len_utf8());
///
/// // as a &str, these two are encoded in UTF-8
/// let tokyo = "東京";
///
/// let len = eastern.len_utf8() + capitol.len_utf8();
///
/// // we can see that they take six bytes total...
/// assert_eq!(6, tokyo.len());
///
/// // ... just like the &str
/// assert_eq!(len, tokyo.len());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn len_utf8(self) -> usize {
C::len_utf8(self)
}
/// Returns the number of 16-bit code units this `char` would need if
/// encoded in UTF-16.
///
/// See the documentation for [`len_utf8`] for more explanation of this
/// concept. This function is a mirror, but for UTF-16 instead of UTF-8.
///
/// [`len_utf8`]: #method.len_utf8
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// let n = 'ß'.len_utf16();
/// assert_eq!(n, 1);
///
/// let len = '💣'.len_utf16();
/// assert_eq!(len, 2);
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn len_utf16(self) -> usize {
C::len_utf16(self)
}
/// Encodes this character as UTF-8 into the provided byte buffer,
/// and then returns the subslice of the buffer that contains the encoded character.
///
/// # Panics
///
/// Panics if the buffer is not large enough.
/// A buffer of length four is large enough to encode any `char`.
///
/// # Examples
///
/// In both of these examples, 'ß' takes two bytes to encode.
///
/// ```
/// let mut b = [0; 2];
///
/// let result = 'ß'.encode_utf8(&mut b);
///
/// assert_eq!(result, "ß");
///
/// assert_eq!(result.len(), 2);
/// ```
///
/// A buffer that's too small:
///
/// ```
/// use std::thread;
///
/// let result = thread::spawn(|| {
/// let mut b = [0; 1];
///
/// // this panics
/// 'ß'.encode_utf8(&mut b);
/// }).join();
///
/// assert!(result.is_err());
/// ```
#[stable(feature = "unicode_encode_char", since = "1.15.0")]
#[inline]
pub fn encode_utf8(self, dst: &mut [u8]) -> &mut str {
C::encode_utf8(self, dst)
}
/// Encodes this character as UTF-16 into the provided `u16` buffer,
/// and then returns the subslice of the buffer that contains the encoded character.
///
/// # Panics
///
/// Panics if the buffer is not large enough.
/// A buffer of length 2 is large enough to encode any `char`.
///
/// # Examples
///
/// In both of these examples, '𝕊' takes two `u16`s to encode.
///
/// ```
/// let mut b = [0; 2];
///
/// let result = '𝕊'.encode_utf16(&mut b);
///
/// assert_eq!(result.len(), 2);
/// ```
///
/// A buffer that's too small:
///
/// ```
/// use std::thread;
///
/// let result = thread::spawn(|| {
/// let mut b = [0; 1];
///
/// // this panics
/// '𝕊'.encode_utf16(&mut b);
/// }).join();
///
/// assert!(result.is_err());
/// ```
#[stable(feature = "unicode_encode_char", since = "1.15.0")]
#[inline]
pub fn encode_utf16(self, dst: &mut [u16]) -> &mut [u16] {
C::encode_utf16(self, dst)
}
/// Returns true if this `char` is an alphabetic code point, and false if not.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert!('a'.is_alphabetic());
/// assert!('京'.is_alphabetic());
///
/// let c = '💝';
/// // love is many things, but it is not alphabetic
/// assert!(!c.is_alphabetic());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn is_alphabetic(self) -> bool {
match self {
'a'...'z' | 'A'...'Z' => true,
c if c > '\x7f' => derived_property::Alphabetic(c),
_ => false,
}
}
/// Returns true if this `char` satisfies the 'XID_Start' Unicode property, and false
/// otherwise.
///
/// 'XID_Start' is a Unicode Derived Property specified in
/// [UAX #31](http://unicode.org/reports/tr31/#NFKC_Modifications),
/// mostly similar to `ID_Start` but modified for closure under `NFKx`.
#[unstable(feature = "unicode",
reason = "mainly needed for compiler internals",
issue = "0")]
#[inline]
pub fn is_xid_start(self) -> bool {
derived_property::XID_Start(self)
}
/// Returns true if this `char` satisfies the 'XID_Continue' Unicode property, and false
/// otherwise.
///
/// 'XID_Continue' is a Unicode Derived Property specified in
/// [UAX #31](http://unicode.org/reports/tr31/#NFKC_Modifications),
/// mostly similar to 'ID_Continue' but modified for closure under NFKx.
#[unstable(feature = "unicode",
reason = "mainly needed for compiler internals",
issue = "0")]
#[inline]
pub fn is_xid_continue(self) -> bool {
derived_property::XID_Continue(self)
}
/// Returns true if this `char` is lowercase, and false otherwise.
///
/// 'Lowercase' is defined according to the terms of the Unicode Derived Core
/// Property `Lowercase`.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert!('a'.is_lowercase());
/// assert!('δ'.is_lowercase());
/// assert!(!'A'.is_lowercase());
/// assert!(!'Δ'.is_lowercase());
///
/// // The various Chinese scripts do not have case, and so:
/// assert!(!'中'.is_lowercase());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn is_lowercase(self) -> bool {
match self {
'a'...'z' => true,
c if c > '\x7f' => derived_property::Lowercase(c),
_ => false,
}
}
/// Returns true if this `char` is uppercase, and false otherwise.
///
/// 'Uppercase' is defined according to the terms of the Unicode Derived Core
/// Property `Uppercase`.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert!(!'a'.is_uppercase());
/// assert!(!'δ'.is_uppercase());
/// assert!('A'.is_uppercase());
/// assert!('Δ'.is_uppercase());
///
/// // The various Chinese scripts do not have case, and so:
/// assert!(!'中'.is_uppercase());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn is_uppercase(self) -> bool {
match self {
'A'...'Z' => true,
c if c > '\x7f' => derived_property::Uppercase(c),
_ => false,
}
}
/// Returns true if this `char` is whitespace, and false otherwise.
///
/// 'Whitespace' is defined according to the terms of the Unicode Derived Core
/// Property `White_Space`.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert!(' '.is_whitespace());
///
/// // a non-breaking space
/// assert!('\u{A0}'.is_whitespace());
///
/// assert!(!'越'.is_whitespace());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn is_whitespace(self) -> bool {
match self {
' ' | '\x09'...'\x0d' => true,
c if c > '\x7f' => property::White_Space(c),
_ => false,
}
}
/// Returns true if this `char` is alphanumeric, and false otherwise.
///
/// 'Alphanumeric'-ness is defined in terms of the Unicode General Categories
/// 'Nd', 'Nl', 'No' and the Derived Core Property 'Alphabetic'.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert!('٣'.is_alphanumeric());
/// assert!('7'.is_alphanumeric());
/// assert!('৬'.is_alphanumeric());
/// assert!('K'.is_alphanumeric());
/// assert!('و'.is_alphanumeric());
/// assert!('藏'.is_alphanumeric());
/// assert!(!'¾'.is_alphanumeric());
/// assert!(!'①'.is_alphanumeric());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn is_alphanumeric(self) -> bool {
self.is_alphabetic() || self.is_numeric()
}
/// Returns true if this `char` is a control code point, and false otherwise.
///
/// 'Control code point' is defined in terms of the Unicode General
/// Category `Cc`.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// // U+009C, STRING TERMINATOR
/// assert!('œ'.is_control());
/// assert!(!'q'.is_control());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn is_control(self) -> bool {
general_category::Cc(self)
}
/// Returns true if this `char` is numeric, and false otherwise.
///
/// 'Numeric'-ness is defined in terms of the Unicode General Categories
/// 'Nd', 'Nl', 'No'.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// assert!('٣'.is_numeric());
/// assert!('7'.is_numeric());
/// assert!('৬'.is_numeric());
/// assert!(!'K'.is_numeric());
/// assert!(!'و'.is_numeric());
/// assert!(!'藏'.is_numeric());
/// assert!(!'¾'.is_numeric());
/// assert!(!'①'.is_numeric());
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn is_numeric(self) -> bool {
match self {
'0'...'9' => true,
c if c > '\x7f' => general_category::N(c),
_ => false,
}
}
/// Returns an iterator that yields the lowercase equivalent of a `char`
/// as one or more `char`s.
///
/// If a character does not have a lowercase equivalent, the same character
/// will be returned back by the iterator.
///
/// This performs complex unconditional mappings with no tailoring: it maps
/// one Unicode character to its lowercase equivalent according to the
/// [Unicode database] and the additional complex mappings
/// [`SpecialCasing.txt`]. Conditional mappings (based on context or
/// language) are not considered here.
///
/// For a full reference, see [here][reference].
///
/// [Unicode database]: ftp://ftp.unicode.org/Public/UNIDATA/UnicodeData.txt
///
/// [`SpecialCasing.txt`]: ftp://ftp.unicode.org/Public/UNIDATA/SpecialCasing.txt
///
/// [reference]: http://www.unicode.org/versions/Unicode7.0.0/ch03.pdf#G33992
///
/// # Examples
///
/// As an iterator:
///
/// ```
/// for c in 'İ'.to_lowercase() {
/// print!("{}", c);
/// }
/// println!();
/// ```
///
/// Using `println!` directly:
///
/// ```
/// println!("{}", 'İ'.to_lowercase());
/// ```
///
/// Both are equivalent to:
///
/// ```
/// println!("i\u{307}");
/// ```
///
/// Using `to_string`:
///
/// ```
/// assert_eq!('C'.to_lowercase().to_string(), "c");
///
/// // Sometimes the result is more than one character:
/// assert_eq!('İ'.to_lowercase().to_string(), "i\u{307}");
///
/// // Characters that do not have both uppercase and lowercase
/// // convert into themselves.
/// assert_eq!('山'.to_lowercase().to_string(), "山");
/// ```
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn to_lowercase(self) -> ToLowercase {
ToLowercase(CaseMappingIter::new(conversions::to_lower(self)))
}
/// Returns an iterator that yields the uppercase equivalent of a `char`
/// as one or more `char`s.
///
/// If a character does not have an uppercase equivalent, the same character
/// will be returned back by the iterator.
///
/// This performs complex unconditional mappings with no tailoring: it maps
/// one Unicode character to its uppercase equivalent according to the
/// [Unicode database] and the additional complex mappings
/// [`SpecialCasing.txt`]. Conditional mappings (based on context or
/// language) are not considered here.
///
/// For a full reference, see [here][reference].
///
/// [Unicode database]: ftp://ftp.unicode.org/Public/UNIDATA/UnicodeData.txt
///
/// [`SpecialCasing.txt`]: ftp://ftp.unicode.org/Public/UNIDATA/SpecialCasing.txt
///
/// [reference]: http://www.unicode.org/versions/Unicode7.0.0/ch03.pdf#G33992
///
/// # Examples
///
/// As an iterator:
///
/// ```
/// for c in 'ß'.to_uppercase() {
/// print!("{}", c);
/// }
/// println!();
/// ```
///
/// Using `println!` directly:
///
/// ```
/// println!("{}", 'ß'.to_uppercase());
/// ```
///
/// Both are equivalent to:
///
/// ```
/// println!("SS");
/// ```
///
/// Using `to_string`:
///
/// ```
/// assert_eq!('c'.to_uppercase().to_string(), "C");
///
/// // Sometimes the result is more than one character:
/// assert_eq!('ß'.to_uppercase().to_string(), "SS");
///
/// // Characters that do not have both uppercase and lowercase
/// // convert into themselves.
/// assert_eq!('山'.to_uppercase().to_string(), "山");
/// ```
///
/// # Note on locale
///
/// In Turkish, the equivalent of 'i' in Latin has five forms instead of two:
///
/// * 'Dotless': I / ı, sometimes written ï
/// * 'Dotted': İ / i
///
/// Note that the lowercase dotted 'i' is the same as the Latin. Therefore:
///
/// ```
/// let upper_i = 'i'.to_uppercase().to_string();
/// ```
///
/// The value of `upper_i` here relies on the language of the text: if we're
/// in `en-US`, it should be `"I"`, but if we're in `tr_TR`, it should
/// be `"İ"`. `to_uppercase()` does not take this into account, and so:
///
/// ```
/// let upper_i = 'i'.to_uppercase().to_string();
///
/// assert_eq!(upper_i, "I");
/// ```
///
/// holds across languages.
#[stable(feature = "rust1", since = "1.0.0")]
#[inline]
pub fn to_uppercase(self) -> ToUppercase {
ToUppercase(CaseMappingIter::new(conversions::to_upper(self)))
}
}
/// An iterator that decodes UTF-16 encoded code points from an iterator of `u16`s.
#[stable(feature = "decode_utf16", since = "1.9.0")]
#[derive(Clone)]
pub struct DecodeUtf16<I>
where I: Iterator<Item = u16>
{
iter: I,
buf: Option<u16>,
}
/// An iterator that decodes UTF-16 encoded code points from an iterator of `u16`s.
#[stable(feature = "decode_utf16", since = "1.9.0")]
#[derive(Debug, Clone, Eq, PartialEq)]
pub struct DecodeUtf16Error {
code: u16,
}
/// Create an iterator over the UTF-16 encoded code points in `iter`,
/// returning unpaired surrogates as `Err`s.
///
/// # Examples
///
/// Basic usage:
///
/// ```
/// use std::char::decode_utf16;
///
/// fn main() {
/// // 𝄞mus<invalid>ic<invalid>
/// let v = [0xD834, 0xDD1E, 0x006d, 0x0075,
/// 0x0073, 0xDD1E, 0x0069, 0x0063,
/// 0xD834];
///
/// assert_eq!(decode_utf16(v.iter().cloned())
/// .map(|r| r.map_err(|e| e.unpaired_surrogate()))
/// .collect::<Vec<_>>(),
/// vec![Ok('𝄞'),
/// Ok('m'), Ok('u'), Ok('s'),
/// Err(0xDD1E),
/// Ok('i'), Ok('c'),
/// Err(0xD834)]);
/// }
/// ```
///
/// A lossy decoder can be obtained by replacing `Err` results with the replacement character:
///
/// ```
/// use std::char::{decode_utf16, REPLACEMENT_CHARACTER};
///
/// fn main() {
/// // 𝄞mus<invalid>ic<invalid>
/// let v = [0xD834, 0xDD1E, 0x006d, 0x0075,
/// 0x0073, 0xDD1E, 0x0069, 0x0063,
/// 0xD834];
///
/// assert_eq!(decode_utf16(v.iter().cloned())
/// .map(|r| r.unwrap_or(REPLACEMENT_CHARACTER))
/// .collect::<String>(),
/// "𝄞mus<75>ic<69>");
/// }
/// ```
#[stable(feature = "decode_utf16", since = "1.9.0")]
#[inline]
pub fn decode_utf16<I: IntoIterator<Item = u16>>(iter: I) -> DecodeUtf16<I::IntoIter> {
DecodeUtf16 {
iter: iter.into_iter(),
buf: None,
}
}
#[stable(feature = "decode_utf16", since = "1.9.0")]
impl<I: Iterator<Item = u16>> Iterator for DecodeUtf16<I> {
type Item = Result<char, DecodeUtf16Error>;
fn next(&mut self) -> Option<Result<char, DecodeUtf16Error>> {
let u = match self.buf.take() {
Some(buf) => buf,
None => {
match self.iter.next() {
Some(u) => u,
None => return None,
}
}
};
if u < 0xD800 || 0xDFFF < u {
// not a surrogate
Some(Ok(unsafe { from_u32_unchecked(u as u32) }))
} else if u >= 0xDC00 {
// a trailing surrogate
Some(Err(DecodeUtf16Error { code: u }))
} else {
let u2 = match self.iter.next() {
Some(u2) => u2,
// eof
None => return Some(Err(DecodeUtf16Error { code: u })),
};
if u2 < 0xDC00 || u2 > 0xDFFF {
// not a trailing surrogate so we're not a valid
// surrogate pair, so rewind to redecode u2 next time.
self.buf = Some(u2);
return Some(Err(DecodeUtf16Error { code: u }));
}
// all ok, so lets decode it.
let c = (((u - 0xD800) as u32) << 10 | (u2 - 0xDC00) as u32) + 0x1_0000;
Some(Ok(unsafe { from_u32_unchecked(c) }))
}
}
#[inline]
fn size_hint(&self) -> (usize, Option<usize>) {
let (low, high) = self.iter.size_hint();
// we could be entirely valid surrogates (2 elements per
// char), or entirely non-surrogates (1 element per char)
(low / 2, high)
}
}
impl DecodeUtf16Error {
/// Returns the unpaired surrogate which caused this error.
#[stable(feature = "decode_utf16", since = "1.9.0")]
pub fn unpaired_surrogate(&self) -> u16 {
self.code
}
}
#[stable(feature = "decode_utf16", since = "1.9.0")]
impl fmt::Display for DecodeUtf16Error {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
write!(f, "unpaired surrogate found: {:x}", self.code)
}
}
/// `U+FFFD REPLACEMENT CHARACTER` (<28>) is used in Unicode to represent a
/// decoding error.
///
/// It can occur, for example, when giving ill-formed UTF-8 bytes to
/// [`String::from_utf8_lossy`](../../std/string/struct.String.html#method.from_utf8_lossy).
#[stable(feature = "decode_utf16", since = "1.9.0")]
pub const REPLACEMENT_CHARACTER: char = '\u{FFFD}';