1 //! Implementation of [the WTF-8 encoding](https://simonsapin.github.io/wtf-8/).
3 //! This library uses Rust’s type system to maintain
4 //! [well-formedness](https://simonsapin.github.io/wtf-8/#well-formed),
5 //! like the `String` and `&str` types do for UTF-8.
7 //! Since [WTF-8 must not be used
8 //! for interchange](https://simonsapin.github.io/wtf-8/#intended-audience),
9 //! this library deliberately does not provide access to the underlying bytes
11 //! nor can it decode WTF-8 from arbitrary bytes.
12 //! WTF-8 strings can be obtained from UTF-8, UTF-16, or code points.
14 // this module is imported from @SimonSapin's repo and has tons of dead code on
15 // unix (it's mostly used on windows), so don't worry about dead code here.
21 use core::str::next_code_point;
23 use crate::borrow::Cow;
25 use crate::collections::TryReserveError;
27 use crate::hash::{Hash, Hasher};
28 use crate::iter::FusedIterator;
35 use crate::sys_common::AsInner;
37 const UTF8_REPLACEMENT_CHARACTER: &str = "\u{FFFD}";
39 /// A Unicode code point: from U+0000 to U+10FFFF.
41 /// Compares with the `char` type,
42 /// which represents a Unicode scalar value:
43 /// a code point that is not a surrogate (U+D800 to U+DFFF).
44 #[derive(Eq, PartialEq, Ord, PartialOrd, Clone, Copy)]
45 pub struct CodePoint {
49 /// Format the code point as `U+` followed by four to six hexadecimal digits.
50 /// Example: `U+1F4A9`
51 impl fmt::Debug for CodePoint {
53 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
54 write!(formatter, "U+{:04X}", self.value)
59 /// Unsafely creates a new `CodePoint` without checking the value.
61 /// Only use when `value` is known to be less than or equal to 0x10FFFF.
63 pub unsafe fn from_u32_unchecked(value: u32) -> CodePoint {
67 /// Creates a new `CodePoint` if the value is a valid code point.
69 /// Returns `None` if `value` is above 0x10FFFF.
71 pub fn from_u32(value: u32) -> Option<CodePoint> {
73 0..=0x10FFFF => Some(CodePoint { value }),
78 /// Creates a new `CodePoint` from a `char`.
80 /// Since all Unicode scalar values are code points, this always succeeds.
82 pub fn from_char(value: char) -> CodePoint {
83 CodePoint { value: value as u32 }
86 /// Returns the numeric value of the code point.
88 pub fn to_u32(&self) -> u32 {
92 /// Optionally returns a Unicode scalar value for the code point.
94 /// Returns `None` if the code point is a surrogate (from U+D800 to U+DFFF).
96 pub fn to_char(&self) -> Option<char> {
98 0xD800..=0xDFFF => None,
99 _ => Some(unsafe { char::from_u32_unchecked(self.value) }),
103 /// Returns a Unicode scalar value for the code point.
105 /// Returns `'\u{FFFD}'` (the replacement character “�”)
106 /// if the code point is a surrogate (from U+D800 to U+DFFF).
108 pub fn to_char_lossy(&self) -> char {
109 self.to_char().unwrap_or('\u{FFFD}')
113 /// An owned, growable string of well-formed WTF-8 data.
115 /// Similar to `String`, but can additionally contain surrogate code points
116 /// if they’re not in a surrogate pair.
117 #[derive(Eq, PartialEq, Ord, PartialOrd, Clone)]
122 impl ops::Deref for Wtf8Buf {
125 fn deref(&self) -> &Wtf8 {
130 impl ops::DerefMut for Wtf8Buf {
131 fn deref_mut(&mut self) -> &mut Wtf8 {
136 /// Format the string with double quotes,
137 /// and surrogates as `\u` followed by four hexadecimal digits.
138 /// Example: `"a\u{D800}"` for a string with code points [U+0061, U+D800]
139 impl fmt::Debug for Wtf8Buf {
141 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
142 fmt::Debug::fmt(&**self, formatter)
147 /// Creates a new, empty WTF-8 string.
149 pub fn new() -> Wtf8Buf {
150 Wtf8Buf { bytes: Vec::new() }
153 /// Creates a new, empty WTF-8 string with pre-allocated capacity for `capacity` bytes.
155 pub fn with_capacity(capacity: usize) -> Wtf8Buf {
156 Wtf8Buf { bytes: Vec::with_capacity(capacity) }
159 /// Creates a WTF-8 string from a UTF-8 `String`.
161 /// This takes ownership of the `String` and does not copy.
163 /// Since WTF-8 is a superset of UTF-8, this always succeeds.
165 pub fn from_string(string: String) -> Wtf8Buf {
166 Wtf8Buf { bytes: string.into_bytes() }
169 /// Creates a WTF-8 string from a UTF-8 `&str` slice.
171 /// This copies the content of the slice.
173 /// Since WTF-8 is a superset of UTF-8, this always succeeds.
175 pub fn from_str(str: &str) -> Wtf8Buf {
176 Wtf8Buf { bytes: <[_]>::to_vec(str.as_bytes()) }
179 pub fn clear(&mut self) {
183 /// Creates a WTF-8 string from a potentially ill-formed UTF-16 slice of 16-bit code units.
185 /// This is lossless: calling `.encode_wide()` on the resulting string
186 /// will always return the original code units.
187 pub fn from_wide(v: &[u16]) -> Wtf8Buf {
188 let mut string = Wtf8Buf::with_capacity(v.len());
189 for item in char::decode_utf16(v.iter().cloned()) {
191 Ok(ch) => string.push_char(ch),
193 let surrogate = surrogate.unpaired_surrogate();
194 // Surrogates are known to be in the code point range.
195 let code_point = unsafe { CodePoint::from_u32_unchecked(surrogate as u32) };
196 // Skip the WTF-8 concatenation check,
197 // surrogate pairs are already decoded by decode_utf16
198 string.push_code_point_unchecked(code_point)
205 /// Copied from String::push
206 /// This does **not** include the WTF-8 concatenation check.
207 fn push_code_point_unchecked(&mut self, code_point: CodePoint) {
208 let mut bytes = [0; 4];
209 let bytes = char::encode_utf8_raw(code_point.value, &mut bytes);
210 self.bytes.extend_from_slice(bytes)
214 pub fn as_slice(&self) -> &Wtf8 {
215 unsafe { Wtf8::from_bytes_unchecked(&self.bytes) }
219 pub fn as_mut_slice(&mut self) -> &mut Wtf8 {
220 unsafe { Wtf8::from_mut_bytes_unchecked(&mut self.bytes) }
223 /// Reserves capacity for at least `additional` more bytes to be inserted
224 /// in the given `Wtf8Buf`.
225 /// The collection may reserve more space to avoid frequent reallocations.
229 /// Panics if the new capacity overflows `usize`.
231 pub fn reserve(&mut self, additional: usize) {
232 self.bytes.reserve(additional)
235 /// Tries to reserve capacity for at least `additional` more length units
236 /// in the given `Wtf8Buf`. The `Wtf8Buf` may reserve more space to avoid
237 /// frequent reallocations. After calling `try_reserve`, capacity will be
238 /// greater than or equal to `self.len() + additional`. Does nothing if
239 /// capacity is already sufficient. This method preserves the contents even
240 /// if an error occurs.
244 /// If the capacity overflows, or the allocator reports a failure, then an error
247 pub fn try_reserve(&mut self, additional: usize) -> Result<(), TryReserveError> {
248 self.bytes.try_reserve(additional)
252 pub fn reserve_exact(&mut self, additional: usize) {
253 self.bytes.reserve_exact(additional)
256 /// Tries to reserve the minimum capacity for exactly `additional`
257 /// length units in the given `Wtf8Buf`. After calling
258 /// `try_reserve_exact`, capacity will be greater than or equal to
259 /// `self.len() + additional` if it returns `Ok(())`.
260 /// Does nothing if the capacity is already sufficient.
262 /// Note that the allocator may give the `Wtf8Buf` more space than it
263 /// requests. Therefore, capacity can not be relied upon to be precisely
264 /// minimal. Prefer [`try_reserve`] if future insertions are expected.
266 /// [`try_reserve`]: Wtf8Buf::try_reserve
270 /// If the capacity overflows, or the allocator reports a failure, then an error
273 pub fn try_reserve_exact(&mut self, additional: usize) -> Result<(), TryReserveError> {
274 self.bytes.try_reserve_exact(additional)
278 pub fn shrink_to_fit(&mut self) {
279 self.bytes.shrink_to_fit()
283 pub fn shrink_to(&mut self, min_capacity: usize) {
284 self.bytes.shrink_to(min_capacity)
287 /// Returns the number of bytes that this string buffer can hold without reallocating.
289 pub fn capacity(&self) -> usize {
290 self.bytes.capacity()
293 /// Append a UTF-8 slice at the end of the string.
295 pub fn push_str(&mut self, other: &str) {
296 self.bytes.extend_from_slice(other.as_bytes())
299 /// Append a WTF-8 slice at the end of the string.
301 /// This replaces newly paired surrogates at the boundary
302 /// with a supplementary code point,
303 /// like concatenating ill-formed UTF-16 strings effectively would.
305 pub fn push_wtf8(&mut self, other: &Wtf8) {
306 match ((&*self).final_lead_surrogate(), other.initial_trail_surrogate()) {
307 // Replace newly paired surrogates by a supplementary code point.
308 (Some(lead), Some(trail)) => {
309 let len_without_lead_surrogate = self.len() - 3;
310 self.bytes.truncate(len_without_lead_surrogate);
311 let other_without_trail_surrogate = &other.bytes[3..];
312 // 4 bytes for the supplementary code point
313 self.bytes.reserve(4 + other_without_trail_surrogate.len());
314 self.push_char(decode_surrogate_pair(lead, trail));
315 self.bytes.extend_from_slice(other_without_trail_surrogate);
317 _ => self.bytes.extend_from_slice(&other.bytes),
321 /// Append a Unicode scalar value at the end of the string.
323 pub fn push_char(&mut self, c: char) {
324 self.push_code_point_unchecked(CodePoint::from_char(c))
327 /// Append a code point at the end of the string.
329 /// This replaces newly paired surrogates at the boundary
330 /// with a supplementary code point,
331 /// like concatenating ill-formed UTF-16 strings effectively would.
333 pub fn push(&mut self, code_point: CodePoint) {
334 if let trail @ 0xDC00..=0xDFFF = code_point.to_u32() {
335 if let Some(lead) = (&*self).final_lead_surrogate() {
336 let len_without_lead_surrogate = self.len() - 3;
337 self.bytes.truncate(len_without_lead_surrogate);
338 self.push_char(decode_surrogate_pair(lead, trail as u16));
343 // No newly paired surrogates at the boundary.
344 self.push_code_point_unchecked(code_point)
347 /// Shortens a string to the specified length.
351 /// Panics if `new_len` > current length,
352 /// or if `new_len` is not a code point boundary.
354 pub fn truncate(&mut self, new_len: usize) {
355 assert!(is_code_point_boundary(self, new_len));
356 self.bytes.truncate(new_len)
359 /// Consumes the WTF-8 string and tries to convert it to UTF-8.
361 /// This does not copy the data.
363 /// If the contents are not well-formed UTF-8
364 /// (that is, if the string contains surrogates),
365 /// the original WTF-8 string is returned instead.
366 pub fn into_string(self) -> Result<String, Wtf8Buf> {
367 match self.next_surrogate(0) {
368 None => Ok(unsafe { String::from_utf8_unchecked(self.bytes) }),
369 Some(_) => Err(self),
373 /// Consumes the WTF-8 string and converts it lossily to UTF-8.
375 /// This does not copy the data (but may overwrite parts of it in place).
377 /// Surrogates are replaced with `"\u{FFFD}"` (the replacement character “�”)
378 pub fn into_string_lossy(mut self) -> String {
381 match self.next_surrogate(pos) {
382 Some((surrogate_pos, _)) => {
383 pos = surrogate_pos + 3;
384 self.bytes[surrogate_pos..pos]
385 .copy_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
387 None => return unsafe { String::from_utf8_unchecked(self.bytes) },
392 /// Converts this `Wtf8Buf` into a boxed `Wtf8`.
394 pub fn into_box(self) -> Box<Wtf8> {
395 unsafe { mem::transmute(self.bytes.into_boxed_slice()) }
398 /// Converts a `Box<Wtf8>` into a `Wtf8Buf`.
399 pub fn from_box(boxed: Box<Wtf8>) -> Wtf8Buf {
400 let bytes: Box<[u8]> = unsafe { mem::transmute(boxed) };
401 Wtf8Buf { bytes: bytes.into_vec() }
405 /// Creates a new WTF-8 string from an iterator of code points.
407 /// This replaces surrogate code point pairs with supplementary code points,
408 /// like concatenating ill-formed UTF-16 strings effectively would.
409 impl FromIterator<CodePoint> for Wtf8Buf {
410 fn from_iter<T: IntoIterator<Item = CodePoint>>(iter: T) -> Wtf8Buf {
411 let mut string = Wtf8Buf::new();
417 /// Append code points from an iterator to the string.
419 /// This replaces surrogate code point pairs with supplementary code points,
420 /// like concatenating ill-formed UTF-16 strings effectively would.
421 impl Extend<CodePoint> for Wtf8Buf {
422 fn extend<T: IntoIterator<Item = CodePoint>>(&mut self, iter: T) {
423 let iterator = iter.into_iter();
424 let (low, _high) = iterator.size_hint();
425 // Lower bound of one byte per code point (ASCII only)
426 self.bytes.reserve(low);
427 iterator.for_each(move |code_point| self.push(code_point));
431 fn extend_one(&mut self, code_point: CodePoint) {
432 self.push(code_point);
436 fn extend_reserve(&mut self, additional: usize) {
437 // Lower bound of one byte per code point (ASCII only)
438 self.bytes.reserve(additional);
442 /// A borrowed slice of well-formed WTF-8 data.
444 /// Similar to `&str`, but can additionally contain surrogate code points
445 /// if they’re not in a surrogate pair.
446 #[derive(Eq, Ord, PartialEq, PartialOrd)]
451 impl AsInner<[u8]> for Wtf8 {
452 fn as_inner(&self) -> &[u8] {
457 /// Format the slice with double quotes,
458 /// and surrogates as `\u` followed by four hexadecimal digits.
459 /// Example: `"a\u{D800}"` for a slice with code points [U+0061, U+D800]
460 impl fmt::Debug for Wtf8 {
461 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
462 fn write_str_escaped(f: &mut fmt::Formatter<'_>, s: &str) -> fmt::Result {
463 use crate::fmt::Write;
464 for c in s.chars().flat_map(|c| c.escape_debug()) {
470 formatter.write_str("\"")?;
472 while let Some((surrogate_pos, surrogate)) = self.next_surrogate(pos) {
473 write_str_escaped(formatter, unsafe {
474 str::from_utf8_unchecked(&self.bytes[pos..surrogate_pos])
476 write!(formatter, "\\u{{{:x}}}", surrogate)?;
477 pos = surrogate_pos + 3;
479 write_str_escaped(formatter, unsafe { str::from_utf8_unchecked(&self.bytes[pos..]) })?;
480 formatter.write_str("\"")
484 impl fmt::Display for Wtf8 {
485 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
486 let wtf8_bytes = &self.bytes;
489 match self.next_surrogate(pos) {
490 Some((surrogate_pos, _)) => {
491 formatter.write_str(unsafe {
492 str::from_utf8_unchecked(&wtf8_bytes[pos..surrogate_pos])
494 formatter.write_str(UTF8_REPLACEMENT_CHARACTER)?;
495 pos = surrogate_pos + 3;
498 let s = unsafe { str::from_utf8_unchecked(&wtf8_bytes[pos..]) };
499 if pos == 0 { return s.fmt(formatter) } else { return formatter.write_str(s) }
507 /// Creates a WTF-8 slice from a UTF-8 `&str` slice.
509 /// Since WTF-8 is a superset of UTF-8, this always succeeds.
511 pub fn from_str(value: &str) -> &Wtf8 {
512 unsafe { Wtf8::from_bytes_unchecked(value.as_bytes()) }
515 /// Creates a WTF-8 slice from a WTF-8 byte slice.
517 /// Since the byte slice is not checked for valid WTF-8, this functions is
520 unsafe fn from_bytes_unchecked(value: &[u8]) -> &Wtf8 {
521 mem::transmute(value)
524 /// Creates a mutable WTF-8 slice from a mutable WTF-8 byte slice.
526 /// Since the byte slice is not checked for valid WTF-8, this functions is
529 unsafe fn from_mut_bytes_unchecked(value: &mut [u8]) -> &mut Wtf8 {
530 mem::transmute(value)
533 /// Returns the length, in WTF-8 bytes.
535 pub fn len(&self) -> usize {
540 pub fn is_empty(&self) -> bool {
541 self.bytes.is_empty()
544 /// Returns the code point at `position` if it is in the ASCII range,
545 /// or `b'\xFF' otherwise.
549 /// Panics if `position` is beyond the end of the string.
551 pub fn ascii_byte_at(&self, position: usize) -> u8 {
552 match self.bytes[position] {
553 ascii_byte @ 0x00..=0x7F => ascii_byte,
558 /// Returns an iterator for the string’s code points.
560 pub fn code_points(&self) -> Wtf8CodePoints<'_> {
561 Wtf8CodePoints { bytes: self.bytes.iter() }
564 /// Tries to convert the string to UTF-8 and return a `&str` slice.
566 /// Returns `None` if the string contains surrogates.
568 /// This does not copy the data.
570 pub fn as_str(&self) -> Option<&str> {
571 // Well-formed WTF-8 is also well-formed UTF-8
572 // if and only if it contains no surrogate.
573 match self.next_surrogate(0) {
574 None => Some(unsafe { str::from_utf8_unchecked(&self.bytes) }),
579 /// Lossily converts the string to UTF-8.
580 /// Returns a UTF-8 `&str` slice if the contents are well-formed in UTF-8.
582 /// Surrogates are replaced with `"\u{FFFD}"` (the replacement character “�”).
584 /// This only copies the data if necessary (if it contains any surrogate).
585 pub fn to_string_lossy(&self) -> Cow<'_, str> {
586 let surrogate_pos = match self.next_surrogate(0) {
587 None => return Cow::Borrowed(unsafe { str::from_utf8_unchecked(&self.bytes) }),
588 Some((pos, _)) => pos,
590 let wtf8_bytes = &self.bytes;
591 let mut utf8_bytes = Vec::with_capacity(self.len());
592 utf8_bytes.extend_from_slice(&wtf8_bytes[..surrogate_pos]);
593 utf8_bytes.extend_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
594 let mut pos = surrogate_pos + 3;
596 match self.next_surrogate(pos) {
597 Some((surrogate_pos, _)) => {
598 utf8_bytes.extend_from_slice(&wtf8_bytes[pos..surrogate_pos]);
599 utf8_bytes.extend_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
600 pos = surrogate_pos + 3;
603 utf8_bytes.extend_from_slice(&wtf8_bytes[pos..]);
604 return Cow::Owned(unsafe { String::from_utf8_unchecked(utf8_bytes) });
610 /// Converts the WTF-8 string to potentially ill-formed UTF-16
611 /// and return an iterator of 16-bit code units.
613 /// This is lossless:
614 /// calling `Wtf8Buf::from_ill_formed_utf16` on the resulting code units
615 /// would always return the original WTF-8 string.
617 pub fn encode_wide(&self) -> EncodeWide<'_> {
618 EncodeWide { code_points: self.code_points(), extra: 0 }
622 fn next_surrogate(&self, mut pos: usize) -> Option<(usize, u16)> {
623 let mut iter = self.bytes[pos..].iter();
625 let b = *iter.next()?;
631 } else if b == 0xED {
632 match (iter.next(), iter.next()) {
633 (Some(&b2), Some(&b3)) if b2 >= 0xA0 => {
634 return Some((pos, decode_surrogate(b2, b3)));
652 fn final_lead_surrogate(&self) -> Option<u16> {
654 [.., 0xED, b2 @ 0xA0..=0xAF, b3] => Some(decode_surrogate(b2, b3)),
660 fn initial_trail_surrogate(&self) -> Option<u16> {
662 [0xED, b2 @ 0xB0..=0xBF, b3, ..] => Some(decode_surrogate(b2, b3)),
667 pub fn clone_into(&self, buf: &mut Wtf8Buf) {
668 self.bytes.clone_into(&mut buf.bytes)
671 /// Boxes this `Wtf8`.
673 pub fn into_box(&self) -> Box<Wtf8> {
674 let boxed: Box<[u8]> = self.bytes.into();
675 unsafe { mem::transmute(boxed) }
678 /// Creates a boxed, empty `Wtf8`.
679 pub fn empty_box() -> Box<Wtf8> {
680 let boxed: Box<[u8]> = Default::default();
681 unsafe { mem::transmute(boxed) }
685 pub fn into_arc(&self) -> Arc<Wtf8> {
686 let arc: Arc<[u8]> = Arc::from(&self.bytes);
687 unsafe { Arc::from_raw(Arc::into_raw(arc) as *const Wtf8) }
691 pub fn into_rc(&self) -> Rc<Wtf8> {
692 let rc: Rc<[u8]> = Rc::from(&self.bytes);
693 unsafe { Rc::from_raw(Rc::into_raw(rc) as *const Wtf8) }
697 pub fn make_ascii_lowercase(&mut self) {
698 self.bytes.make_ascii_lowercase()
702 pub fn make_ascii_uppercase(&mut self) {
703 self.bytes.make_ascii_uppercase()
707 pub fn to_ascii_lowercase(&self) -> Wtf8Buf {
708 Wtf8Buf { bytes: self.bytes.to_ascii_lowercase() }
712 pub fn to_ascii_uppercase(&self) -> Wtf8Buf {
713 Wtf8Buf { bytes: self.bytes.to_ascii_uppercase() }
717 pub fn is_ascii(&self) -> bool {
718 self.bytes.is_ascii()
722 pub fn eq_ignore_ascii_case(&self, other: &Self) -> bool {
723 self.bytes.eq_ignore_ascii_case(&other.bytes)
727 /// Returns a slice of the given string for the byte range \[`begin`..`end`).
731 /// Panics when `begin` and `end` do not point to code point boundaries,
732 /// or point beyond the end of the string.
733 impl ops::Index<ops::Range<usize>> for Wtf8 {
737 fn index(&self, range: ops::Range<usize>) -> &Wtf8 {
738 // is_code_point_boundary checks that the index is in [0, .len()]
739 if range.start <= range.end
740 && is_code_point_boundary(self, range.start)
741 && is_code_point_boundary(self, range.end)
743 unsafe { slice_unchecked(self, range.start, range.end) }
745 slice_error_fail(self, range.start, range.end)
750 /// Returns a slice of the given string from byte `begin` to its end.
754 /// Panics when `begin` is not at a code point boundary,
755 /// or is beyond the end of the string.
756 impl ops::Index<ops::RangeFrom<usize>> for Wtf8 {
760 fn index(&self, range: ops::RangeFrom<usize>) -> &Wtf8 {
761 // is_code_point_boundary checks that the index is in [0, .len()]
762 if is_code_point_boundary(self, range.start) {
763 unsafe { slice_unchecked(self, range.start, self.len()) }
765 slice_error_fail(self, range.start, self.len())
770 /// Returns a slice of the given string from its beginning to byte `end`.
774 /// Panics when `end` is not at a code point boundary,
775 /// or is beyond the end of the string.
776 impl ops::Index<ops::RangeTo<usize>> for Wtf8 {
780 fn index(&self, range: ops::RangeTo<usize>) -> &Wtf8 {
781 // is_code_point_boundary checks that the index is in [0, .len()]
782 if is_code_point_boundary(self, range.end) {
783 unsafe { slice_unchecked(self, 0, range.end) }
785 slice_error_fail(self, 0, range.end)
790 impl ops::Index<ops::RangeFull> for Wtf8 {
794 fn index(&self, _range: ops::RangeFull) -> &Wtf8 {
800 fn decode_surrogate(second_byte: u8, third_byte: u8) -> u16 {
801 // The first byte is assumed to be 0xED
802 0xD800 | (second_byte as u16 & 0x3F) << 6 | third_byte as u16 & 0x3F
806 fn decode_surrogate_pair(lead: u16, trail: u16) -> char {
807 let code_point = 0x10000 + ((((lead - 0xD800) as u32) << 10) | (trail - 0xDC00) as u32);
808 unsafe { char::from_u32_unchecked(code_point) }
811 /// Copied from core::str::StrPrelude::is_char_boundary
813 pub fn is_code_point_boundary(slice: &Wtf8, index: usize) -> bool {
814 if index == slice.len() {
817 match slice.bytes.get(index) {
819 Some(&b) => b < 128 || b >= 192,
823 /// Copied from core::str::raw::slice_unchecked
825 pub unsafe fn slice_unchecked(s: &Wtf8, begin: usize, end: usize) -> &Wtf8 {
826 // memory layout of a &[u8] and &Wtf8 are the same
827 Wtf8::from_bytes_unchecked(slice::from_raw_parts(s.bytes.as_ptr().add(begin), end - begin))
830 /// Copied from core::str::raw::slice_error_fail
832 pub fn slice_error_fail(s: &Wtf8, begin: usize, end: usize) -> ! {
833 assert!(begin <= end);
834 panic!("index {begin} and/or {end} in `{s:?}` do not lie on character boundary");
837 /// Iterator for the code points of a WTF-8 string.
839 /// Created with the method `.code_points()`.
841 pub struct Wtf8CodePoints<'a> {
842 bytes: slice::Iter<'a, u8>,
845 impl<'a> Iterator for Wtf8CodePoints<'a> {
846 type Item = CodePoint;
849 fn next(&mut self) -> Option<CodePoint> {
850 // SAFETY: `self.bytes` has been created from a WTF-8 string
851 unsafe { next_code_point(&mut self.bytes).map(|c| CodePoint { value: c }) }
855 fn size_hint(&self) -> (usize, Option<usize>) {
856 let len = self.bytes.len();
857 (len.saturating_add(3) / 4, Some(len))
861 /// Generates a wide character sequence for potentially ill-formed UTF-16.
862 #[stable(feature = "rust1", since = "1.0.0")]
864 pub struct EncodeWide<'a> {
865 code_points: Wtf8CodePoints<'a>,
869 // Copied from libunicode/u_str.rs
870 #[stable(feature = "rust1", since = "1.0.0")]
871 impl<'a> Iterator for EncodeWide<'a> {
875 fn next(&mut self) -> Option<u16> {
877 let tmp = self.extra;
882 let mut buf = [0; 2];
883 self.code_points.next().map(|code_point| {
884 let n = char::encode_utf16_raw(code_point.value, &mut buf).len();
893 fn size_hint(&self) -> (usize, Option<usize>) {
894 let (low, high) = self.code_points.size_hint();
895 let ext = (self.extra != 0) as usize;
896 // every code point gets either one u16 or two u16,
897 // so this iterator is between 1 or 2 times as
898 // long as the underlying iterator.
899 (low + ext, high.and_then(|n| n.checked_mul(2)).and_then(|n| n.checked_add(ext)))
903 #[stable(feature = "encode_wide_fused_iterator", since = "1.62.0")]
904 impl FusedIterator for EncodeWide<'_> {}
906 impl Hash for CodePoint {
908 fn hash<H: Hasher>(&self, state: &mut H) {
909 self.value.hash(state)
913 impl Hash for Wtf8Buf {
915 fn hash<H: Hasher>(&self, state: &mut H) {
916 state.write(&self.bytes);
923 fn hash<H: Hasher>(&self, state: &mut H) {
924 state.write(&self.bytes);