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.
243 /// If the capacity overflows, or the allocator reports a failure, then an error
246 pub fn try_reserve(&mut self, additional: usize) -> Result<(), TryReserveError> {
247 self.bytes.try_reserve(additional)
251 pub fn reserve_exact(&mut self, additional: usize) {
252 self.bytes.reserve_exact(additional)
255 /// Tries to reserve the minimum capacity for exactly `additional`
256 /// length units in the given `Wtf8Buf`. After calling
257 /// `try_reserve_exact`, capacity will be greater than or equal to
258 /// `self.len() + additional` if it returns `Ok(())`.
259 /// Does nothing if the capacity is already sufficient.
261 /// Note that the allocator may give the `Wtf8Buf` more space than it
262 /// requests. Therefore, capacity can not be relied upon to be precisely
263 /// minimal. Prefer [`try_reserve`] if future insertions are expected.
265 /// [`try_reserve`]: Wtf8Buf::try_reserve
269 /// If the capacity overflows, or the allocator reports a failure, then an error
272 pub fn try_reserve_exact(&mut self, additional: usize) -> Result<(), TryReserveError> {
273 self.bytes.try_reserve_exact(additional)
277 pub fn shrink_to_fit(&mut self) {
278 self.bytes.shrink_to_fit()
282 pub fn shrink_to(&mut self, min_capacity: usize) {
283 self.bytes.shrink_to(min_capacity)
286 /// Returns the number of bytes that this string buffer can hold without reallocating.
288 pub fn capacity(&self) -> usize {
289 self.bytes.capacity()
292 /// Append a UTF-8 slice at the end of the string.
294 pub fn push_str(&mut self, other: &str) {
295 self.bytes.extend_from_slice(other.as_bytes())
298 /// Append a WTF-8 slice at the end of the string.
300 /// This replaces newly paired surrogates at the boundary
301 /// with a supplementary code point,
302 /// like concatenating ill-formed UTF-16 strings effectively would.
304 pub fn push_wtf8(&mut self, other: &Wtf8) {
305 match ((&*self).final_lead_surrogate(), other.initial_trail_surrogate()) {
306 // Replace newly paired surrogates by a supplementary code point.
307 (Some(lead), Some(trail)) => {
308 let len_without_lead_surrogate = self.len() - 3;
309 self.bytes.truncate(len_without_lead_surrogate);
310 let other_without_trail_surrogate = &other.bytes[3..];
311 // 4 bytes for the supplementary code point
312 self.bytes.reserve(4 + other_without_trail_surrogate.len());
313 self.push_char(decode_surrogate_pair(lead, trail));
314 self.bytes.extend_from_slice(other_without_trail_surrogate);
316 _ => self.bytes.extend_from_slice(&other.bytes),
320 /// Append a Unicode scalar value at the end of the string.
322 pub fn push_char(&mut self, c: char) {
323 self.push_code_point_unchecked(CodePoint::from_char(c))
326 /// Append a code point at the end of the string.
328 /// This replaces newly paired surrogates at the boundary
329 /// with a supplementary code point,
330 /// like concatenating ill-formed UTF-16 strings effectively would.
332 pub fn push(&mut self, code_point: CodePoint) {
333 if let trail @ 0xDC00..=0xDFFF = code_point.to_u32() {
334 if let Some(lead) = (&*self).final_lead_surrogate() {
335 let len_without_lead_surrogate = self.len() - 3;
336 self.bytes.truncate(len_without_lead_surrogate);
337 self.push_char(decode_surrogate_pair(lead, trail as u16));
342 // No newly paired surrogates at the boundary.
343 self.push_code_point_unchecked(code_point)
346 /// Shortens a string to the specified length.
350 /// Panics if `new_len` > current length,
351 /// or if `new_len` is not a code point boundary.
353 pub fn truncate(&mut self, new_len: usize) {
354 assert!(is_code_point_boundary(self, new_len));
355 self.bytes.truncate(new_len)
358 /// Consumes the WTF-8 string and tries to convert it to UTF-8.
360 /// This does not copy the data.
362 /// If the contents are not well-formed UTF-8
363 /// (that is, if the string contains surrogates),
364 /// the original WTF-8 string is returned instead.
365 pub fn into_string(self) -> Result<String, Wtf8Buf> {
366 match self.next_surrogate(0) {
367 None => Ok(unsafe { String::from_utf8_unchecked(self.bytes) }),
368 Some(_) => Err(self),
372 /// Consumes the WTF-8 string and converts it lossily to UTF-8.
374 /// This does not copy the data (but may overwrite parts of it in place).
376 /// Surrogates are replaced with `"\u{FFFD}"` (the replacement character “�”)
377 pub fn into_string_lossy(mut self) -> String {
380 match self.next_surrogate(pos) {
381 Some((surrogate_pos, _)) => {
382 pos = surrogate_pos + 3;
383 self.bytes[surrogate_pos..pos]
384 .copy_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
386 None => return unsafe { String::from_utf8_unchecked(self.bytes) },
391 /// Converts this `Wtf8Buf` into a boxed `Wtf8`.
393 pub fn into_box(self) -> Box<Wtf8> {
394 unsafe { mem::transmute(self.bytes.into_boxed_slice()) }
397 /// Converts a `Box<Wtf8>` into a `Wtf8Buf`.
398 pub fn from_box(boxed: Box<Wtf8>) -> Wtf8Buf {
399 let bytes: Box<[u8]> = unsafe { mem::transmute(boxed) };
400 Wtf8Buf { bytes: bytes.into_vec() }
404 /// Creates a new WTF-8 string from an iterator of code points.
406 /// This replaces surrogate code point pairs with supplementary code points,
407 /// like concatenating ill-formed UTF-16 strings effectively would.
408 impl FromIterator<CodePoint> for Wtf8Buf {
409 fn from_iter<T: IntoIterator<Item = CodePoint>>(iter: T) -> Wtf8Buf {
410 let mut string = Wtf8Buf::new();
416 /// Append code points from an iterator to the string.
418 /// This replaces surrogate code point pairs with supplementary code points,
419 /// like concatenating ill-formed UTF-16 strings effectively would.
420 impl Extend<CodePoint> for Wtf8Buf {
421 fn extend<T: IntoIterator<Item = CodePoint>>(&mut self, iter: T) {
422 let iterator = iter.into_iter();
423 let (low, _high) = iterator.size_hint();
424 // Lower bound of one byte per code point (ASCII only)
425 self.bytes.reserve(low);
426 iterator.for_each(move |code_point| self.push(code_point));
430 fn extend_one(&mut self, code_point: CodePoint) {
431 self.push(code_point);
435 fn extend_reserve(&mut self, additional: usize) {
436 // Lower bound of one byte per code point (ASCII only)
437 self.bytes.reserve(additional);
441 /// A borrowed slice of well-formed WTF-8 data.
443 /// Similar to `&str`, but can additionally contain surrogate code points
444 /// if they’re not in a surrogate pair.
445 #[derive(Eq, Ord, PartialEq, PartialOrd)]
450 impl AsInner<[u8]> for Wtf8 {
451 fn as_inner(&self) -> &[u8] {
456 /// Format the slice with double quotes,
457 /// and surrogates as `\u` followed by four hexadecimal digits.
458 /// Example: `"a\u{D800}"` for a slice with code points [U+0061, U+D800]
459 impl fmt::Debug for Wtf8 {
460 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
461 fn write_str_escaped(f: &mut fmt::Formatter<'_>, s: &str) -> fmt::Result {
462 use crate::fmt::Write;
463 for c in s.chars().flat_map(|c| c.escape_debug()) {
469 formatter.write_str("\"")?;
471 while let Some((surrogate_pos, surrogate)) = self.next_surrogate(pos) {
472 write_str_escaped(formatter, unsafe {
473 str::from_utf8_unchecked(&self.bytes[pos..surrogate_pos])
475 write!(formatter, "\\u{{{:x}}}", surrogate)?;
476 pos = surrogate_pos + 3;
478 write_str_escaped(formatter, unsafe { str::from_utf8_unchecked(&self.bytes[pos..]) })?;
479 formatter.write_str("\"")
483 impl fmt::Display for Wtf8 {
484 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
485 let wtf8_bytes = &self.bytes;
488 match self.next_surrogate(pos) {
489 Some((surrogate_pos, _)) => {
490 formatter.write_str(unsafe {
491 str::from_utf8_unchecked(&wtf8_bytes[pos..surrogate_pos])
493 formatter.write_str(UTF8_REPLACEMENT_CHARACTER)?;
494 pos = surrogate_pos + 3;
497 let s = unsafe { str::from_utf8_unchecked(&wtf8_bytes[pos..]) };
498 if pos == 0 { return s.fmt(formatter) } else { return formatter.write_str(s) }
506 /// Creates a WTF-8 slice from a UTF-8 `&str` slice.
508 /// Since WTF-8 is a superset of UTF-8, this always succeeds.
510 pub fn from_str(value: &str) -> &Wtf8 {
511 unsafe { Wtf8::from_bytes_unchecked(value.as_bytes()) }
514 /// Creates a WTF-8 slice from a WTF-8 byte slice.
516 /// Since the byte slice is not checked for valid WTF-8, this functions is
519 unsafe fn from_bytes_unchecked(value: &[u8]) -> &Wtf8 {
520 mem::transmute(value)
523 /// Creates a mutable WTF-8 slice from a mutable WTF-8 byte slice.
525 /// Since the byte slice is not checked for valid WTF-8, this functions is
528 unsafe fn from_mut_bytes_unchecked(value: &mut [u8]) -> &mut Wtf8 {
529 mem::transmute(value)
532 /// Returns the length, in WTF-8 bytes.
534 pub fn len(&self) -> usize {
539 pub fn is_empty(&self) -> bool {
540 self.bytes.is_empty()
543 /// Returns the code point at `position` if it is in the ASCII range,
544 /// or `b'\xFF' otherwise.
548 /// Panics if `position` is beyond the end of the string.
550 pub fn ascii_byte_at(&self, position: usize) -> u8 {
551 match self.bytes[position] {
552 ascii_byte @ 0x00..=0x7F => ascii_byte,
557 /// Returns an iterator for the string’s code points.
559 pub fn code_points(&self) -> Wtf8CodePoints<'_> {
560 Wtf8CodePoints { bytes: self.bytes.iter() }
563 /// Tries to convert the string to UTF-8 and return a `&str` slice.
565 /// Returns `None` if the string contains surrogates.
567 /// This does not copy the data.
569 pub fn as_str(&self) -> Option<&str> {
570 // Well-formed WTF-8 is also well-formed UTF-8
571 // if and only if it contains no surrogate.
572 match self.next_surrogate(0) {
573 None => Some(unsafe { str::from_utf8_unchecked(&self.bytes) }),
578 /// Lossily converts the string to UTF-8.
579 /// Returns a UTF-8 `&str` slice if the contents are well-formed in UTF-8.
581 /// Surrogates are replaced with `"\u{FFFD}"` (the replacement character “�”).
583 /// This only copies the data if necessary (if it contains any surrogate).
584 pub fn to_string_lossy(&self) -> Cow<'_, str> {
585 let surrogate_pos = match self.next_surrogate(0) {
586 None => return Cow::Borrowed(unsafe { str::from_utf8_unchecked(&self.bytes) }),
587 Some((pos, _)) => pos,
589 let wtf8_bytes = &self.bytes;
590 let mut utf8_bytes = Vec::with_capacity(self.len());
591 utf8_bytes.extend_from_slice(&wtf8_bytes[..surrogate_pos]);
592 utf8_bytes.extend_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
593 let mut pos = surrogate_pos + 3;
595 match self.next_surrogate(pos) {
596 Some((surrogate_pos, _)) => {
597 utf8_bytes.extend_from_slice(&wtf8_bytes[pos..surrogate_pos]);
598 utf8_bytes.extend_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
599 pos = surrogate_pos + 3;
602 utf8_bytes.extend_from_slice(&wtf8_bytes[pos..]);
603 return Cow::Owned(unsafe { String::from_utf8_unchecked(utf8_bytes) });
609 /// Converts the WTF-8 string to potentially ill-formed UTF-16
610 /// and return an iterator of 16-bit code units.
612 /// This is lossless:
613 /// calling `Wtf8Buf::from_ill_formed_utf16` on the resulting code units
614 /// would always return the original WTF-8 string.
616 pub fn encode_wide(&self) -> EncodeWide<'_> {
617 EncodeWide { code_points: self.code_points(), extra: 0 }
621 fn next_surrogate(&self, mut pos: usize) -> Option<(usize, u16)> {
622 let mut iter = self.bytes[pos..].iter();
624 let b = *iter.next()?;
630 } else if b == 0xED {
631 match (iter.next(), iter.next()) {
632 (Some(&b2), Some(&b3)) if b2 >= 0xA0 => {
633 return Some((pos, decode_surrogate(b2, b3)));
651 fn final_lead_surrogate(&self) -> Option<u16> {
653 [.., 0xED, b2 @ 0xA0..=0xAF, b3] => Some(decode_surrogate(b2, b3)),
659 fn initial_trail_surrogate(&self) -> Option<u16> {
661 [0xED, b2 @ 0xB0..=0xBF, b3, ..] => Some(decode_surrogate(b2, b3)),
666 pub fn clone_into(&self, buf: &mut Wtf8Buf) {
667 self.bytes.clone_into(&mut buf.bytes)
670 /// Boxes this `Wtf8`.
672 pub fn into_box(&self) -> Box<Wtf8> {
673 let boxed: Box<[u8]> = self.bytes.into();
674 unsafe { mem::transmute(boxed) }
677 /// Creates a boxed, empty `Wtf8`.
678 pub fn empty_box() -> Box<Wtf8> {
679 let boxed: Box<[u8]> = Default::default();
680 unsafe { mem::transmute(boxed) }
684 pub fn into_arc(&self) -> Arc<Wtf8> {
685 let arc: Arc<[u8]> = Arc::from(&self.bytes);
686 unsafe { Arc::from_raw(Arc::into_raw(arc) as *const Wtf8) }
690 pub fn into_rc(&self) -> Rc<Wtf8> {
691 let rc: Rc<[u8]> = Rc::from(&self.bytes);
692 unsafe { Rc::from_raw(Rc::into_raw(rc) as *const Wtf8) }
696 pub fn make_ascii_lowercase(&mut self) {
697 self.bytes.make_ascii_lowercase()
701 pub fn make_ascii_uppercase(&mut self) {
702 self.bytes.make_ascii_uppercase()
706 pub fn to_ascii_lowercase(&self) -> Wtf8Buf {
707 Wtf8Buf { bytes: self.bytes.to_ascii_lowercase() }
711 pub fn to_ascii_uppercase(&self) -> Wtf8Buf {
712 Wtf8Buf { bytes: self.bytes.to_ascii_uppercase() }
716 pub fn is_ascii(&self) -> bool {
717 self.bytes.is_ascii()
721 pub fn eq_ignore_ascii_case(&self, other: &Self) -> bool {
722 self.bytes.eq_ignore_ascii_case(&other.bytes)
726 /// Returns a slice of the given string for the byte range \[`begin`..`end`).
730 /// Panics when `begin` and `end` do not point to code point boundaries,
731 /// or point beyond the end of the string.
732 impl ops::Index<ops::Range<usize>> for Wtf8 {
736 fn index(&self, range: ops::Range<usize>) -> &Wtf8 {
737 // is_code_point_boundary checks that the index is in [0, .len()]
738 if range.start <= range.end
739 && is_code_point_boundary(self, range.start)
740 && is_code_point_boundary(self, range.end)
742 unsafe { slice_unchecked(self, range.start, range.end) }
744 slice_error_fail(self, range.start, range.end)
749 /// Returns a slice of the given string from byte `begin` to its end.
753 /// Panics when `begin` is not at a code point boundary,
754 /// or is beyond the end of the string.
755 impl ops::Index<ops::RangeFrom<usize>> for Wtf8 {
759 fn index(&self, range: ops::RangeFrom<usize>) -> &Wtf8 {
760 // is_code_point_boundary checks that the index is in [0, .len()]
761 if is_code_point_boundary(self, range.start) {
762 unsafe { slice_unchecked(self, range.start, self.len()) }
764 slice_error_fail(self, range.start, self.len())
769 /// Returns a slice of the given string from its beginning to byte `end`.
773 /// Panics when `end` is not at a code point boundary,
774 /// or is beyond the end of the string.
775 impl ops::Index<ops::RangeTo<usize>> for Wtf8 {
779 fn index(&self, range: ops::RangeTo<usize>) -> &Wtf8 {
780 // is_code_point_boundary checks that the index is in [0, .len()]
781 if is_code_point_boundary(self, range.end) {
782 unsafe { slice_unchecked(self, 0, range.end) }
784 slice_error_fail(self, 0, range.end)
789 impl ops::Index<ops::RangeFull> for Wtf8 {
793 fn index(&self, _range: ops::RangeFull) -> &Wtf8 {
799 fn decode_surrogate(second_byte: u8, third_byte: u8) -> u16 {
800 // The first byte is assumed to be 0xED
801 0xD800 | (second_byte as u16 & 0x3F) << 6 | third_byte as u16 & 0x3F
805 fn decode_surrogate_pair(lead: u16, trail: u16) -> char {
806 let code_point = 0x10000 + ((((lead - 0xD800) as u32) << 10) | (trail - 0xDC00) as u32);
807 unsafe { char::from_u32_unchecked(code_point) }
810 /// Copied from core::str::StrPrelude::is_char_boundary
812 pub fn is_code_point_boundary(slice: &Wtf8, index: usize) -> bool {
813 if index == slice.len() {
816 match slice.bytes.get(index) {
818 Some(&b) => b < 128 || b >= 192,
822 /// Copied from core::str::raw::slice_unchecked
824 pub unsafe fn slice_unchecked(s: &Wtf8, begin: usize, end: usize) -> &Wtf8 {
825 // memory layout of a &[u8] and &Wtf8 are the same
826 Wtf8::from_bytes_unchecked(slice::from_raw_parts(s.bytes.as_ptr().add(begin), end - begin))
829 /// Copied from core::str::raw::slice_error_fail
831 pub fn slice_error_fail(s: &Wtf8, begin: usize, end: usize) -> ! {
832 assert!(begin <= end);
833 panic!("index {begin} and/or {end} in `{s:?}` do not lie on character boundary");
836 /// Iterator for the code points of a WTF-8 string.
838 /// Created with the method `.code_points()`.
840 pub struct Wtf8CodePoints<'a> {
841 bytes: slice::Iter<'a, u8>,
844 impl<'a> Iterator for Wtf8CodePoints<'a> {
845 type Item = CodePoint;
848 fn next(&mut self) -> Option<CodePoint> {
849 // SAFETY: `self.bytes` has been created from a WTF-8 string
850 unsafe { next_code_point(&mut self.bytes).map(|c| CodePoint { value: c }) }
854 fn size_hint(&self) -> (usize, Option<usize>) {
855 let len = self.bytes.len();
856 (len.saturating_add(3) / 4, Some(len))
860 /// Generates a wide character sequence for potentially ill-formed UTF-16.
861 #[stable(feature = "rust1", since = "1.0.0")]
863 pub struct EncodeWide<'a> {
864 code_points: Wtf8CodePoints<'a>,
868 // Copied from libunicode/u_str.rs
869 #[stable(feature = "rust1", since = "1.0.0")]
870 impl<'a> Iterator for EncodeWide<'a> {
874 fn next(&mut self) -> Option<u16> {
876 let tmp = self.extra;
881 let mut buf = [0; 2];
882 self.code_points.next().map(|code_point| {
883 let n = char::encode_utf16_raw(code_point.value, &mut buf).len();
892 fn size_hint(&self) -> (usize, Option<usize>) {
893 let (low, high) = self.code_points.size_hint();
894 let ext = (self.extra != 0) as usize;
895 // every code point gets either one u16 or two u16,
896 // so this iterator is between 1 or 2 times as
897 // long as the underlying iterator.
898 (low + ext, high.and_then(|n| n.checked_mul(2)).and_then(|n| n.checked_add(ext)))
902 #[stable(feature = "encode_wide_fused_iterator", since = "1.62.0")]
903 impl FusedIterator for EncodeWide<'_> {}
905 impl Hash for CodePoint {
907 fn hash<H: Hasher>(&self, state: &mut H) {
908 self.value.hash(state)
912 impl Hash for Wtf8Buf {
914 fn hash<H: Hasher>(&self, state: &mut H) {
915 state.write(&self.bytes);
922 fn hash<H: Hasher>(&self, state: &mut H) {
923 state.write(&self.bytes);