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 /// Returns the numeric value of the code point if it is a leading surrogate.
94 pub fn to_lead_surrogate(&self) -> Option<u16> {
96 lead @ 0xD800..=0xDBFF => Some(lead as u16),
101 /// Returns the numeric value of the code point if it is a trailing surrogate.
103 pub fn to_trail_surrogate(&self) -> Option<u16> {
105 trail @ 0xDC00..=0xDFFF => Some(trail as u16),
110 /// Optionally returns a Unicode scalar value for the code point.
112 /// Returns `None` if the code point is a surrogate (from U+D800 to U+DFFF).
114 pub fn to_char(&self) -> Option<char> {
116 0xD800..=0xDFFF => None,
117 _ => Some(unsafe { char::from_u32_unchecked(self.value) }),
121 /// Returns a Unicode scalar value for the code point.
123 /// Returns `'\u{FFFD}'` (the replacement character “�”)
124 /// if the code point is a surrogate (from U+D800 to U+DFFF).
126 pub fn to_char_lossy(&self) -> char {
127 self.to_char().unwrap_or('\u{FFFD}')
131 /// An owned, growable string of well-formed WTF-8 data.
133 /// Similar to `String`, but can additionally contain surrogate code points
134 /// if they’re not in a surrogate pair.
135 #[derive(Eq, PartialEq, Ord, PartialOrd, Clone)]
139 /// Do we know that `bytes` holds a valid UTF-8 encoding? We can easily
140 /// know this if we're constructed from a `String` or `&str`.
142 /// It is possible for `bytes` to have valid UTF-8 without this being
143 /// set, such as when we're concatenating `&Wtf8`'s and surrogates become
144 /// paired, as we don't bother to rescan the entire string.
148 impl ops::Deref for Wtf8Buf {
151 fn deref(&self) -> &Wtf8 {
156 impl ops::DerefMut for Wtf8Buf {
157 fn deref_mut(&mut self) -> &mut Wtf8 {
162 /// Format the string with double quotes,
163 /// and surrogates as `\u` followed by four hexadecimal digits.
164 /// Example: `"a\u{D800}"` for a string with code points [U+0061, U+D800]
165 impl fmt::Debug for Wtf8Buf {
167 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
168 fmt::Debug::fmt(&**self, formatter)
173 /// Creates a new, empty WTF-8 string.
175 pub fn new() -> Wtf8Buf {
176 Wtf8Buf { bytes: Vec::new(), is_known_utf8: true }
179 /// Creates a new, empty WTF-8 string with pre-allocated capacity for `capacity` bytes.
181 pub fn with_capacity(capacity: usize) -> Wtf8Buf {
182 Wtf8Buf { bytes: Vec::with_capacity(capacity), is_known_utf8: true }
185 /// Creates a WTF-8 string from a UTF-8 `String`.
187 /// This takes ownership of the `String` and does not copy.
189 /// Since WTF-8 is a superset of UTF-8, this always succeeds.
191 pub fn from_string(string: String) -> Wtf8Buf {
192 Wtf8Buf { bytes: string.into_bytes(), is_known_utf8: true }
195 /// Creates a WTF-8 string from a UTF-8 `&str` slice.
197 /// This copies the content of the slice.
199 /// Since WTF-8 is a superset of UTF-8, this always succeeds.
201 pub fn from_str(str: &str) -> Wtf8Buf {
202 Wtf8Buf { bytes: <[_]>::to_vec(str.as_bytes()), is_known_utf8: true }
205 pub fn clear(&mut self) {
207 self.is_known_utf8 = true;
210 /// Creates a WTF-8 string from a potentially ill-formed UTF-16 slice of 16-bit code units.
212 /// This is lossless: calling `.encode_wide()` on the resulting string
213 /// will always return the original code units.
214 pub fn from_wide(v: &[u16]) -> Wtf8Buf {
215 let mut string = Wtf8Buf::with_capacity(v.len());
216 for item in char::decode_utf16(v.iter().cloned()) {
218 Ok(ch) => string.push_char(ch),
220 let surrogate = surrogate.unpaired_surrogate();
221 // Surrogates are known to be in the code point range.
222 let code_point = unsafe { CodePoint::from_u32_unchecked(surrogate as u32) };
223 // The string will now contain an unpaired surrogate.
224 string.is_known_utf8 = false;
225 // Skip the WTF-8 concatenation check,
226 // surrogate pairs are already decoded by decode_utf16
227 string.push_code_point_unchecked(code_point);
234 /// Copied from String::push
235 /// This does **not** include the WTF-8 concatenation check or `is_known_utf8` check.
236 fn push_code_point_unchecked(&mut self, code_point: CodePoint) {
237 let mut bytes = [0; 4];
238 let bytes = char::encode_utf8_raw(code_point.value, &mut bytes);
239 self.bytes.extend_from_slice(bytes)
243 pub fn as_slice(&self) -> &Wtf8 {
244 unsafe { Wtf8::from_bytes_unchecked(&self.bytes) }
248 pub fn as_mut_slice(&mut self) -> &mut Wtf8 {
249 // Safety: `Wtf8` doesn't expose any way to mutate the bytes that would
250 // cause them to change from well-formed UTF-8 to ill-formed UTF-8,
251 // which would break the assumptions of the `is_known_utf8` field.
252 unsafe { Wtf8::from_mut_bytes_unchecked(&mut self.bytes) }
255 /// Reserves capacity for at least `additional` more bytes to be inserted
256 /// in the given `Wtf8Buf`.
257 /// The collection may reserve more space to avoid frequent reallocations.
261 /// Panics if the new capacity overflows `usize`.
263 pub fn reserve(&mut self, additional: usize) {
264 self.bytes.reserve(additional)
267 /// Tries to reserve capacity for at least `additional` more length units
268 /// in the given `Wtf8Buf`. The `Wtf8Buf` may reserve more space to avoid
269 /// frequent reallocations. After calling `try_reserve`, capacity will be
270 /// greater than or equal to `self.len() + additional`. Does nothing if
271 /// capacity is already sufficient. This method preserves the contents even
272 /// if an error occurs.
276 /// If the capacity overflows, or the allocator reports a failure, then an error
279 pub fn try_reserve(&mut self, additional: usize) -> Result<(), TryReserveError> {
280 self.bytes.try_reserve(additional)
284 pub fn reserve_exact(&mut self, additional: usize) {
285 self.bytes.reserve_exact(additional)
288 /// Tries to reserve the minimum capacity for exactly `additional`
289 /// length units in the given `Wtf8Buf`. After calling
290 /// `try_reserve_exact`, capacity will be greater than or equal to
291 /// `self.len() + additional` if it returns `Ok(())`.
292 /// Does nothing if the capacity is already sufficient.
294 /// Note that the allocator may give the `Wtf8Buf` more space than it
295 /// requests. Therefore, capacity can not be relied upon to be precisely
296 /// minimal. Prefer [`try_reserve`] if future insertions are expected.
298 /// [`try_reserve`]: Wtf8Buf::try_reserve
302 /// If the capacity overflows, or the allocator reports a failure, then an error
305 pub fn try_reserve_exact(&mut self, additional: usize) -> Result<(), TryReserveError> {
306 self.bytes.try_reserve_exact(additional)
310 pub fn shrink_to_fit(&mut self) {
311 self.bytes.shrink_to_fit()
315 pub fn shrink_to(&mut self, min_capacity: usize) {
316 self.bytes.shrink_to(min_capacity)
319 /// Returns the number of bytes that this string buffer can hold without reallocating.
321 pub fn capacity(&self) -> usize {
322 self.bytes.capacity()
325 /// Append a UTF-8 slice at the end of the string.
327 pub fn push_str(&mut self, other: &str) {
328 self.bytes.extend_from_slice(other.as_bytes())
331 /// Append a WTF-8 slice at the end of the string.
333 /// This replaces newly paired surrogates at the boundary
334 /// with a supplementary code point,
335 /// like concatenating ill-formed UTF-16 strings effectively would.
337 pub fn push_wtf8(&mut self, other: &Wtf8) {
338 match ((&*self).final_lead_surrogate(), other.initial_trail_surrogate()) {
339 // Replace newly paired surrogates by a supplementary code point.
340 (Some(lead), Some(trail)) => {
341 let len_without_lead_surrogate = self.len() - 3;
342 self.bytes.truncate(len_without_lead_surrogate);
343 let other_without_trail_surrogate = &other.bytes[3..];
344 // 4 bytes for the supplementary code point
345 self.bytes.reserve(4 + other_without_trail_surrogate.len());
346 self.push_char(decode_surrogate_pair(lead, trail));
347 self.bytes.extend_from_slice(other_without_trail_surrogate);
350 // If we'll be pushing a string containing a surrogate, we may
351 // no longer have UTF-8.
352 if other.next_surrogate(0).is_some() {
353 self.is_known_utf8 = false;
356 self.bytes.extend_from_slice(&other.bytes);
361 /// Append a Unicode scalar value at the end of the string.
363 pub fn push_char(&mut self, c: char) {
364 self.push_code_point_unchecked(CodePoint::from_char(c))
367 /// Append a code point at the end of the string.
369 /// This replaces newly paired surrogates at the boundary
370 /// with a supplementary code point,
371 /// like concatenating ill-formed UTF-16 strings effectively would.
373 pub fn push(&mut self, code_point: CodePoint) {
374 if let Some(trail) = code_point.to_trail_surrogate() {
375 if let Some(lead) = (&*self).final_lead_surrogate() {
376 let len_without_lead_surrogate = self.len() - 3;
377 self.bytes.truncate(len_without_lead_surrogate);
378 self.push_char(decode_surrogate_pair(lead, trail));
382 // We're pushing a trailing surrogate.
383 self.is_known_utf8 = false;
384 } else if code_point.to_lead_surrogate().is_some() {
385 // We're pushing a leading surrogate.
386 self.is_known_utf8 = false;
389 // No newly paired surrogates at the boundary.
390 self.push_code_point_unchecked(code_point)
393 /// Shortens a string to the specified length.
397 /// Panics if `new_len` > current length,
398 /// or if `new_len` is not a code point boundary.
400 pub fn truncate(&mut self, new_len: usize) {
401 assert!(is_code_point_boundary(self, new_len));
402 self.bytes.truncate(new_len)
405 /// Consumes the WTF-8 string and tries to convert it to UTF-8.
407 /// This does not copy the data.
409 /// If the contents are not well-formed UTF-8
410 /// (that is, if the string contains surrogates),
411 /// the original WTF-8 string is returned instead.
412 pub fn into_string(self) -> Result<String, Wtf8Buf> {
413 if self.is_known_utf8 || self.next_surrogate(0).is_none() {
414 Ok(unsafe { String::from_utf8_unchecked(self.bytes) })
420 /// Consumes the WTF-8 string and converts it lossily to UTF-8.
422 /// This does not copy the data (but may overwrite parts of it in place).
424 /// Surrogates are replaced with `"\u{FFFD}"` (the replacement character “�”)
425 pub fn into_string_lossy(mut self) -> String {
426 // Fast path: If we already have UTF-8, we can return it immediately.
427 if self.is_known_utf8 {
428 return unsafe { String::from_utf8_unchecked(self.bytes) };
433 match self.next_surrogate(pos) {
434 Some((surrogate_pos, _)) => {
435 pos = surrogate_pos + 3;
436 self.bytes[surrogate_pos..pos]
437 .copy_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
439 None => return unsafe { String::from_utf8_unchecked(self.bytes) },
444 /// Converts this `Wtf8Buf` into a boxed `Wtf8`.
446 pub fn into_box(self) -> Box<Wtf8> {
447 unsafe { mem::transmute(self.bytes.into_boxed_slice()) }
450 /// Converts a `Box<Wtf8>` into a `Wtf8Buf`.
451 pub fn from_box(boxed: Box<Wtf8>) -> Wtf8Buf {
452 let bytes: Box<[u8]> = unsafe { mem::transmute(boxed) };
453 Wtf8Buf { bytes: bytes.into_vec(), is_known_utf8: false }
457 /// Creates a new WTF-8 string from an iterator of code points.
459 /// This replaces surrogate code point pairs with supplementary code points,
460 /// like concatenating ill-formed UTF-16 strings effectively would.
461 impl FromIterator<CodePoint> for Wtf8Buf {
462 fn from_iter<T: IntoIterator<Item = CodePoint>>(iter: T) -> Wtf8Buf {
463 let mut string = Wtf8Buf::new();
469 /// Append code points from an iterator to the string.
471 /// This replaces surrogate code point pairs with supplementary code points,
472 /// like concatenating ill-formed UTF-16 strings effectively would.
473 impl Extend<CodePoint> for Wtf8Buf {
474 fn extend<T: IntoIterator<Item = CodePoint>>(&mut self, iter: T) {
475 let iterator = iter.into_iter();
476 let (low, _high) = iterator.size_hint();
477 // Lower bound of one byte per code point (ASCII only)
478 self.bytes.reserve(low);
479 iterator.for_each(move |code_point| self.push(code_point));
483 fn extend_one(&mut self, code_point: CodePoint) {
484 self.push(code_point);
488 fn extend_reserve(&mut self, additional: usize) {
489 // Lower bound of one byte per code point (ASCII only)
490 self.bytes.reserve(additional);
494 /// A borrowed slice of well-formed WTF-8 data.
496 /// Similar to `&str`, but can additionally contain surrogate code points
497 /// if they’re not in a surrogate pair.
498 #[derive(Eq, Ord, PartialEq, PartialOrd)]
503 impl AsInner<[u8]> for Wtf8 {
504 fn as_inner(&self) -> &[u8] {
509 /// Format the slice with double quotes,
510 /// and surrogates as `\u` followed by four hexadecimal digits.
511 /// Example: `"a\u{D800}"` for a slice with code points [U+0061, U+D800]
512 impl fmt::Debug for Wtf8 {
513 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
514 fn write_str_escaped(f: &mut fmt::Formatter<'_>, s: &str) -> fmt::Result {
515 use crate::fmt::Write;
516 for c in s.chars().flat_map(|c| c.escape_debug()) {
522 formatter.write_str("\"")?;
524 while let Some((surrogate_pos, surrogate)) = self.next_surrogate(pos) {
525 write_str_escaped(formatter, unsafe {
526 str::from_utf8_unchecked(&self.bytes[pos..surrogate_pos])
528 write!(formatter, "\\u{{{:x}}}", surrogate)?;
529 pos = surrogate_pos + 3;
531 write_str_escaped(formatter, unsafe { str::from_utf8_unchecked(&self.bytes[pos..]) })?;
532 formatter.write_str("\"")
536 impl fmt::Display for Wtf8 {
537 fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
538 let wtf8_bytes = &self.bytes;
541 match self.next_surrogate(pos) {
542 Some((surrogate_pos, _)) => {
543 formatter.write_str(unsafe {
544 str::from_utf8_unchecked(&wtf8_bytes[pos..surrogate_pos])
546 formatter.write_str(UTF8_REPLACEMENT_CHARACTER)?;
547 pos = surrogate_pos + 3;
550 let s = unsafe { str::from_utf8_unchecked(&wtf8_bytes[pos..]) };
551 if pos == 0 { return s.fmt(formatter) } else { return formatter.write_str(s) }
559 /// Creates a WTF-8 slice from a UTF-8 `&str` slice.
561 /// Since WTF-8 is a superset of UTF-8, this always succeeds.
563 pub fn from_str(value: &str) -> &Wtf8 {
564 unsafe { Wtf8::from_bytes_unchecked(value.as_bytes()) }
567 /// Creates a WTF-8 slice from a WTF-8 byte slice.
569 /// Since the byte slice is not checked for valid WTF-8, this functions is
572 unsafe fn from_bytes_unchecked(value: &[u8]) -> &Wtf8 {
573 mem::transmute(value)
576 /// Creates a mutable WTF-8 slice from a mutable WTF-8 byte slice.
578 /// Since the byte slice is not checked for valid WTF-8, this functions is
581 unsafe fn from_mut_bytes_unchecked(value: &mut [u8]) -> &mut Wtf8 {
582 mem::transmute(value)
585 /// Returns the length, in WTF-8 bytes.
587 pub fn len(&self) -> usize {
592 pub fn is_empty(&self) -> bool {
593 self.bytes.is_empty()
596 /// Returns the code point at `position` if it is in the ASCII range,
597 /// or `b'\xFF' otherwise.
601 /// Panics if `position` is beyond the end of the string.
603 pub fn ascii_byte_at(&self, position: usize) -> u8 {
604 match self.bytes[position] {
605 ascii_byte @ 0x00..=0x7F => ascii_byte,
610 /// Returns an iterator for the string’s code points.
612 pub fn code_points(&self) -> Wtf8CodePoints<'_> {
613 Wtf8CodePoints { bytes: self.bytes.iter() }
616 /// Tries to convert the string to UTF-8 and return a `&str` slice.
618 /// Returns `None` if the string contains surrogates.
620 /// This does not copy the data.
622 pub fn as_str(&self) -> Option<&str> {
623 // Well-formed WTF-8 is also well-formed UTF-8
624 // if and only if it contains no surrogate.
625 match self.next_surrogate(0) {
626 None => Some(unsafe { str::from_utf8_unchecked(&self.bytes) }),
631 /// Creates an owned `Wtf8Buf` from a borrowed `Wtf8`.
632 pub fn to_owned(&self) -> Wtf8Buf {
633 Wtf8Buf { bytes: self.bytes.to_vec(), is_known_utf8: false }
636 /// Lossily converts the string to UTF-8.
637 /// Returns a UTF-8 `&str` slice if the contents are well-formed in UTF-8.
639 /// Surrogates are replaced with `"\u{FFFD}"` (the replacement character “�”).
641 /// This only copies the data if necessary (if it contains any surrogate).
642 pub fn to_string_lossy(&self) -> Cow<'_, str> {
643 let surrogate_pos = match self.next_surrogate(0) {
644 None => return Cow::Borrowed(unsafe { str::from_utf8_unchecked(&self.bytes) }),
645 Some((pos, _)) => pos,
647 let wtf8_bytes = &self.bytes;
648 let mut utf8_bytes = Vec::with_capacity(self.len());
649 utf8_bytes.extend_from_slice(&wtf8_bytes[..surrogate_pos]);
650 utf8_bytes.extend_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
651 let mut pos = surrogate_pos + 3;
653 match self.next_surrogate(pos) {
654 Some((surrogate_pos, _)) => {
655 utf8_bytes.extend_from_slice(&wtf8_bytes[pos..surrogate_pos]);
656 utf8_bytes.extend_from_slice(UTF8_REPLACEMENT_CHARACTER.as_bytes());
657 pos = surrogate_pos + 3;
660 utf8_bytes.extend_from_slice(&wtf8_bytes[pos..]);
661 return Cow::Owned(unsafe { String::from_utf8_unchecked(utf8_bytes) });
667 /// Converts the WTF-8 string to potentially ill-formed UTF-16
668 /// and return an iterator of 16-bit code units.
670 /// This is lossless:
671 /// calling `Wtf8Buf::from_ill_formed_utf16` on the resulting code units
672 /// would always return the original WTF-8 string.
674 pub fn encode_wide(&self) -> EncodeWide<'_> {
675 EncodeWide { code_points: self.code_points(), extra: 0 }
679 fn next_surrogate(&self, mut pos: usize) -> Option<(usize, u16)> {
680 let mut iter = self.bytes[pos..].iter();
682 let b = *iter.next()?;
688 } else if b == 0xED {
689 match (iter.next(), iter.next()) {
690 (Some(&b2), Some(&b3)) if b2 >= 0xA0 => {
691 return Some((pos, decode_surrogate(b2, b3)));
709 fn final_lead_surrogate(&self) -> Option<u16> {
711 [.., 0xED, b2 @ 0xA0..=0xAF, b3] => Some(decode_surrogate(b2, b3)),
717 fn initial_trail_surrogate(&self) -> Option<u16> {
719 [0xED, b2 @ 0xB0..=0xBF, b3, ..] => Some(decode_surrogate(b2, b3)),
724 pub fn clone_into(&self, buf: &mut Wtf8Buf) {
725 buf.is_known_utf8 = false;
726 self.bytes.clone_into(&mut buf.bytes);
729 /// Boxes this `Wtf8`.
731 pub fn into_box(&self) -> Box<Wtf8> {
732 let boxed: Box<[u8]> = self.bytes.into();
733 unsafe { mem::transmute(boxed) }
736 /// Creates a boxed, empty `Wtf8`.
737 pub fn empty_box() -> Box<Wtf8> {
738 let boxed: Box<[u8]> = Default::default();
739 unsafe { mem::transmute(boxed) }
743 pub fn into_arc(&self) -> Arc<Wtf8> {
744 let arc: Arc<[u8]> = Arc::from(&self.bytes);
745 unsafe { Arc::from_raw(Arc::into_raw(arc) as *const Wtf8) }
749 pub fn into_rc(&self) -> Rc<Wtf8> {
750 let rc: Rc<[u8]> = Rc::from(&self.bytes);
751 unsafe { Rc::from_raw(Rc::into_raw(rc) as *const Wtf8) }
755 pub fn make_ascii_lowercase(&mut self) {
756 self.bytes.make_ascii_lowercase()
760 pub fn make_ascii_uppercase(&mut self) {
761 self.bytes.make_ascii_uppercase()
765 pub fn to_ascii_lowercase(&self) -> Wtf8Buf {
766 Wtf8Buf { bytes: self.bytes.to_ascii_lowercase(), is_known_utf8: false }
770 pub fn to_ascii_uppercase(&self) -> Wtf8Buf {
771 Wtf8Buf { bytes: self.bytes.to_ascii_uppercase(), is_known_utf8: false }
775 pub fn is_ascii(&self) -> bool {
776 self.bytes.is_ascii()
780 pub fn eq_ignore_ascii_case(&self, other: &Self) -> bool {
781 self.bytes.eq_ignore_ascii_case(&other.bytes)
785 /// Returns a slice of the given string for the byte range \[`begin`..`end`).
789 /// Panics when `begin` and `end` do not point to code point boundaries,
790 /// or point beyond the end of the string.
791 impl ops::Index<ops::Range<usize>> for Wtf8 {
795 fn index(&self, range: ops::Range<usize>) -> &Wtf8 {
796 // is_code_point_boundary checks that the index is in [0, .len()]
797 if range.start <= range.end
798 && is_code_point_boundary(self, range.start)
799 && is_code_point_boundary(self, range.end)
801 unsafe { slice_unchecked(self, range.start, range.end) }
803 slice_error_fail(self, range.start, range.end)
808 /// Returns a slice of the given string from byte `begin` to its end.
812 /// Panics when `begin` is not at a code point boundary,
813 /// or is beyond the end of the string.
814 impl ops::Index<ops::RangeFrom<usize>> for Wtf8 {
818 fn index(&self, range: ops::RangeFrom<usize>) -> &Wtf8 {
819 // is_code_point_boundary checks that the index is in [0, .len()]
820 if is_code_point_boundary(self, range.start) {
821 unsafe { slice_unchecked(self, range.start, self.len()) }
823 slice_error_fail(self, range.start, self.len())
828 /// Returns a slice of the given string from its beginning to byte `end`.
832 /// Panics when `end` is not at a code point boundary,
833 /// or is beyond the end of the string.
834 impl ops::Index<ops::RangeTo<usize>> for Wtf8 {
838 fn index(&self, range: ops::RangeTo<usize>) -> &Wtf8 {
839 // is_code_point_boundary checks that the index is in [0, .len()]
840 if is_code_point_boundary(self, range.end) {
841 unsafe { slice_unchecked(self, 0, range.end) }
843 slice_error_fail(self, 0, range.end)
848 impl ops::Index<ops::RangeFull> for Wtf8 {
852 fn index(&self, _range: ops::RangeFull) -> &Wtf8 {
858 fn decode_surrogate(second_byte: u8, third_byte: u8) -> u16 {
859 // The first byte is assumed to be 0xED
860 0xD800 | (second_byte as u16 & 0x3F) << 6 | third_byte as u16 & 0x3F
864 fn decode_surrogate_pair(lead: u16, trail: u16) -> char {
865 let code_point = 0x10000 + ((((lead - 0xD800) as u32) << 10) | (trail - 0xDC00) as u32);
866 unsafe { char::from_u32_unchecked(code_point) }
869 /// Copied from core::str::StrPrelude::is_char_boundary
871 pub fn is_code_point_boundary(slice: &Wtf8, index: usize) -> bool {
872 if index == slice.len() {
875 match slice.bytes.get(index) {
877 Some(&b) => b < 128 || b >= 192,
881 /// Copied from core::str::raw::slice_unchecked
883 pub unsafe fn slice_unchecked(s: &Wtf8, begin: usize, end: usize) -> &Wtf8 {
884 // memory layout of a &[u8] and &Wtf8 are the same
885 Wtf8::from_bytes_unchecked(slice::from_raw_parts(s.bytes.as_ptr().add(begin), end - begin))
888 /// Copied from core::str::raw::slice_error_fail
890 pub fn slice_error_fail(s: &Wtf8, begin: usize, end: usize) -> ! {
891 assert!(begin <= end);
892 panic!("index {begin} and/or {end} in `{s:?}` do not lie on character boundary");
895 /// Iterator for the code points of a WTF-8 string.
897 /// Created with the method `.code_points()`.
899 pub struct Wtf8CodePoints<'a> {
900 bytes: slice::Iter<'a, u8>,
903 impl<'a> Iterator for Wtf8CodePoints<'a> {
904 type Item = CodePoint;
907 fn next(&mut self) -> Option<CodePoint> {
908 // SAFETY: `self.bytes` has been created from a WTF-8 string
909 unsafe { next_code_point(&mut self.bytes).map(|c| CodePoint { value: c }) }
913 fn size_hint(&self) -> (usize, Option<usize>) {
914 let len = self.bytes.len();
915 (len.saturating_add(3) / 4, Some(len))
919 /// Generates a wide character sequence for potentially ill-formed UTF-16.
920 #[stable(feature = "rust1", since = "1.0.0")]
922 pub struct EncodeWide<'a> {
923 code_points: Wtf8CodePoints<'a>,
927 // Copied from libunicode/u_str.rs
928 #[stable(feature = "rust1", since = "1.0.0")]
929 impl<'a> Iterator for EncodeWide<'a> {
933 fn next(&mut self) -> Option<u16> {
935 let tmp = self.extra;
940 let mut buf = [0; 2];
941 self.code_points.next().map(|code_point| {
942 let n = char::encode_utf16_raw(code_point.value, &mut buf).len();
951 fn size_hint(&self) -> (usize, Option<usize>) {
952 let (low, high) = self.code_points.size_hint();
953 let ext = (self.extra != 0) as usize;
954 // every code point gets either one u16 or two u16,
955 // so this iterator is between 1 or 2 times as
956 // long as the underlying iterator.
957 (low + ext, high.and_then(|n| n.checked_mul(2)).and_then(|n| n.checked_add(ext)))
961 #[stable(feature = "encode_wide_fused_iterator", since = "1.62.0")]
962 impl FusedIterator for EncodeWide<'_> {}
964 impl Hash for CodePoint {
966 fn hash<H: Hasher>(&self, state: &mut H) {
967 self.value.hash(state)
971 impl Hash for Wtf8Buf {
973 fn hash<H: Hasher>(&self, state: &mut H) {
974 state.write(&self.bytes);
981 fn hash<H: Hasher>(&self, state: &mut H) {
982 state.write(&self.bytes);