Untersuchungsergebnis.rs Download desUnknown {[0] [0] [0]}zum Wurzelverzeichnis wechseln
// Copyright Mozilla Foundation. See the COPYRIGHT
// file at the top-level directory of this distribution.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
//
https://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or
https://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.
//! Functions for converting between different in-RAM representations of text
//! and for quickly checking if the Unicode Bidirectional Algorithm can be
//! avoided.
//!
//! By using slices for output, the functions here seek to enable by-register
//! (ALU register or SIMD register as available) operations in order to
//! outperform iterator-based conversions available in the Rust standard
//! library.
//!
//! _Note:_ "Latin1" in this module refers to the Unicode range from U+0000 to
//! U+00FF, inclusive, and does not refer to the windows-1252 range. This
//! in-memory encoding is sometimes used as a storage optimization of text
//! when UTF-16 indexing and length semantics are exposed.
//!
//! The FFI binding for this module are in the
//! [encoding_c_mem crate](
https://github.com/hsivonen/encoding_c_mem).
#[cfg(feature = "alloc")]
use alloc::borrow::Cow;
#[cfg(feature = "alloc")]
use alloc::string::String;
#[cfg(feature = "alloc")]
use alloc::vec::Vec;
use super::in_inclusive_range16;
use super::in_inclusive_range32;
use super::in_inclusive_range8;
use super::in_range16;
use super::in_range32;
use super::DecoderResult;
use crate::ascii::*;
use crate::utf_8::*;
macro_rules! non_fuzz_debug_assert {
($($arg:tt)*) => (if !cfg!(fuzzing) { debug_assert!($($arg)*); })
}
cfg_if! {
if #[cfg(feature = "simd-accel")] {
use ::core::intrinsics::likely;
use ::core::intrinsics::unlikely;
} else {
#[inline(always)]
fn likely(b: bool) -> bool {
b
}
#[inline(always)]
fn unlikely(b: bool) -> bool {
b
}
}
}
/// Classification of text as Latin1 (all code points are below U+0100),
/// left-to-right with some non-Latin1 characters or as containing at least
/// some right-to-left characters.
#[must_use]
#[derive(Debug, PartialEq, Eq)]
#[repr(C)]
pub enum Latin1Bidi {
/// Every character is below U+0100.
Latin1 = 0,
/// There is at least one character that's U+0100 or higher, but there
/// are no right-to-left characters.
LeftToRight = 1,
/// There is at least one right-to-left character.
Bidi = 2,
}
// `as` truncates, so works on 32-bit, too.
#[allow(dead_code)]
const LATIN1_MASK: usize = 0xFF00_FF00_FF00_FF00u64 as usize;
#[allow(unused_macros)]
macro_rules! by_unit_check_alu {
($name:ident, $unit:ty, $bound:expr, $mask:ident) => {
#[cfg_attr(feature = "cargo-clippy", allow(cast_ptr_alignment))]
#[inline(always)]
fn $name(buffer: &[$unit]) -> bool {
let mut offset = 0usize;
let mut accu = 0usize;
let unit_size = ::core::mem::size_of::<$unit>();
let len = buffer.len();
if len >= ALU_ALIGNMENT / unit_size {
// The most common reason to return `false` is for the first code
// unit to fail the test, so check that first.
if buffer[0] >= $bound {
return false;
}
let src = buffer.as_ptr();
let mut until_alignment = ((ALU_ALIGNMENT - ((src as usize) & ALU_ALIGNMENT_MASK))
& ALU_ALIGNMENT_MASK)
/ unit_size;
if until_alignment + ALU_ALIGNMENT / unit_size <= len {
if until_alignment != 0 {
accu |= buffer[offset] as usize;
offset += 1;
until_alignment -= 1;
while until_alignment != 0 {
accu |= buffer[offset] as usize;
offset += 1;
until_alignment -= 1;
}
if accu >= $bound {
return false;
}
}
let len_minus_stride = len - ALU_ALIGNMENT / unit_size;
if offset + (4 * (ALU_ALIGNMENT / unit_size)) <= len {
// Safety: the above check lets us perform 4 consecutive reads of
// length ALU_ALIGNMENT / unit_size. ALU_ALIGNMENT is the size of usize, and unit_size
// is the size of the `src` pointer, so this is equal to performing four usize reads.
//
// This invariant is upheld on all loop iterations
let len_minus_unroll = len - (4 * (ALU_ALIGNMENT / unit_size));
loop {
let unroll_accu = unsafe { *(src.add(offset) as *const usize) }
| unsafe {
*(src.add(offset + (ALU_ALIGNMENT / unit_size)) as *const usize)
}
| unsafe {
*(src.add(offset + (2 * (ALU_ALIGNMENT / unit_size)))
as *const usize)
}
| unsafe {
*(src.add(offset + (3 * (ALU_ALIGNMENT / unit_size)))
as *const usize)
};
if unroll_accu & $mask != 0 {
return false;
}
offset += 4 * (ALU_ALIGNMENT / unit_size);
// Safety: this check lets us continue to perform the 4 reads earlier
if offset > len_minus_unroll {
break;
}
}
}
while offset <= len_minus_stride {
// Safety: the above check lets us perform one usize read.
accu |= unsafe { *(src.add(offset) as *const usize) };
offset += ALU_ALIGNMENT / unit_size;
}
}
}
for &unit in &buffer[offset..] {
accu |= unit as usize;
}
accu & $mask == 0
}
};
}
#[allow(unused_macros)]
macro_rules! by_unit_check_simd {
($name:ident, $unit:ty, $splat:expr, $simd_ty:ty, $bound:expr, $func:ident) => {
#[inline(always)]
fn $name(buffer: &[$unit]) -> bool {
let mut offset = 0usize;
let mut accu = 0usize;
let unit_size = ::core::mem::size_of::<$unit>();
let len = buffer.len();
if len >= SIMD_STRIDE_SIZE / unit_size {
// The most common reason to return `false` is for the first code
// unit to fail the test, so check that first.
if buffer[0] >= $bound {
return false;
}
let src = buffer.as_ptr();
let mut until_alignment = ((SIMD_ALIGNMENT
- ((src as usize) & SIMD_ALIGNMENT_MASK))
& SIMD_ALIGNMENT_MASK)
/ unit_size;
if until_alignment + SIMD_STRIDE_SIZE / unit_size <= len {
if until_alignment != 0 {
accu |= buffer[offset] as usize;
offset += 1;
until_alignment -= 1;
while until_alignment != 0 {
accu |= buffer[offset] as usize;
offset += 1;
until_alignment -= 1;
}
if accu >= $bound {
return false;
}
}
let len_minus_stride = len - SIMD_STRIDE_SIZE / unit_size;
if offset + (4 * (SIMD_STRIDE_SIZE / unit_size)) <= len {
// Safety: the above check lets us perform 4 consecutive reads of
// length SIMD_STRIDE_SIZE / unit_size. SIMD_STRIDE_SIZE is the size of $simd_ty, and unit_s
ize
// is the size of the `src` pointer, so this is equal to performing four $simd_ty reads.
//
// This invariant is upheld on all loop iterations
let len_minus_unroll = len - (4 * (SIMD_STRIDE_SIZE / unit_size));
loop {
let unroll_accu = unsafe { *(src.add(offset) as *const $simd_ty) }
| unsafe {
*(src.add(offset + (SIMD_STRIDE_SIZE / unit_size))
as *const $simd_ty)
}
| unsafe {
*(src.add(offset + (2 * (SIMD_STRIDE_SIZE / unit_size)))
as *const $simd_ty)
}
| unsafe {
*(src.add(offset + (3 * (SIMD_STRIDE_SIZE / unit_size)))
as *const $simd_ty)
};
if !$func(unroll_accu) {
return false;
}
offset += 4 * (SIMD_STRIDE_SIZE / unit_size);
// Safety: this check lets us continue to perform the 4 reads earlier
if offset > len_minus_unroll {
break;
}
}
}
let mut simd_accu = $splat;
while offset <= len_minus_stride {
// Safety: the above check lets us perform one $simd_ty read.
simd_accu = simd_accu | unsafe { *(src.add(offset) as *const $simd_ty) };
offset += SIMD_STRIDE_SIZE / unit_size;
}
if !$func(simd_accu) {
return false;
}
}
}
for &unit in &buffer[offset..] {
accu |= unit as usize;
}
accu < $bound
}
};
}
cfg_if! {
if #[cfg(all(feature = "simd-accel", any(target_feature = "sse2", all(target_endian = "little", target_arch = "aarch64"), all(target_endian = "little", target_feature = "neon"))))] {
use crate::simd_funcs::*;
use core::simd::u8x16;
use core::simd::u16x8;
const SIMD_ALIGNMENT: usize = 16;
const SIMD_ALIGNMENT_MASK: usize = 15;
by_unit_check_simd!(is_ascii_impl, u8, u8x16::splat(0), u8x16, 0x80, simd_is_ascii);
by_unit_check_simd!(is_basic_latin_impl, u16, u16x8::splat(0), u16x8, 0x80, simd_is_basic_latin);
by_unit_check_simd!(is_utf16_latin1_impl, u16, u16x8::splat(0), u16x8, 0x100, simd_is_latin1);
#[inline(always)]
fn utf16_valid_up_to_impl(buffer: &[u16]) -> usize {
// This function is a mess, because it simultaneously tries to do
// only aligned SIMD (perhaps misguidedly) and needs to deal with
// the last code unit in a SIMD stride being part of a valid
// surrogate pair.
let unit_size = ::core::mem::size_of::<u16>();
let src = buffer.as_ptr();
let len = buffer.len();
let mut offset = 0usize;
'outer: loop {
let until_alignment = ((SIMD_ALIGNMENT - ((unsafe { src.add(offset) } as usize) & SIMD_ALIGNMENT_MASK)) &
SIMD_ALIGNMENT_MASK) / unit_size;
if until_alignment == 0 {
if offset + SIMD_STRIDE_SIZE / unit_size > len {
break;
}
} else {
let offset_plus_until_alignment = offset + until_alignment;
let offset_plus_until_alignment_plus_one = offset_plus_until_alignment + 1;
if offset_plus_until_alignment_plus_one + SIMD_STRIDE_SIZE / unit_size > len {
break;
}
let (up_to, last_valid_low) = utf16_valid_up_to_alu(&buffer[offset..offset_plus_until_alignment_plus_one]);
if up_to < until_alignment {
return offset + up_to;
}
if last_valid_low {
offset = offset_plus_until_alignment_plus_one;
continue;
}
offset = offset_plus_until_alignment;
}
let len_minus_stride = len - SIMD_STRIDE_SIZE / unit_size;
loop {
let offset_plus_stride = offset + SIMD_STRIDE_SIZE / unit_size;
if contains_surrogates(unsafe { *(src.add(offset) as *const u16x8) }) {
if offset_plus_stride == len {
break 'outer;
}
let offset_plus_stride_plus_one = offset_plus_stride + 1;
let (up_to, last_valid_low) = utf16_valid_up_to_alu(&buffer[offset..offset_plus_stride_plus_one]);
if up_to < SIMD_STRIDE_SIZE / unit_size {
return offset + up_to;
}
if last_valid_low {
offset = offset_plus_stride_plus_one;
continue 'outer;
}
}
offset = offset_plus_stride;
if offset > len_minus_stride {
break 'outer;
}
}
}
let (up_to, _) = utf16_valid_up_to_alu(&buffer[offset..]);
offset + up_to
}
} else {
by_unit_check_alu!(is_ascii_impl, u8, 0x80, ASCII_MASK);
by_unit_check_alu!(is_basic_latin_impl, u16, 0x80, BASIC_LATIN_MASK);
by_unit_check_alu!(is_utf16_latin1_impl, u16, 0x100, LATIN1_MASK);
#[inline(always)]
fn utf16_valid_up_to_impl(buffer: &[u16]) -> usize {
let (up_to, _) = utf16_valid_up_to_alu(buffer);
up_to
}
}
}
/// The second return value is true iff the last code unit of the slice was
/// reached and turned out to be a low surrogate that is part of a valid pair.
#[cfg_attr(feature = "cargo-clippy", allow(collapsible_if))]
#[inline(always)]
fn utf16_valid_up_to_alu(buffer: &[u16]) -> (usize, bool) {
let len = buffer.len();
if len == 0 {
return (0, false);
}
let mut offset = 0usize;
loop {
let unit = buffer[offset];
let next = offset + 1;
let unit_minus_surrogate_start = unit.wrapping_sub(0xD800);
if unit_minus_surrogate_start > (0xDFFF - 0xD800) {
// Not a surrogate
offset = next;
if offset == len {
return (offset, false);
}
continue;
}
if unit_minus_surrogate_start <= (0xDBFF - 0xD800) {
// high surrogate
if next < len {
let second = buffer[next];
let second_minus_low_surrogate_start = second.wrapping_sub(0xDC00);
if second_minus_low_surrogate_start <= (0xDFFF - 0xDC00) {
// The next code unit is a low surrogate. Advance position.
offset = next + 1;
if offset == len {
return (offset, true);
}
continue;
}
// The next code unit is not a low surrogate. Don't advance
// position and treat the high surrogate as unpaired.
// fall through
}
// Unpaired, fall through
}
// Unpaired surrogate
return (offset, false);
}
}
cfg_if! {
if #[cfg(all(feature = "simd-accel", any(target_feature = "sse2", all(target_endian = "little", target_arch = "aarch64"), all(target_endian = "little", target_feature = "neon"))))] {
#[inline(always)]
fn is_str_latin1_impl(buffer: &str) -> Option<usize> {
let mut offset = 0usize;
let bytes = buffer.as_bytes();
let len = bytes.len();
if len >= SIMD_STRIDE_SIZE {
let src = bytes.as_ptr();
let mut until_alignment = (SIMD_ALIGNMENT - ((src as usize) & SIMD_ALIGNMENT_MASK)) &
SIMD_ALIGNMENT_MASK;
if until_alignment + SIMD_STRIDE_SIZE <= len {
while until_alignment != 0 {
if bytes[offset] > 0xC3 {
return Some(offset);
}
offset += 1;
until_alignment -= 1;
}
let len_minus_stride = len - SIMD_STRIDE_SIZE;
loop {
if !simd_is_str_latin1(unsafe { *(src.add(offset) as *const u8x16) }) {
// TODO: Ensure this compiles away when inlined into `is_str_latin1()`.
while bytes[offset] & 0xC0 == 0x80 {
offset += 1;
}
return Some(offset);
}
offset += SIMD_STRIDE_SIZE;
if offset > len_minus_stride {
break;
}
}
}
}
for i in offset..len {
if bytes[i] > 0xC3 {
return Some(i);
}
}
None
}
} else {
#[inline(always)]
fn is_str_latin1_impl(buffer: &str) -> Option<usize> {
let mut bytes = buffer.as_bytes();
let mut total = 0;
loop {
if let Some((byte, offset)) = validate_ascii(bytes) {
total += offset;
if byte > 0xC3 {
return Some(total);
}
bytes = &bytes[offset + 2..];
total += 2;
} else {
return None;
}
}
}
}
}
#[inline(always)]
fn is_utf8_latin1_impl(buffer: &[u8]) -> Option<usize> {
let mut bytes = buffer;
let mut total = 0;
loop {
if let Some((byte, offset)) = validate_ascii(bytes) {
total += offset;
if in_inclusive_range8(byte, 0xC2, 0xC3) {
let next = offset + 1;
if next == bytes.len() {
return Some(total);
}
if bytes[next] & 0xC0 != 0x80 {
return Some(total);
}
bytes = &bytes[offset + 2..];
total += 2;
} else {
return Some(total);
}
} else {
return None;
}
}
}
cfg_if! {
if #[cfg(all(feature = "simd-accel", any(target_feature = "sse2", all(target_endian = "little", target_arch = "aarch64"), all(target_endian = "little", target_feature = "neon"))))] {
#[inline(always)]
fn is_utf16_bidi_impl(buffer: &[u16]) -> bool {
let mut offset = 0usize;
let len = buffer.len();
if len >= SIMD_STRIDE_SIZE / 2 {
let src = buffer.as_ptr();
let mut until_alignment = ((SIMD_ALIGNMENT - ((src as usize) & SIMD_ALIGNMENT_MASK)) &
SIMD_ALIGNMENT_MASK) / 2;
if until_alignment + (SIMD_STRIDE_SIZE / 2) <= len {
while until_alignment != 0 {
if is_utf16_code_unit_bidi(buffer[offset]) {
return true;
}
offset += 1;
until_alignment -= 1;
}
let len_minus_stride = len - (SIMD_STRIDE_SIZE / 2);
loop {
if is_u16x8_bidi(unsafe { *(src.add(offset) as *const u16x8) }) {
return true;
}
offset += SIMD_STRIDE_SIZE / 2;
if offset > len_minus_stride {
break;
}
}
}
}
for &u in &buffer[offset..] {
if is_utf16_code_unit_bidi(u) {
return true;
}
}
false
}
} else {
#[inline(always)]
fn is_utf16_bidi_impl(buffer: &[u16]) -> bool {
for &u in buffer {
if is_utf16_code_unit_bidi(u) {
return true;
}
}
false
}
}
}
cfg_if! {
if #[cfg(all(feature = "simd-accel", any(target_feature = "sse2", all(target_endian = "little", target_arch = "aarch64"), all(target_endian = "little", target_feature = "neon"))))] {
#[inline(always)]
fn check_utf16_for_latin1_and_bidi_impl(buffer: &[u16]) -> Latin1Bidi {
let mut offset = 0usize;
let len = buffer.len();
if len >= SIMD_STRIDE_SIZE / 2 {
let src = buffer.as_ptr();
let mut until_alignment = ((SIMD_ALIGNMENT - ((src as usize) & SIMD_ALIGNMENT_MASK)) &
SIMD_ALIGNMENT_MASK) / 2;
if until_alignment + (SIMD_STRIDE_SIZE / 2) <= len {
while until_alignment != 0 {
if buffer[offset] > 0xFF {
// This transition isn't optimal, since the aligment is recomputing
// but not tweaking further today.
if is_utf16_bidi_impl(&buffer[offset..]) {
return Latin1Bidi::Bidi;
}
return Latin1Bidi::LeftToRight;
}
offset += 1;
until_alignment -= 1;
}
let len_minus_stride = len - (SIMD_STRIDE_SIZE / 2);
loop {
let mut s = unsafe { *(src.add(offset) as *const u16x8) };
if !simd_is_latin1(s) {
loop {
if is_u16x8_bidi(s) {
return Latin1Bidi::Bidi;
}
offset += SIMD_STRIDE_SIZE / 2;
if offset > len_minus_stride {
for &u in &buffer[offset..] {
if is_utf16_code_unit_bidi(u) {
return Latin1Bidi::Bidi;
}
}
return Latin1Bidi::LeftToRight;
}
s = unsafe { *(src.add(offset) as *const u16x8) };
}
}
offset += SIMD_STRIDE_SIZE / 2;
if offset > len_minus_stride {
break;
}
}
}
}
let mut iter = (&buffer[offset..]).iter();
loop {
if let Some(&u) = iter.next() {
if u > 0xFF {
let mut inner_u = u;
loop {
if is_utf16_code_unit_bidi(inner_u) {
return Latin1Bidi::Bidi;
}
if let Some(&code_unit) = iter.next() {
inner_u = code_unit;
} else {
return Latin1Bidi::LeftToRight;
}
}
}
} else {
return Latin1Bidi::Latin1;
}
}
}
} else {
#[cfg_attr(feature = "cargo-clippy", allow(cast_ptr_alignment))]
#[inline(always)]
fn check_utf16_for_latin1_and_bidi_impl(buffer: &[u16]) -> Latin1Bidi {
let mut offset = 0usize;
let len = buffer.len();
if len >= ALU_ALIGNMENT / 2 {
let src = buffer.as_ptr();
let mut until_alignment = ((ALU_ALIGNMENT - ((src as usize) & ALU_ALIGNMENT_MASK)) &
ALU_ALIGNMENT_MASK) / 2;
if until_alignment + ALU_ALIGNMENT / 2 <= len {
while until_alignment != 0 {
if buffer[offset] > 0xFF {
if is_utf16_bidi_impl(&buffer[offset..]) {
return Latin1Bidi::Bidi;
}
return Latin1Bidi::LeftToRight;
}
offset += 1;
until_alignment -= 1;
}
let len_minus_stride = len - ALU_ALIGNMENT / 2;
loop {
if unsafe { *(src.add(offset) as *const usize) } & LATIN1_MASK != 0 {
if is_utf16_bidi_impl(&buffer[offset..]) {
return Latin1Bidi::Bidi;
}
return Latin1Bidi::LeftToRight;
}
offset += ALU_ALIGNMENT / 2;
if offset > len_minus_stride {
break;
}
}
}
}
let mut iter = (&buffer[offset..]).iter();
loop {
if let Some(&u) = iter.next() {
if u > 0xFF {
let mut inner_u = u;
loop {
if is_utf16_code_unit_bidi(inner_u) {
return Latin1Bidi::Bidi;
}
if let Some(&code_unit) = iter.next() {
inner_u = code_unit;
} else {
return Latin1Bidi::LeftToRight;
}
}
}
} else {
return Latin1Bidi::Latin1;
}
}
}
}
}
/// Checks whether the buffer is all-ASCII.
///
/// May read the entire buffer even if it isn't all-ASCII. (I.e. the function
/// is not guaranteed to fail fast.)
pub fn is_ascii(buffer: &[u8]) -> bool {
is_ascii_impl(buffer)
}
/// Checks whether the buffer is all-Basic Latin (i.e. UTF-16 representing
/// only ASCII characters).
///
/// May read the entire buffer even if it isn't all-ASCII. (I.e. the function
/// is not guaranteed to fail fast.)
pub fn is_basic_latin(buffer: &[u16]) -> bool {
is_basic_latin_impl(buffer)
}
/// Checks whether the buffer is valid UTF-8 representing only code points
/// less than or equal to U+00FF.
///
/// Fails fast. (I.e. returns before having read the whole buffer if UTF-8
/// invalidity or code points above U+00FF are discovered.
pub fn is_utf8_latin1(buffer: &[u8]) -> bool {
is_utf8_latin1_impl(buffer).is_none()
}
/// Checks whether the buffer represents only code points less than or equal
/// to U+00FF.
///
/// Fails fast. (I.e. returns before having read the whole buffer if code
/// points above U+00FF are discovered.
pub fn is_str_latin1(buffer: &str) -> bool {
is_str_latin1_impl(buffer).is_none()
}
/// Checks whether the buffer represents only code point less than or equal
/// to U+00FF.
///
/// May read the entire buffer even if it isn't all-Latin1. (I.e. the function
/// is not guaranteed to fail fast.)
pub fn is_utf16_latin1(buffer: &[u16]) -> bool {
is_utf16_latin1_impl(buffer)
}
/// Checks whether a potentially-invalid UTF-8 buffer contains code points
/// that trigger right-to-left processing.
///
/// The check is done on a Unicode block basis without regard to assigned
/// vs. unassigned code points in the block. Hebrew presentation forms in
/// the Alphabetic Presentation Forms block are treated as if they formed
/// a block on their own (i.e. it treated as right-to-left). Additionally,
/// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked
/// for. Control characters that are technically bidi controls but do not
/// cause right-to-left behavior without the presence of right-to-left
/// characters or right-to-left controls are not checked for. As a special
/// case, U+FEFF is excluded from Arabic Presentation Forms-B.
///
/// Returns `true` if the input is invalid UTF-8 or the input contains an
/// RTL character. Returns `false` if the input is valid UTF-8 and contains
/// no RTL characters.
#[cfg_attr(feature = "cargo-clippy", allow(collapsible_if, cyclomatic_complexity))]
#[inline]
pub fn is_utf8_bidi(buffer: &[u8]) -> bool {
// As of rustc 1.25.0-nightly (73ac5d6a8 2018-01-11), this is faster
// than UTF-8 validation followed by `is_str_bidi()` for German,
// Russian and Japanese. However, this is considerably slower for Thai.
// Chances are that the compiler makes some branch predictions that are
// unfortunate for Thai. Not spending the time to manually optimize
// further at this time, since it's unclear if this variant even has
// use cases. However, this is worth revisiting once Rust gets the
// ability to annotate relative priorities of match arms.
// U+058F: D6 8F
// U+0590: D6 90
// U+08FF: E0 A3 BF
// U+0900: E0 A4 80
//
// U+200F: E2 80 8F
// U+202B: E2 80 AB
// U+202E: E2 80 AE
// U+2067: E2 81 A7
//
// U+FB1C: EF AC 9C
// U+FB1D: EF AC 9D
// U+FDFF: EF B7 BF
// U+FE00: EF B8 80
//
// U+FE6F: EF B9 AF
// U+FE70: EF B9 B0
// U+FEFE: EF BB BE
// U+FEFF: EF BB BF
//
// U+107FF: F0 90 9F BF
// U+10800: F0 90 A0 80
// U+10FFF: F0 90 BF BF
// U+11000: F0 91 80 80
//
// U+1E7FF: F0 9E 9F BF
// U+1E800: F0 9E A0 80
// U+1EFFF: F0 9E BF BF
// U+1F000: F0 9F 80 80
let mut src = buffer;
'outer: loop {
if let Some((mut byte, mut read)) = validate_ascii(src) {
// Check for the longest sequence to avoid checking twice for the
// multi-byte sequences.
if read + 4 <= src.len() {
'inner: loop {
// At this point, `byte` is not included in `read`.
match byte {
0..=0x7F => {
// ASCII: go back to SIMD.
read += 1;
src = &src[read..];
continue 'outer;
}
0xC2..=0xD5 => {
// Two-byte
let second = unsafe { *(src.get_unchecked(read + 1)) };
if !in_inclusive_range8(second, 0x80, 0xBF) {
return true;
}
read += 2;
}
0xD6 => {
// Two-byte
let second = unsafe { *(src.get_unchecked(read + 1)) };
if !in_inclusive_range8(second, 0x80, 0xBF) {
return true;
}
// XXX consider folding the above and below checks
if second > 0x8F {
return true;
}
read += 2;
}
// two-byte starting with 0xD7 and above is bidi
0xE1 | 0xE3..=0xEC | 0xEE => {
// Three-byte normal
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe {
*(UTF8_DATA.table.get_unchecked(byte as usize + 0x80))
})
| (third >> 6))
!= 2
{
return true;
}
read += 3;
}
0xE2 => {
// Three-byte normal, potentially bidi
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe {
*(UTF8_DATA.table.get_unchecked(byte as usize + 0x80))
})
| (third >> 6))
!= 2
{
return true;
}
if second == 0x80 {
if third == 0x8F || third == 0xAB || third == 0xAE {
return true;
}
} else if second == 0x81 {
if third == 0xA7 {
return true;
}
}
read += 3;
}
0xEF => {
// Three-byte normal, potentially bidi
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe {
*(UTF8_DATA.table.get_unchecked(byte as usize + 0x80))
})
| (third >> 6))
!= 2
{
return true;
}
if in_inclusive_range8(second, 0xAC, 0xB7) {
if second == 0xAC {
if third > 0x9C {
return true;
}
} else {
return true;
}
} else if in_inclusive_range8(second, 0xB9, 0xBB) {
if second == 0xB9 {
if third > 0xAF {
return true;
}
} else if second == 0xBB {
if third != 0xBF {
return true;
}
} else {
return true;
}
}
read += 3;
}
0xE0 => {
// Three-byte special lower bound, potentially bidi
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe {
*(UTF8_DATA.table.get_unchecked(byte as usize + 0x80))
})
| (third >> 6))
!= 2
{
return true;
}
// XXX can this be folded into the above validity check
if second < 0xA4 {
return true;
}
read += 3;
}
0xED => {
// Three-byte special upper bound
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe {
*(UTF8_DATA.table.get_unchecked(byte as usize + 0x80))
})
| (third >> 6))
!= 2
{
return true;
}
read += 3;
}
0xF1..=0xF4 => {
// Four-byte normal
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
let fourth = unsafe { *(src.get_unchecked(read + 3)) };
if (u16::from(
UTF8_DATA.table[usize::from(second)]
& unsafe {
*(UTF8_DATA.table.get_unchecked(byte as usize + 0x80))
},
) | u16::from(third >> 6)
| (u16::from(fourth & 0xC0) << 2))
!= 0x202
{
return true;
}
read += 4;
}
0xF0 => {
// Four-byte special lower bound, potentially bidi
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
let fourth = unsafe { *(src.get_unchecked(read + 3)) };
if (u16::from(
UTF8_DATA.table[usize::from(second)]
& unsafe {
*(UTF8_DATA.table.get_unchecked(byte as usize + 0x80))
},
) | u16::from(third >> 6)
| (u16::from(fourth & 0xC0) << 2))
!= 0x202
{
return true;
}
if unlikely(second == 0x90 || second == 0x9E) {
let third = src[read + 2];
if third >= 0xA0 {
return true;
}
}
read += 4;
}
_ => {
// Invalid lead or bidi-only lead
return true;
}
}
if read + 4 > src.len() {
if read == src.len() {
return false;
}
byte = src[read];
break 'inner;
}
byte = src[read];
continue 'inner;
}
}
// We can't have a complete 4-byte sequence, but we could still have
// a complete shorter sequence.
// At this point, `byte` is not included in `read`.
match byte {
0..=0x7F => {
// ASCII: go back to SIMD.
read += 1;
src = &src[read..];
continue 'outer;
}
0xC2..=0xD5 => {
// Two-byte
let new_read = read + 2;
if new_read > src.len() {
return true;
}
let second = unsafe { *(src.get_unchecked(read + 1)) };
if !in_inclusive_range8(second, 0x80, 0xBF) {
return true;
}
read = new_read;
// We need to deal with the case where we came here with 3 bytes
// left, so we need to take a look at the last one.
src = &src[read..];
continue 'outer;
}
0xD6 => {
// Two-byte, potentially bidi
let new_read = read + 2;
if new_read > src.len() {
return true;
}
let second = unsafe { *(src.get_unchecked(read + 1)) };
if !in_inclusive_range8(second, 0x80, 0xBF) {
return true;
}
// XXX consider folding the above and below checks
if second > 0x8F {
return true;
}
read = new_read;
// We need to deal with the case where we came here with 3 bytes
// left, so we need to take a look at the last one.
src = &src[read..];
continue 'outer;
}
// two-byte starting with 0xD7 and above is bidi
0xE1 | 0xE3..=0xEC | 0xEE => {
// Three-byte normal
let new_read = read + 3;
if new_read > src.len() {
return true;
}
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) })
| (third >> 6))
!= 2
{
return true;
}
}
0xE2 => {
// Three-byte normal, potentially bidi
let new_read = read + 3;
if new_read > src.len() {
return true;
}
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) })
| (third >> 6))
!= 2
{
return true;
}
if second == 0x80 {
if third == 0x8F || third == 0xAB || third == 0xAE {
return true;
}
} else if second == 0x81 {
if third == 0xA7 {
return true;
}
}
}
0xEF => {
// Three-byte normal, potentially bidi
let new_read = read + 3;
if new_read > src.len() {
return true;
}
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) })
| (third >> 6))
!= 2
{
return true;
}
if in_inclusive_range8(second, 0xAC, 0xB7) {
if second == 0xAC {
if third > 0x9C {
return true;
}
} else {
return true;
}
} else if in_inclusive_range8(second, 0xB9, 0xBB) {
if second == 0xB9 {
if third > 0xAF {
return true;
}
} else if second == 0xBB {
if third != 0xBF {
return true;
}
} else {
return true;
}
}
}
0xE0 => {
// Three-byte special lower bound, potentially bidi
let new_read = read + 3;
if new_read > src.len() {
return true;
}
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) })
| (third >> 6))
!= 2
{
return true;
}
// XXX can this be folded into the above validity check
if second < 0xA4 {
return true;
}
}
0xED => {
// Three-byte special upper bound
let new_read = read + 3;
if new_read > src.len() {
return true;
}
let second = unsafe { *(src.get_unchecked(read + 1)) };
let third = unsafe { *(src.get_unchecked(read + 2)) };
if ((UTF8_DATA.table[usize::from(second)]
& unsafe { *(UTF8_DATA.table.get_unchecked(byte as usize + 0x80)) })
| (third >> 6))
!= 2
{
return true;
}
}
_ => {
// Invalid lead, 4-byte lead or 2-byte bidi-only lead
return true;
}
}
return false;
} else {
return false;
}
}
}
/// Checks whether a valid UTF-8 buffer contains code points that trigger
/// right-to-left processing.
///
/// The check is done on a Unicode block basis without regard to assigned
/// vs. unassigned code points in the block. Hebrew presentation forms in
/// the Alphabetic Presentation Forms block are treated as if they formed
/// a block on their own (i.e. it treated as right-to-left). Additionally,
/// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked
/// for. Control characters that are technically bidi controls but do not
/// cause right-to-left behavior without the presence of right-to-left
/// characters or right-to-left controls are not checked for. As a special
/// case, U+FEFF is excluded from Arabic Presentation Forms-B.
#[cfg_attr(feature = "cargo-clippy", allow(collapsible_if))]
#[inline]
pub fn is_str_bidi(buffer: &str) -> bool {
// U+058F: D6 8F
// U+0590: D6 90
// U+08FF: E0 A3 BF
// U+0900: E0 A4 80
//
// U+200F: E2 80 8F
// U+202B: E2 80 AB
// U+202E: E2 80 AE
// U+2067: E2 81 A7
//
// U+FB1C: EF AC 9C
// U+FB1D: EF AC 9D
// U+FDFF: EF B7 BF
// U+FE00: EF B8 80
//
// U+FE6F: EF B9 AF
// U+FE70: EF B9 B0
// U+FEFE: EF BB BE
// U+FEFF: EF BB BF
//
// U+107FF: F0 90 9F BF
// U+10800: F0 90 A0 80
// U+10FFF: F0 90 BF BF
// U+11000: F0 91 80 80
//
// U+1E7FF: F0 9E 9F BF
// U+1E800: F0 9E A0 80
// U+1EFFF: F0 9E BF BF
// U+1F000: F0 9F 80 80
let mut bytes = buffer.as_bytes();
'outer: loop {
// TODO: Instead of just validating ASCII using SIMD, use SIMD
// to check for non-ASCII lead bytes, too, to quickly conclude
// that the vector consist entirely of CJK and below-Hebrew
// code points.
// Unfortunately, scripts above Arabic but below CJK share
// lead bytes with RTL.
if let Some((mut byte, mut read)) = validate_ascii(bytes) {
'inner: loop {
// At this point, `byte` is not included in `read`.
if byte < 0xE0 {
if byte >= 0x80 {
// Two-byte
// Adding `unlikely` here improved throughput on
// Russian plain text by 33%!
if unlikely(byte >= 0xD6) {
if byte == 0xD6 {
let second = bytes[read + 1];
if second > 0x8F {
return true;
}
} else {
return true;
}
}
read += 2;
} else {
// ASCII: write and go back to SIMD.
read += 1;
// Intuitively, we should go back to the outer loop only
// if byte is 0x30 or above, so as to avoid trashing on
// ASCII space, comma and period in non-Latin context.
// However, the extra branch seems to cost more than it's
// worth.
bytes = &bytes[read..];
continue 'outer;
}
} else if byte < 0xF0 {
// Three-byte
if unlikely(!in_inclusive_range8(byte, 0xE3, 0xEE) && byte != 0xE1) {
let second = bytes[read + 1];
if byte == 0xE0 {
if second < 0xA4 {
return true;
}
} else if byte == 0xE2 {
let third = bytes[read + 2];
if second == 0x80 {
if third == 0x8F || third == 0xAB || third == 0xAE {
return true;
}
} else if second == 0x81 {
if third == 0xA7 {
return true;
}
}
} else {
debug_assert_eq!(byte, 0xEF);
if in_inclusive_range8(second, 0xAC, 0xB7) {
if second == 0xAC {
let third = bytes[read + 2];
if third > 0x9C {
return true;
}
} else {
return true;
}
} else if in_inclusive_range8(second, 0xB9, 0xBB) {
if second == 0xB9 {
let third = bytes[read + 2];
if third > 0xAF {
return true;
}
} else if second == 0xBB {
let third = bytes[read + 2];
if third != 0xBF {
return true;
}
} else {
return true;
}
}
}
}
read += 3;
} else {
// Four-byte
let second = bytes[read + 1];
if unlikely(byte == 0xF0 && (second == 0x90 || second == 0x9E)) {
let third = bytes[read + 2];
if third >= 0xA0 {
return true;
}
}
read += 4;
}
// The comparison is always < or == and never >, but including
// > here to let the compiler assume that < is true if this
// comparison is false.
if read >= bytes.len() {
return false;
}
byte = bytes[read];
continue 'inner;
}
} else {
return false;
}
}
}
/// Checks whether a UTF-16 buffer contains code points that trigger
/// right-to-left processing.
///
/// The check is done on a Unicode block basis without regard to assigned
/// vs. unassigned code points in the block. Hebrew presentation forms in
/// the Alphabetic Presentation Forms block are treated as if they formed
/// a block on their own (i.e. it treated as right-to-left). Additionally,
/// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked
/// for. Control characters that are technically bidi controls but do not
/// cause right-to-left behavior without the presence of right-to-left
/// characters or right-to-left controls are not checked for. As a special
/// case, U+FEFF is excluded from Arabic Presentation Forms-B.
///
/// Returns `true` if the input contains an RTL character or an unpaired
/// high surrogate that could be the high half of an RTL character.
/// Returns `false` if the input contains neither RTL characters nor
/// unpaired high surrogates that could be higher halves of RTL characters.
pub fn is_utf16_bidi(buffer: &[u16]) -> bool {
is_utf16_bidi_impl(buffer)
}
/// Checks whether a scalar value triggers right-to-left processing.
///
/// The check is done on a Unicode block basis without regard to assigned
/// vs. unassigned code points in the block. Hebrew presentation forms in
/// the Alphabetic Presentation Forms block are treated as if they formed
/// a block on their own (i.e. it treated as right-to-left). Additionally,
/// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked
/// for. Control characters that are technically bidi controls but do not
/// cause right-to-left behavior without the presence of right-to-left
/// characters or right-to-left controls are not checked for. As a special
/// case, U+FEFF is excluded from Arabic Presentation Forms-B.
#[inline(always)]
pub fn is_char_bidi(c: char) -> bool {
// Controls:
// Every control with RIGHT-TO-LEFT in its name in
// https://www.unicode.org/charts/PDF/U2000.pdf
// U+200F RLM
// U+202B RLE
// U+202E RLO
// U+2067 RLI
//
// BMP RTL:
// https://www.unicode.org/roadmaps/bmp/
// U+0590...U+08FF
// U+FB1D...U+FDFF Hebrew presentation forms and
// Arabic Presentation Forms A
// U+FE70...U+FEFE Arabic Presentation Forms B (excl. BOM)
//
// Supplementary RTL:
// https://www.unicode.org/roadmaps/smp/
// U+10800...U+10FFF (Lead surrogate U+D802 or U+D803)
// U+1E800...U+1EFFF (Lead surrogate U+D83A or U+D83B)
let code_point = u32::from(c);
if code_point < 0x0590 {
// Below Hebrew
return false;
}
if in_range32(code_point, 0x0900, 0xFB1D) {
// Above Arabic Extended-A and below Hebrew presentation forms
if in_inclusive_range32(code_point, 0x200F, 0x2067) {
// In the range that contains the RTL controls
return code_point == 0x200F
|| code_point == 0x202B
|| code_point == 0x202E
|| code_point == 0x2067;
}
return false;
}
if code_point > 0x1EFFF {
// Above second astral RTL. (Emoji is here.)
return false;
}
if in_range32(code_point, 0x11000, 0x1E800) {
// Between astral RTL blocks
return false;
}
if in_range32(code_point, 0xFEFF, 0x10800) {
// Above Arabic Presentations Forms B (excl. BOM) and below first
// astral RTL
return false;
}
if in_range32(code_point, 0xFE00, 0xFE70) {
// Between Arabic Presentations Forms
return false;
}
true
}
/// Checks whether a UTF-16 code unit triggers right-to-left processing.
///
/// The check is done on a Unicode block basis without regard to assigned
/// vs. unassigned code points in the block. Hebrew presentation forms in
/// the Alphabetic Presentation Forms block are treated as if they formed
/// a block on their own (i.e. it treated as right-to-left). Additionally,
/// the four RIGHT-TO-LEFT FOO controls in General Punctuation are checked
/// for. Control characters that are technically bidi controls but do not
/// cause right-to-left behavior without the presence of right-to-left
/// characters or right-to-left controls are not checked for. As a special
/// case, U+FEFF is excluded from Arabic Presentation Forms-B.
///
/// Since supplementary-plane right-to-left blocks are identifiable from the
/// high surrogate without examining the low surrogate, this function returns
/// `true` for such high surrogates making the function suitable for handling
/// supplementary-plane text without decoding surrogate pairs to scalar
/// values. Obviously, such high surrogates are then reported as right-to-left
/// even if actually unpaired.
#[inline(always)]
pub fn is_utf16_code_unit_bidi(u: u16) -> bool {
if u < 0x0590 {
// Below Hebrew
return false;
}
if in_range16(u, 0x0900, 0xD802) {
// Above Arabic Extended-A and below first RTL surrogate
if in_inclusive_range16(u, 0x200F, 0x2067) {
// In the range that contains the RTL controls
return u == 0x200F || u == 0x202B || u == 0x202E || u == 0x2067;
}
return false;
}
if in_range16(u, 0xD83C, 0xFB1D) {
// Between astral RTL high surrogates and Hebrew presentation forms
// (Emoji is here)
return false;
}
if in_range16(u, 0xD804, 0xD83A) {
// Between RTL high surragates
return false;
}
if u > 0xFEFE {
// Above Arabic Presentation Forms (excl. BOM)
return false;
}
if in_range16(u, 0xFE00, 0xFE70) {
// Between Arabic Presentations Forms
return false;
}
true
}
/// Checks whether a potentially invalid UTF-8 buffer contains code points
/// that trigger right-to-left processing or is all-Latin1.
///
/// Possibly more efficient than performing the checks separately.
///
/// Returns `Latin1Bidi::Latin1` if `is_utf8_latin1()` would return `true`.
/// Otherwise, returns `Latin1Bidi::Bidi` if `is_utf8_bidi()` would return
/// `true`. Otherwise, returns `Latin1Bidi::LeftToRight`.
pub fn check_utf8_for_latin1_and_bidi(buffer: &[u8]) -> Latin1Bidi {
if let Some(offset) = is_utf8_latin1_impl(buffer) {
if is_utf8_bidi(&buffer[offset..]) {
Latin1Bidi::Bidi
} else {
Latin1Bidi::LeftToRight
}
} else {
Latin1Bidi::Latin1
}
}
/// Checks whether a valid UTF-8 buffer contains code points
/// that trigger right-to-left processing or is all-Latin1.
///
/// Possibly more efficient than performing the checks separately.
///
/// Returns `Latin1Bidi::Latin1` if `is_str_latin1()` would return `true`.
/// Otherwise, returns `Latin1Bidi::Bidi` if `is_str_bidi()` would return
/// `true`. Otherwise, returns `Latin1Bidi::LeftToRight`.
pub fn check_str_for_latin1_and_bidi(buffer: &str) -> Latin1Bidi {
// The transition from the latin1 check to the bidi check isn't
// optimal but not tweaking it to perfection today.
if let Some(offset) = is_str_latin1_impl(buffer) {
if is_str_bidi(&buffer[offset..]) {
Latin1Bidi::Bidi
} else {
Latin1Bidi::LeftToRight
}
} else {
Latin1Bidi::Latin1
}
}
/// Checks whether a potentially invalid UTF-16 buffer contains code points
/// that trigger right-to-left processing or is all-Latin1.
///
/// Possibly more efficient than performing the checks separately.
///
/// Returns `Latin1Bidi::Latin1` if `is_utf16_latin1()` would return `true`.
/// Otherwise, returns `Latin1Bidi::Bidi` if `is_utf16_bidi()` would return
/// `true`. Otherwise, returns `Latin1Bidi::LeftToRight`.
pub fn check_utf16_for_latin1_and_bidi(buffer: &[u16]) -> Latin1Bidi {
check_utf16_for_latin1_and_bidi_impl(buffer)
}
/// Converts potentially-invalid UTF-8 to valid UTF-16 with errors replaced
/// with the REPLACEMENT CHARACTER.
///
/// The length of the destination buffer must be at least the length of the
/// source buffer _plus one_.
///
/// Returns the number of `u16`s written.
///
/// # Panics
///
/// Panics if the destination buffer is shorter than stated above.
pub fn convert_utf8_to_utf16(src: &[u8], dst: &mut [u16]) -> usize {
// TODO: Can the requirement for dst to be at least one unit longer
// be eliminated?
assert!(dst.len() > src.len());
let mut decoder = Utf8Decoder::new_inner();
let mut total_read = 0usize;
let mut total_written = 0usize;
loop {
let (result, read, written) =
decoder.decode_to_utf16_raw(&src[total_read..], &mut dst[total_written..], true);
total_read += read;
total_written += written;
match result {
DecoderResult::InputEmpty => {
return total_written;
}
DecoderResult::OutputFull => {
unreachable!("The assert at the top of the function should have caught this.");
}
DecoderResult::Malformed(_, _) => {
// There should always be space for the U+FFFD, because
// otherwise we'd have gotten OutputFull already.
dst[total_written] = 0xFFFD;
total_written += 1;
}
}
}
}
/// Converts valid UTF-8 to valid UTF-16.
///
/// The length of the destination buffer must be at least the length of the
/// source buffer.
///
/// Returns the number of `u16`s written.
///
/// # Panics
///
/// Panics if the destination buffer is shorter than stated above.
pub fn convert_str_to_utf16(src: &str, dst: &mut [u16]) -> usize {
assert!(
dst.len() >= src.len(),
"Destination must not be shorter than the source."
);
let bytes = src.as_bytes();
let mut read = 0;
let mut written = 0;
'outer: loop {
let mut byte = {
let src_remaining = &bytes[read..];
let dst_remaining = &mut dst[written..];
let length = src_remaining.len();
match unsafe {
ascii_to_basic_latin(src_remaining.as_ptr(), dst_remaining.as_mut_ptr(), length)
} {
None => {
written += length;
return written;
}
Some((non_ascii, consumed)) => {
--> --------------------
--> maximum size reached
--> --------------------
