ladybird/AK/Utf8View.h
Ali Mohammad Pur 5e1499d104 Everywhere: Rename {Deprecated => Byte}String
This commit un-deprecates DeprecatedString, and repurposes it as a byte
string.
As the null state has already been removed, there are no other
particularly hairy blockers in repurposing this type as a byte string
(what it _really_ is).

This commit is auto-generated:
  $ xs=$(ack -l \bDeprecatedString\b\|deprecated_string AK Userland \
    Meta Ports Ladybird Tests Kernel)
  $ perl -pie 's/\bDeprecatedString\b/ByteString/g;
    s/deprecated_string/byte_string/g' $xs
  $ clang-format --style=file -i \
    $(git diff --name-only | grep \.cpp\|\.h)
  $ gn format $(git ls-files '*.gn' '*.gni')
2023-12-17 18:25:10 +03:30

284 lines
8.0 KiB
C++

/*
* Copyright (c) 2019-2020, Sergey Bugaev <bugaevc@serenityos.org>
* Copyright (c) 2021, Max Wipfli <mail@maxwipfli.ch>
*
* SPDX-License-Identifier: BSD-2-Clause
*/
#pragma once
#include <AK/Format.h>
#include <AK/StringView.h>
#include <AK/Types.h>
#ifndef KERNEL
# include <AK/ByteString.h>
#endif
namespace AK {
class Utf8View;
class Utf8CodePointIterator {
friend class Utf8View;
public:
Utf8CodePointIterator() = default;
~Utf8CodePointIterator() = default;
bool operator==(Utf8CodePointIterator const&) const = default;
bool operator!=(Utf8CodePointIterator const&) const = default;
Utf8CodePointIterator& operator++();
u32 operator*() const;
// NOTE: This returns {} if the peek is at or past EOF.
Optional<u32> peek(size_t offset = 0) const;
ssize_t operator-(Utf8CodePointIterator const& other) const
{
return m_ptr - other.m_ptr;
}
u8 const* ptr() const { return m_ptr; }
// Note : These methods return the information about the underlying UTF-8 bytes.
// If the UTF-8 string encoding is not valid at the iterator's position, then the underlying bytes might be different from the
// decoded character's re-encoded bytes (which will be an `0xFFFD REPLACEMENT CHARACTER` with an UTF-8 length of three bytes).
// If your code relies on the decoded character being equivalent to the re-encoded character, use the `UTF8View::validate()`
// method on the view prior to using its iterator.
size_t underlying_code_point_length_in_bytes() const;
ReadonlyBytes underlying_code_point_bytes() const;
bool done() const { return m_length == 0; }
private:
Utf8CodePointIterator(u8 const* ptr, size_t length)
: m_ptr(ptr)
, m_length(length)
{
}
u8 const* m_ptr { nullptr };
size_t m_length { 0 };
};
class Utf8View {
public:
using Iterator = Utf8CodePointIterator;
Utf8View() = default;
explicit constexpr Utf8View(StringView string)
: m_string(string)
{
}
#ifndef KERNEL
explicit Utf8View(ByteString& string)
: m_string(string.view())
{
}
explicit Utf8View(ByteString&&) = delete;
#endif
~Utf8View() = default;
StringView as_string() const { return m_string; }
Utf8CodePointIterator begin() const { return { begin_ptr(), m_string.length() }; }
Utf8CodePointIterator end() const { return { end_ptr(), 0 }; }
Utf8CodePointIterator iterator_at_byte_offset(size_t) const;
Utf8CodePointIterator iterator_at_byte_offset_without_validation(size_t) const;
unsigned char const* bytes() const { return begin_ptr(); }
size_t byte_length() const { return m_string.length(); }
size_t byte_offset_of(Utf8CodePointIterator const&) const;
size_t byte_offset_of(size_t code_point_offset) const;
Utf8View substring_view(size_t byte_offset, size_t byte_length) const { return Utf8View { m_string.substring_view(byte_offset, byte_length) }; }
Utf8View substring_view(size_t byte_offset) const { return substring_view(byte_offset, byte_length() - byte_offset); }
Utf8View unicode_substring_view(size_t code_point_offset, size_t code_point_length) const;
Utf8View unicode_substring_view(size_t code_point_offset) const { return unicode_substring_view(code_point_offset, length() - code_point_offset); }
bool is_empty() const { return m_string.is_empty(); }
bool is_null() const { return m_string.is_null(); }
bool starts_with(Utf8View const&) const;
bool contains(u32) const;
Utf8View trim(Utf8View const& characters, TrimMode mode = TrimMode::Both) const;
size_t iterator_offset(Utf8CodePointIterator const& it) const
{
return byte_offset_of(it);
}
size_t length() const
{
if (!m_have_length) {
m_length = calculate_length();
m_have_length = true;
}
return m_length;
}
constexpr bool validate() const
{
size_t valid_bytes = 0;
return validate(valid_bytes);
}
constexpr bool validate(size_t& valid_bytes) const
{
valid_bytes = 0;
for (auto it = m_string.begin(); it != m_string.end(); ++it) {
auto [byte_length, code_point, is_valid] = decode_leading_byte(static_cast<u8>(*it));
if (!is_valid)
return false;
for (size_t i = 1; i < byte_length; ++i) {
if (++it == m_string.end())
return false;
auto [code_point_bits, is_valid] = decode_continuation_byte(static_cast<u8>(*it));
if (!is_valid)
return false;
code_point <<= 6;
code_point |= code_point_bits;
}
if (!is_valid_code_point(code_point, byte_length))
return false;
valid_bytes += byte_length;
}
return true;
}
private:
friend class Utf8CodePointIterator;
u8 const* begin_ptr() const { return reinterpret_cast<u8 const*>(m_string.characters_without_null_termination()); }
u8 const* end_ptr() const { return begin_ptr() + m_string.length(); }
size_t calculate_length() const;
struct Utf8EncodedByteData {
size_t byte_length { 0 };
u8 encoding_bits { 0 };
u8 encoding_mask { 0 };
u32 first_code_point { 0 };
u32 last_code_point { 0 };
};
static constexpr Array<Utf8EncodedByteData, 4> utf8_encoded_byte_data { {
{ 1, 0b0000'0000, 0b1000'0000, 0x0000, 0x007F },
{ 2, 0b1100'0000, 0b1110'0000, 0x0080, 0x07FF },
{ 3, 0b1110'0000, 0b1111'0000, 0x0800, 0xFFFF },
{ 4, 0b1111'0000, 0b1111'1000, 0x10000, 0x10FFFF },
} };
struct LeadingByte {
size_t byte_length { 0 };
u32 code_point_bits { 0 };
bool is_valid { false };
};
static constexpr LeadingByte decode_leading_byte(u8 byte)
{
for (auto const& data : utf8_encoded_byte_data) {
if ((byte & data.encoding_mask) != data.encoding_bits)
continue;
byte &= ~data.encoding_mask;
return { data.byte_length, byte, true };
}
return { .is_valid = false };
}
struct ContinuationByte {
u32 code_point_bits { 0 };
bool is_valid { false };
};
static constexpr ContinuationByte decode_continuation_byte(u8 byte)
{
constexpr u8 continuation_byte_encoding_bits = 0b1000'0000;
constexpr u8 continuation_byte_encoding_mask = 0b1100'0000;
if ((byte & continuation_byte_encoding_mask) == continuation_byte_encoding_bits) {
byte &= ~continuation_byte_encoding_mask;
return { byte, true };
}
return { .is_valid = false };
}
static constexpr bool is_valid_code_point(u32 code_point, size_t byte_length)
{
for (auto const& data : utf8_encoded_byte_data) {
if (code_point >= data.first_code_point && code_point <= data.last_code_point)
return byte_length == data.byte_length;
}
return false;
}
StringView m_string;
mutable size_t m_length { 0 };
mutable bool m_have_length { false };
};
#ifndef KERNEL
class DeprecatedStringCodePointIterator {
public:
Optional<u32> next()
{
if (m_it.done())
return {};
auto value = *m_it;
++m_it;
return value;
}
[[nodiscard]] Optional<u32> peek() const
{
if (m_it.done())
return {};
return *m_it;
}
[[nodiscard]] size_t byte_offset() const
{
return Utf8View(m_string).byte_offset_of(m_it);
}
DeprecatedStringCodePointIterator(ByteString string)
: m_string(move(string))
, m_it(Utf8View(m_string).begin())
{
}
private:
ByteString m_string;
Utf8CodePointIterator m_it;
};
#endif
template<>
struct Formatter<Utf8View> : Formatter<StringView> {
ErrorOr<void> format(FormatBuilder&, Utf8View const&);
};
}
#if USING_AK_GLOBALLY
# ifndef KERNEL
using AK::DeprecatedStringCodePointIterator;
# endif
using AK::Utf8CodePointIterator;
using AK::Utf8View;
#endif