mirror of
https://github.com/LadybirdBrowser/ladybird.git
synced 2024-11-22 15:40:19 +00:00
156 lines
4.8 KiB
C++
156 lines
4.8 KiB
C++
/*
|
|
* Copyright (c) 2021-2024, Tim Flynn <trflynn89@serenityos.org>
|
|
*
|
|
* SPDX-License-Identifier: BSD-2-Clause
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include <AK/ByteString.h>
|
|
#include <AK/Endian.h>
|
|
#include <AK/Error.h>
|
|
#include <AK/Format.h>
|
|
#include <AK/Forward.h>
|
|
#include <AK/Optional.h>
|
|
#include <AK/Span.h>
|
|
#include <AK/String.h>
|
|
#include <AK/Types.h>
|
|
#include <AK/Vector.h>
|
|
|
|
namespace AK {
|
|
|
|
using Utf16Data = Vector<u16, 1>;
|
|
|
|
ErrorOr<Utf16Data> utf8_to_utf16(StringView, Endianness = Endianness::Host);
|
|
ErrorOr<Utf16Data> utf8_to_utf16(Utf8View const&, Endianness = Endianness::Host);
|
|
ErrorOr<Utf16Data> utf32_to_utf16(Utf32View const&, Endianness = Endianness::Host);
|
|
ErrorOr<void> code_point_to_utf16(Utf16Data&, u32, Endianness = Endianness::Host);
|
|
|
|
size_t utf16_code_unit_length_from_utf8(StringView);
|
|
|
|
class Utf16View;
|
|
|
|
class Utf16CodePointIterator {
|
|
friend class Utf16View;
|
|
|
|
public:
|
|
Utf16CodePointIterator() = default;
|
|
~Utf16CodePointIterator() = default;
|
|
|
|
bool operator==(Utf16CodePointIterator const& other) const
|
|
{
|
|
return (m_ptr == other.m_ptr) && (m_remaining_code_units == other.m_remaining_code_units);
|
|
}
|
|
|
|
Utf16CodePointIterator& operator++();
|
|
u32 operator*() const;
|
|
|
|
size_t length_in_code_units() const;
|
|
|
|
private:
|
|
Utf16CodePointIterator(u16 const* ptr, size_t length, Endianness endianness)
|
|
: m_ptr(ptr)
|
|
, m_remaining_code_units(length)
|
|
, m_endianness(endianness)
|
|
{
|
|
}
|
|
|
|
u16 const* m_ptr { nullptr };
|
|
size_t m_remaining_code_units { 0 };
|
|
Endianness m_endianness { Endianness::Host };
|
|
};
|
|
|
|
class Utf16View {
|
|
public:
|
|
using Iterator = Utf16CodePointIterator;
|
|
|
|
static bool is_high_surrogate(u16);
|
|
static bool is_low_surrogate(u16);
|
|
static u32 decode_surrogate_pair(u16 high_surrogate, u16 low_surrogate);
|
|
|
|
Utf16View() = default;
|
|
~Utf16View() = default;
|
|
|
|
explicit Utf16View(ReadonlySpan<u16> code_units, Endianness endianness = Endianness::Host)
|
|
: m_code_units(code_units)
|
|
, m_endianness(endianness)
|
|
{
|
|
}
|
|
|
|
template<size_t Size>
|
|
Utf16View(char16_t const (&code_units)[Size], Endianness endianness = Endianness::Host)
|
|
: m_code_units(
|
|
reinterpret_cast<u16 const*>(&code_units[0]),
|
|
code_units[Size - 1] == u'\0' ? Size - 1 : Size)
|
|
, m_endianness(endianness)
|
|
{
|
|
}
|
|
|
|
bool operator==(Utf16View const& other) const { return m_code_units == other.m_code_units; }
|
|
|
|
enum class AllowInvalidCodeUnits {
|
|
Yes,
|
|
No,
|
|
};
|
|
|
|
ErrorOr<ByteString> to_byte_string(AllowInvalidCodeUnits = AllowInvalidCodeUnits::No) const;
|
|
ErrorOr<String> to_utf8(AllowInvalidCodeUnits = AllowInvalidCodeUnits::No) const;
|
|
|
|
bool is_null() const { return m_code_units.is_null(); }
|
|
bool is_empty() const { return m_code_units.is_empty(); }
|
|
size_t length_in_code_units() const { return m_code_units.size(); }
|
|
size_t length_in_code_points() const;
|
|
|
|
Endianness endianness() const { return m_endianness; }
|
|
|
|
Utf16CodePointIterator begin() const { return { begin_ptr(), m_code_units.size(), m_endianness }; }
|
|
Utf16CodePointIterator end() const { return { end_ptr(), 0, m_endianness }; }
|
|
|
|
u16 const* data() const { return m_code_units.data(); }
|
|
char16_t const* char_data() const { return reinterpret_cast<char16_t const*>(data()); }
|
|
|
|
u16 code_unit_at(size_t index) const;
|
|
u32 code_point_at(size_t index) const;
|
|
|
|
size_t code_point_offset_of(size_t code_unit_offset) const;
|
|
size_t code_unit_offset_of(size_t code_point_offset) const;
|
|
size_t code_unit_offset_of(Utf16CodePointIterator const&) const;
|
|
|
|
Utf16View substring_view(size_t code_unit_offset, size_t code_unit_length) const;
|
|
Utf16View substring_view(size_t code_unit_offset) const { return substring_view(code_unit_offset, length_in_code_units() - code_unit_offset); }
|
|
|
|
Utf16View unicode_substring_view(size_t code_point_offset, size_t code_point_length) const;
|
|
Utf16View unicode_substring_view(size_t code_point_offset) const { return unicode_substring_view(code_point_offset, length_in_code_points() - code_point_offset); }
|
|
|
|
bool starts_with(Utf16View const&) const;
|
|
|
|
bool validate() const;
|
|
bool validate(size_t& valid_code_units) const;
|
|
|
|
bool equals_ignoring_case(Utf16View const&) const;
|
|
|
|
private:
|
|
u16 const* begin_ptr() const { return m_code_units.data(); }
|
|
u16 const* end_ptr() const { return begin_ptr() + m_code_units.size(); }
|
|
|
|
size_t calculate_length_in_code_points() const;
|
|
|
|
ReadonlySpan<u16> m_code_units;
|
|
mutable Optional<size_t> m_length_in_code_points;
|
|
Endianness m_endianness { Endianness::Host };
|
|
};
|
|
|
|
}
|
|
|
|
template<>
|
|
struct AK::Formatter<AK::Utf16View> : Formatter<FormatString> {
|
|
ErrorOr<void> format(FormatBuilder& builder, AK::Utf16View const& value)
|
|
{
|
|
return builder.builder().try_append(value);
|
|
}
|
|
};
|
|
|
|
#if USING_AK_GLOBALLY
|
|
using AK::Utf16Data;
|
|
using AK::Utf16View;
|
|
#endif
|