ladybird/Userland/Libraries/LibUnicode/CharacterTypes.cpp

/*
 * Copyright (c) 2021, Tim Flynn <trflynn89@serenityos.org>
 *
 * SPDX-License-Identifier: BSD-2-Clause
 */

#include <AK/CharacterTypes.h>
#include <AK/Platform.h>
#include <AK/StringBuilder.h>
#include <AK/Types.h>
#include <AK/Utf16View.h>
#include <AK/Utf8View.h>
#include <LibUnicode/CharacterTypes.h>
#include <LibUnicode/Locale.h>

#if ENABLE_UNICODE_DATA
#    include <LibUnicode/UnicodeData.h>
#endif

// For details on the algorithms used here, see Section 3.13 Default Case Algorithms
// https://www.unicode.org/versions/Unicode13.0.0/ch03.pdf

namespace Unicode {

Optional<String> __attribute__((weak)) code_point_display_name(u32) { return {}; }
Optional<StringView> __attribute__((weak)) code_point_block_display_name(u32) { return {}; }
Optional<StringView> __attribute__((weak)) code_point_abbreviation(u32) { return {}; }
u32 __attribute__((weak)) canonical_combining_class(u32) { return {}; }
Span<SpecialCasing const* const> __attribute__((weak)) special_case_mapping(u32) { return {}; }
Span<BlockName const> __attribute__((weak)) block_display_names() { return {}; }

#if ENABLE_UNICODE_DATA

static bool is_after_uppercase_i(Utf8View const& string, size_t index)
{
    // There is an uppercase I before C, and there is no intervening combining character class 230 (Above) or 0.
    auto preceding_view = string.substring_view(0, index);
    bool found_uppercase_i = false;

    // FIXME: Would be better if Utf8View supported reverse iteration.
    for (auto code_point : preceding_view) {
        if (code_point == 'I') {
            found_uppercase_i = true;
            continue;
        }

        u32 combining_class = canonical_combining_class(code_point);

        if (combining_class == 0)
            found_uppercase_i = false;
        else if (combining_class == 230)
            found_uppercase_i = false;
    }

    return found_uppercase_i;
}

static bool is_after_soft_dotted_code_point(Utf8View const& string, size_t index)
{
    // There is a Soft_Dotted character before C, with no intervening character of combining class 0 or 230 (Above).
    auto preceding_view = string.substring_view(0, index);
    bool found_soft_dotted_code_point = false;

    // FIXME: Would be better if Utf8View supported reverse iteration.
    for (auto code_point : preceding_view) {
        if (code_point_has_property(code_point, Property::Soft_Dotted)) {
            found_soft_dotted_code_point = true;
            continue;
        }

        u32 combining_class = canonical_combining_class(code_point);

        if (combining_class == 0)
            found_soft_dotted_code_point = false;
        else if (combining_class == 230)
            found_soft_dotted_code_point = false;
    }

    return found_soft_dotted_code_point;
}

static bool is_final_code_point(Utf8View const& string, size_t index, size_t byte_length)
{
    // C is preceded by a sequence consisting of a cased letter and then zero or more case-ignorable
    // characters, and C is not followed by a sequence consisting of zero or more case-ignorable
    // characters and then a cased letter.
    auto preceding_view = string.substring_view(0, index);
    auto following_view = ((index + byte_length) < string.byte_length())
        ? string.substring_view(index + byte_length)
        : Utf8View {};

    size_t cased_letter_count = 0;

    for (auto code_point : preceding_view) {
        bool is_cased = code_point_has_property(code_point, Property::Cased);
        bool is_case_ignorable = code_point_has_property(code_point, Property::Case_Ignorable);

        if (is_cased && !is_case_ignorable)
            ++cased_letter_count;
        else if (!is_case_ignorable)
            cased_letter_count = 0;
    }

    if (cased_letter_count == 0)
        return false;

    for (auto code_point : following_view) {
        bool is_cased = code_point_has_property(code_point, Property::Cased);
        bool is_case_ignorable = code_point_has_property(code_point, Property::Case_Ignorable);

        if (is_case_ignorable)
            continue;
        if (is_cased)
            return false;

        break;
    }

    return true;
}

static bool is_followed_by_combining_class_above(Utf8View const& string, size_t index, size_t byte_length)
{
    // C is followed by a character of combining class 230 (Above) with no intervening character of combining class 0 or 230 (Above).
    auto following_view = ((index + byte_length) < string.byte_length())
        ? string.substring_view(index + byte_length)
        : Utf8View {};

    for (auto code_point : following_view) {
        u32 combining_class = canonical_combining_class(code_point);

        if (combining_class == 0)
            return false;
        if (combining_class == 230)
            return true;
    }

    return false;
}

static bool is_followed_by_combining_dot_above(Utf8View const& string, size_t index, size_t byte_length)
{
    // C is followed by combining dot above (U+0307). Any sequence of characters with a combining class that is neither 0 nor 230 may
    // intervene between the current character and the combining dot above.
    auto following_view = ((index + byte_length) < string.byte_length())
        ? string.substring_view(index + byte_length)
        : Utf8View {};

    for (auto code_point : following_view) {
        if (code_point == 0x307)
            return true;

        u32 combining_class = canonical_combining_class(code_point);

        if (combining_class == 0)
            return false;
        if (combining_class == 230)
            return false;
    }

    return false;
}

static SpecialCasing const* find_matching_special_case(u32 code_point, Utf8View const& string, Optional<StringView> locale, size_t index, size_t byte_length)
{
    auto requested_locale = Locale::None;

    if (locale.has_value()) {
        if (auto maybe_locale = locale_from_string(*locale); maybe_locale.has_value())
            requested_locale = *maybe_locale;
    }

    auto special_casings = special_case_mapping(code_point);

    for (auto const* special_casing : special_casings) {
        if (special_casing->locale != Locale::None && special_casing->locale != requested_locale)
            continue;

        switch (special_casing->condition) {
        case Condition::None:
            return special_casing;

        case Condition::AfterI:
            if (is_after_uppercase_i(string, index))
                return special_casing;
            break;

        case Condition::AfterSoftDotted:
            if (is_after_soft_dotted_code_point(string, index))
                return special_casing;
            break;

        case Condition::FinalSigma:
            if (is_final_code_point(string, index, byte_length))
                return special_casing;
            break;

        case Condition::MoreAbove:
            if (is_followed_by_combining_class_above(string, index, byte_length))
                return special_casing;
            break;

        case Condition::NotBeforeDot:
            if (!is_followed_by_combining_dot_above(string, index, byte_length))
                return special_casing;
            break;
        }
    }

    return nullptr;
}

#endif

u32 __attribute__((weak)) to_unicode_lowercase(u32 code_point)
{
    return to_ascii_lowercase(code_point);
}

u32 __attribute__((weak)) to_unicode_uppercase(u32 code_point)
{
    return to_ascii_uppercase(code_point);
}

String to_unicode_lowercase_full(StringView string, [[maybe_unused]] Optional<StringView> locale)
{
#if ENABLE_UNICODE_DATA
    Utf8View view { string };
    StringBuilder builder;

    size_t index = 0;
    size_t byte_length = 0;

    for (auto it = view.begin(); it != view.end(); ++it, index += byte_length) {
        u32 code_point = *it;
        byte_length = it.underlying_code_point_length_in_bytes();

        auto const* special_casing = find_matching_special_case(code_point, view, locale, index, byte_length);
        if (!special_casing) {
            builder.append_code_point(to_unicode_lowercase(code_point));
            continue;
        }

        for (size_t i = 0; i < special_casing->lowercase_mapping_size; ++i)
            builder.append_code_point(special_casing->lowercase_mapping[i]);
    }

    return builder.build();
#else
    return string.to_lowercase_string();
#endif
}

String to_unicode_uppercase_full(StringView string, [[maybe_unused]] Optional<StringView> locale)
{
#if ENABLE_UNICODE_DATA
    Utf8View view { string };
    StringBuilder builder;

    size_t index = 0;
    size_t byte_length = 0;

    for (auto it = view.begin(); it != view.end(); ++it, index += byte_length) {
        u32 code_point = *it;
        byte_length = it.underlying_code_point_length_in_bytes();

        auto const* special_casing = find_matching_special_case(code_point, view, locale, index, byte_length);
        if (!special_casing) {
            builder.append_code_point(to_unicode_uppercase(code_point));
            continue;
        }

        for (size_t i = 0; i < special_casing->uppercase_mapping_size; ++i)
            builder.append_code_point(special_casing->uppercase_mapping[i]);
    }

    return builder.build();
#else
    return string.to_uppercase_string();
#endif
}

Optional<GeneralCategory> __attribute__((weak)) general_category_from_string(StringView) { return {}; }
bool __attribute__((weak)) code_point_has_general_category(u32, GeneralCategory) { return {}; }
Optional<Property> __attribute__((weak)) property_from_string(StringView) { return {}; }
bool __attribute__((weak)) code_point_has_property(u32, Property) { return {}; }

bool is_ecma262_property([[maybe_unused]] Property property)
{
#if ENABLE_UNICODE_DATA
    // EMCA-262 only allows a subset of Unicode properties: https://tc39.es/ecma262/#table-binary-unicode-properties
    switch (property) {
    case Unicode::Property::ASCII:
    case Unicode::Property::ASCII_Hex_Digit:
    case Unicode::Property::Alphabetic:
    case Unicode::Property::Any:
    case Unicode::Property::Assigned:
    case Unicode::Property::Bidi_Control:
    case Unicode::Property::Bidi_Mirrored:
    case Unicode::Property::Case_Ignorable:
    case Unicode::Property::Cased:
    case Unicode::Property::Changes_When_Casefolded:
    case Unicode::Property::Changes_When_Casemapped:
    case Unicode::Property::Changes_When_Lowercased:
    case Unicode::Property::Changes_When_NFKC_Casefolded:
    case Unicode::Property::Changes_When_Titlecased:
    case Unicode::Property::Changes_When_Uppercased:
    case Unicode::Property::Dash:
    case Unicode::Property::Default_Ignorable_Code_Point:
    case Unicode::Property::Deprecated:
    case Unicode::Property::Diacritic:
    case Unicode::Property::Emoji:
    case Unicode::Property::Emoji_Component:
    case Unicode::Property::Emoji_Modifier:
    case Unicode::Property::Emoji_Modifier_Base:
    case Unicode::Property::Emoji_Presentation:
    case Unicode::Property::Extended_Pictographic:
    case Unicode::Property::Extender:
    case Unicode::Property::Grapheme_Base:
    case Unicode::Property::Grapheme_Extend:
    case Unicode::Property::Hex_Digit:
    case Unicode::Property::IDS_Binary_Operator:
    case Unicode::Property::IDS_Trinary_Operator:
    case Unicode::Property::ID_Continue:
    case Unicode::Property::ID_Start:
    case Unicode::Property::Ideographic:
    case Unicode::Property::Join_Control:
    case Unicode::Property::Logical_Order_Exception:
    case Unicode::Property::Lowercase:
    case Unicode::Property::Math:
    case Unicode::Property::Noncharacter_Code_Point:
    case Unicode::Property::Pattern_Syntax:
    case Unicode::Property::Pattern_White_Space:
    case Unicode::Property::Quotation_Mark:
    case Unicode::Property::Radical:
    case Unicode::Property::Regional_Indicator:
    case Unicode::Property::Sentence_Terminal:
    case Unicode::Property::Soft_Dotted:
    case Unicode::Property::Terminal_Punctuation:
    case Unicode::Property::Unified_Ideograph:
    case Unicode::Property::Uppercase:
    case Unicode::Property::Variation_Selector:
    case Unicode::Property::White_Space:
    case Unicode::Property::XID_Continue:
    case Unicode::Property::XID_Start:
        return true;
    default:
        return false;
    }
#else
    return false;
#endif
}

Optional<Script> __attribute__((weak)) script_from_string(StringView) { return {}; }
bool __attribute__((weak)) code_point_has_script(u32, Script) { return {}; }
bool __attribute__((weak)) code_point_has_script_extension(u32, Script) { return {}; }

Optional<Block> __attribute__((weak)) block_from_string(StringView) { return {}; }
bool __attribute__((weak)) code_point_has_block(u32, Block) { return {}; }

bool __attribute__((weak)) code_point_has_grapheme_break_property(u32, GraphemeBreakProperty) { return {}; }
bool __attribute__((weak)) code_point_has_word_break_property(u32, WordBreakProperty) { return {}; }
bool __attribute__((weak)) code_point_has_sentence_break_property(u32, SentenceBreakProperty) { return {}; }

Vector<size_t> find_grapheme_segmentation_boundaries([[maybe_unused]] Utf16View const& view)
{
#if ENABLE_UNICODE_DATA
    using GBP = GraphemeBreakProperty;
    Vector<size_t> boundaries;

    // https://www.unicode.org/reports/tr29/#Grapheme_Cluster_Boundary_Rules
    if (view.length_in_code_points() == 0)
        return boundaries;

    auto has_any_gbp = [](u32 code_point, auto&&... properties) {
        return (code_point_has_grapheme_break_property(code_point, properties) || ...);
    };

    // GB1
    boundaries.append(0);

    if (view.length_in_code_points() > 1) {
        auto it = view.begin();
        auto code_point = *it;
        u32 next_code_point;
        auto current_ri_chain = 0;
        auto in_emoji_sequence = false;

        for (++it; it != view.end(); ++it, code_point = next_code_point) {
            next_code_point = *it;

            auto code_point_is_cr = has_any_gbp(code_point, GBP::CR);
            auto next_code_point_is_lf = has_any_gbp(next_code_point, GBP::LF);

            // GB3
            if (code_point_is_cr && next_code_point_is_lf)
                continue;
            // GB4, GB5
            if (code_point_is_cr || next_code_point_is_lf || has_any_gbp(next_code_point, GBP::CR, GBP::Control) || has_any_gbp(code_point, GBP::LF, GBP::Control)) {
                boundaries.append(view.code_unit_offset_of(it));
                continue;
            }

            auto next_code_point_is_v = has_any_gbp(next_code_point, GBP::V);
            auto next_code_point_is_t = has_any_gbp(next_code_point, GBP::T);

            // GB6
            if (has_any_gbp(code_point, GBP::L) && (next_code_point_is_v || has_any_gbp(next_code_point, GBP::L, GBP::LV, GBP::LVT)))
                continue;
            // GB7
            if ((next_code_point_is_v || next_code_point_is_t) && has_any_gbp(code_point, GBP::LV, GBP::V))
                continue;
            // GB8
            if (next_code_point_is_t && has_any_gbp(code_point, GBP::LVT, GBP::T))
                continue;

            auto code_point_is_zwj = has_any_gbp(code_point, GBP::ZWJ);
            if (!in_emoji_sequence && code_point_has_property(code_point, Property::Extended_Pictographic))
                in_emoji_sequence = true;
            else if (in_emoji_sequence && !has_any_gbp(code_point, GBP::Extend) && !code_point_is_zwj)
                in_emoji_sequence = false;

            // GB9
            if (has_any_gbp(next_code_point, GBP::Extend, GBP::ZWJ))
                continue;
            // GB9a
            if (has_any_gbp(next_code_point, GBP::SpacingMark))
                continue;
            // GB9b
            if (has_any_gbp(code_point, GBP::Prepend))
                continue;

            // GB11
            if (in_emoji_sequence && code_point_is_zwj && code_point_has_property(next_code_point, Property::Extended_Pictographic))
                continue;

            auto code_point_is_ri = has_any_gbp(code_point, GBP::Regional_Indicator);
            current_ri_chain = code_point_is_ri ? current_ri_chain + 1 : 0;

            // GB12, GB13
            if (code_point_is_ri && has_any_gbp(next_code_point, GBP::Regional_Indicator) && current_ri_chain % 2 == 1)
                continue;

            // GB999
            boundaries.append(view.code_unit_offset_of(it));
        }
    }

    // GB2
    boundaries.append(view.length_in_code_units());
    return boundaries;
#else
    return {};
#endif
}

Vector<size_t> find_word_segmentation_boundaries([[maybe_unused]] Utf16View const& view)
{
#if ENABLE_UNICODE_DATA
    using WBP = WordBreakProperty;
    Vector<size_t> boundaries;

    // https://www.unicode.org/reports/tr29/#Word_Boundary_Rules
    if (view.length_in_code_points() == 0)
        return boundaries;

    auto has_any_wbp = [](u32 code_point, auto&&... properties) {
        return (code_point_has_word_break_property(code_point, properties) || ...);
    };

    // WB1
    boundaries.append(0);

    if (view.length_in_code_points() > 1) {
        auto it = view.begin();
        auto code_point = *it;
        u32 next_code_point;
        Optional<u32> previous_code_point;
        auto current_ri_chain = 0;

        for (++it; it != view.end(); ++it, previous_code_point = code_point, code_point = next_code_point) {
            next_code_point = *it;

            auto code_point_is_cr = has_any_wbp(code_point, WBP::CR);
            auto next_code_point_is_lf = has_any_wbp(next_code_point, WBP::LF);

            // WB3
            if (code_point_is_cr && next_code_point_is_lf)
                continue;
            // WB3a, WB3b
            if (code_point_is_cr || next_code_point_is_lf || has_any_wbp(next_code_point, WBP::CR, WBP::Newline) || has_any_wbp(code_point, WBP::LF, WBP::Newline)) {
                boundaries.append(view.code_unit_offset_of(it));
                continue;
            }
            // WB3c
            if (has_any_wbp(code_point, WBP::ZWJ) && code_point_has_property(next_code_point, Property::Extended_Pictographic))
                continue;
            // WB3d
            if (has_any_wbp(code_point, WBP::WSegSpace) && has_any_wbp(next_code_point, WBP::WSegSpace))
                continue;

            // WB4
            if (has_any_wbp(next_code_point, WBP::Format, WBP::Extend, WBP::ZWJ))
                continue;

            auto code_point_is_hebrew_letter = has_any_wbp(code_point, WBP::Hebrew_Letter);
            auto code_point_is_ah_letter = code_point_is_hebrew_letter || has_any_wbp(code_point, WBP::ALetter);
            auto next_code_point_is_hebrew_letter = has_any_wbp(next_code_point, WBP::Hebrew_Letter);
            auto next_code_point_is_ah_letter = next_code_point_is_hebrew_letter || has_any_wbp(next_code_point, WBP::ALetter);

            // WB5
            if (code_point_is_ah_letter && next_code_point_is_ah_letter)
                continue;

            Optional<u32> next_next_code_point;
            if (it != view.end()) {
                auto it_copy = it;
                ++it_copy;
                if (it_copy != view.end())
                    next_next_code_point = *it;
            }
            bool next_next_code_point_is_hebrew_letter = next_next_code_point.has_value() && has_any_wbp(*next_next_code_point, WBP::Hebrew_Letter);
            bool next_next_code_point_is_ah_letter = next_next_code_point_is_hebrew_letter || (next_next_code_point.has_value() && has_any_wbp(*next_next_code_point, WBP::ALetter));

            auto next_code_point_is_mid_num_let_q = has_any_wbp(next_code_point, WBP::MidNumLet, WBP::Single_Quote);

            // WB6
            if (code_point_is_ah_letter && next_next_code_point_is_ah_letter && (next_code_point_is_mid_num_let_q || has_any_wbp(next_code_point, WBP::MidLetter)))
                continue;

            auto code_point_is_mid_num_let_q = has_any_wbp(code_point, WBP::MidNumLet, WBP::Single_Quote);
            auto previous_code_point_is_hebrew_letter = previous_code_point.has_value() && has_any_wbp(*previous_code_point, WBP::Hebrew_Letter);
            auto previous_code_point_is_ah_letter = previous_code_point_is_hebrew_letter || (previous_code_point.has_value() && has_any_wbp(*previous_code_point, WBP::ALetter));

            // WB7
            if (previous_code_point_is_ah_letter && next_code_point_is_ah_letter && (code_point_is_mid_num_let_q || has_any_wbp(code_point, WBP::MidLetter)))
                continue;
            // WB7a
            if (code_point_is_hebrew_letter && has_any_wbp(next_code_point, WBP::Single_Quote))
                continue;
            // WB7b
            if (code_point_is_hebrew_letter && next_next_code_point_is_hebrew_letter && has_any_wbp(next_code_point, WBP::Double_Quote))
                continue;
            // WB7c
            if (previous_code_point_is_hebrew_letter && next_code_point_is_hebrew_letter && has_any_wbp(code_point, WBP::Double_Quote))
                continue;

            auto code_point_is_numeric = has_any_wbp(code_point, WBP::Numeric);
            auto next_code_point_is_numeric = has_any_wbp(next_code_point, WBP::Numeric);

            // WB8
            if (code_point_is_numeric && next_code_point_is_numeric)
                continue;
            // WB9
            if (code_point_is_ah_letter && next_code_point_is_numeric)
                continue;
            // WB10
            if (code_point_is_numeric && next_code_point_is_ah_letter)
                continue;

            auto previous_code_point_is_numeric = previous_code_point.has_value() && has_any_wbp(code_point, WBP::Numeric);

            // WB11
            if (previous_code_point_is_numeric && next_code_point_is_numeric && (code_point_is_mid_num_let_q || has_any_wbp(code_point, WBP::MidNum)))
                continue;

            bool next_next_code_point_is_numeric = next_next_code_point.has_value() && has_any_wbp(*next_next_code_point, WBP::Numeric);

            // WB12
            if (code_point_is_numeric && next_next_code_point_is_numeric && (next_code_point_is_mid_num_let_q || has_any_wbp(next_code_point, WBP::MidNum)))
                continue;

            auto code_point_is_katakana = has_any_wbp(code_point, WBP::Katakana);
            auto next_code_point_is_katakana = has_any_wbp(next_code_point, WBP::Katakana);

            // WB13
            if (code_point_is_katakana && next_code_point_is_katakana)
                continue;

            auto code_point_is_extend_num_let = has_any_wbp(code_point, WBP::ExtendNumLet);

            // WB13a
            if ((code_point_is_ah_letter || code_point_is_numeric || code_point_is_katakana || code_point_is_extend_num_let) && has_any_wbp(next_code_point, WBP::ExtendNumLet))
                continue;
            // WB13b
            if (code_point_is_extend_num_let && (next_code_point_is_ah_letter || next_code_point_is_numeric || next_code_point_is_katakana))
                continue;

            auto code_point_is_ri = has_any_wbp(code_point, WBP::Regional_Indicator);
            current_ri_chain = code_point_is_ri ? current_ri_chain + 1 : 0;

            // WB15, WB16
            if (code_point_is_ri && has_any_wbp(next_code_point, WBP::Regional_Indicator) && current_ri_chain % 2 == 1)
                continue;

            // WB999
            boundaries.append(view.code_unit_offset_of(it));
        }
    }

    // WB2
    boundaries.append(view.length_in_code_units());
    return boundaries;
#else
    return {};
#endif
}

Vector<size_t> find_sentence_segmentation_boundaries([[maybe_unused]] Utf16View const& view)
{
#if ENABLE_UNICODE_DATA
    using SBP = SentenceBreakProperty;
    Vector<size_t> boundaries;

    // https://www.unicode.org/reports/tr29/#Grapheme_Cluster_Boundary_Rules
    if (view.length_in_code_points() == 0)
        return boundaries;

    auto has_any_sbp = [](u32 code_point, auto&&... properties) {
        return (code_point_has_sentence_break_property(code_point, properties) || ...);
    };

    // SB1
    boundaries.append(0);

    if (view.length_in_code_points() > 1) {
        auto it = view.begin();
        auto code_point = *it;
        u32 next_code_point;
        Optional<u32> previous_code_point;
        enum class TerminatorSequenceState {
            None,
            Term,
            Close,
            Sp
        } terminator_sequence_state { TerminatorSequenceState::None };
        auto term_was_a_term = false;

        for (++it; it != view.end(); ++it, previous_code_point = code_point, code_point = next_code_point) {
            next_code_point = *it;

            auto code_point_is_cr = has_any_sbp(code_point, SBP::CR);
            auto next_code_point_is_lf = has_any_sbp(next_code_point, SBP::LF);

            // SB3
            if (code_point_is_cr && next_code_point_is_lf)
                continue;

            auto code_point_is_para_sep = code_point_is_cr || has_any_sbp(code_point, SBP::LF, SBP::Sep);

            // SB4
            if (code_point_is_para_sep) {
                boundaries.append(view.code_unit_offset_of(it));
                continue;
            }

            // SB5
            if (has_any_sbp(next_code_point, SBP::Format, SBP::Extend))
                continue;

            auto code_point_is_a_term = has_any_sbp(code_point, SBP::ATerm);

            // SB6
            if (code_point_is_a_term && has_any_sbp(next_code_point, SBP::Numeric))
                continue;
            // SB7
            if (code_point_is_a_term && previous_code_point.has_value() && has_any_sbp(*previous_code_point, SBP::Upper, SBP::Lower) && has_any_sbp(next_code_point, SBP::Upper))
                continue;

            if (code_point_is_a_term || has_any_sbp(code_point, SBP::STerm)) {
                terminator_sequence_state = TerminatorSequenceState::Term;
                term_was_a_term = code_point_is_a_term;
            } else if (terminator_sequence_state >= TerminatorSequenceState::Term && terminator_sequence_state <= TerminatorSequenceState::Close && has_any_sbp(code_point, SBP::Close)) {
                terminator_sequence_state = TerminatorSequenceState::Close;
            } else if (terminator_sequence_state >= TerminatorSequenceState::Term && has_any_sbp(code_point, SBP::Sp)) {
                terminator_sequence_state = TerminatorSequenceState::Sp;
            } else {
                terminator_sequence_state = TerminatorSequenceState::None;
            }

            // SB8
            if (terminator_sequence_state >= TerminatorSequenceState::Term && term_was_a_term) {
                auto it_copy = it;
                bool illegal_sequence = false;
                for (auto sequence_code_point = *it_copy; it_copy != view.end(); ++it_copy) {
                    if (has_any_sbp(sequence_code_point, SBP::Close, SBP::SContinue, SBP::Numeric, SBP::Sp, SBP::Format, SBP::Extend))
                        continue;
                    illegal_sequence = has_any_sbp(sequence_code_point, SBP::Lower);
                }
                if (illegal_sequence)
                    continue;
            }

            // SB8a
            if (terminator_sequence_state >= TerminatorSequenceState::Term && (has_any_sbp(next_code_point, SBP::SContinue, SBP::STerm, SBP::ATerm)))
                continue;

            auto next_code_point_is_sp = has_any_sbp(next_code_point, SBP::Sp);
            auto next_code_point_is_para_sep = has_any_sbp(next_code_point, SBP::Sep, SBP::CR, SBP::LF);

            // SB9
            if (terminator_sequence_state >= TerminatorSequenceState::Term && terminator_sequence_state <= TerminatorSequenceState::Close && (next_code_point_is_sp || next_code_point_is_para_sep || has_any_sbp(next_code_point, SBP::Close)))
                continue;

            // SB10
            if (terminator_sequence_state >= TerminatorSequenceState::Term && (next_code_point_is_sp || next_code_point_is_para_sep))
                continue;

            // SB11
            if (terminator_sequence_state >= TerminatorSequenceState::Term)
                boundaries.append(view.code_unit_offset_of(it));

            // SB998
        }
    }

    // SB2
    boundaries.append(view.length_in_code_units());
    return boundaries;
#else
    return {};
#endif
}

}