/*
 * Copyright (C) 2003 - 2021 Apple Inc. All rights reserved.
 * Copyright (C) 2008 Holger Hans Peter Freyther
 *
 * This library is free software; you can redistribute it and/or
 * modify it under the terms of the GNU Library General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This library is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * Library General Public License for more details.
 *
 * You should have received a copy of the GNU Library General Public License
 * along with this library; see the file COPYING.LIB.  If not, write to
 * the Free Software Foundation, Inc., 51 Franklin Street, Fifth Floor,
 * Boston, MA 02110-1301, USA.
 *
 */

#include "config.h"
#include "WidthIterator.h"

#include "CharacterProperties.h"
#include "ComposedCharacterClusterTextIterator.h"
#include "Font.h"
#include "FontCascade.h"
#include "GlyphBuffer.h"
#include "Latin1TextIterator.h"
#include "SurrogatePairAwareTextIterator.h"
#include <algorithm>
#include <wtf/MathExtras.h>

namespace WebCore {

using namespace WTF::Unicode;

WidthIterator::WidthIterator(const FontCascade& font, const TextRun& run, HashSet<const Font*>* fallbackFonts, bool accountForGlyphBounds, bool forTextEmphasis)
    : m_font(font)
    , m_run(run)
    , m_fallbackFonts(fallbackFonts)
    , m_expansion(run.expansion())
    , m_isAfterExpansion(run.expansionBehavior().left == ExpansionBehavior::Behavior::Forbid)
    , m_accountForGlyphBounds(accountForGlyphBounds)
    , m_enableKerning(font.enableKerning())
    , m_requiresShaping(font.requiresShaping())
    , m_forTextEmphasis(forTextEmphasis)
{
    // FIXME: Should we clamp m_expansion so it can never be negative?

    if (!m_expansion)
        m_expansionPerOpportunity = 0;
    else {
        unsigned expansionOpportunityCount = FontCascade::expansionOpportunityCount(m_run.text(), m_run.ltr() ? TextDirection::LTR : TextDirection::RTL, run.expansionBehavior()).first;

        if (!expansionOpportunityCount)
            m_expansionPerOpportunity = 0;
        else
            m_expansionPerOpportunity = m_expansion / expansionOpportunityCount;
    }
}

struct OriginalAdvancesForCharacterTreatedAsSpace {
    explicit OriginalAdvancesForCharacterTreatedAsSpace(GlyphBufferStringOffset stringOffset, bool isSpace, float advanceBefore, float advanceAt)
        : stringOffset(stringOffset)
        , characterIsSpace(isSpace)
        , advanceBeforeCharacter(advanceBefore)
        , advanceAtCharacter(advanceAt)
    {
    }

    GlyphBufferStringOffset stringOffset;
    bool characterIsSpace;
    float advanceBeforeCharacter;
    float advanceAtCharacter;
};

inline auto WidthIterator::applyFontTransforms(GlyphBuffer& glyphBuffer, unsigned lastGlyphCount, const Font& font, CharactersTreatedAsSpace& charactersTreatedAsSpace) -> ApplyFontTransformsResult
{
    auto glyphBufferSize = glyphBuffer.size();
    ASSERT(lastGlyphCount <= glyphBufferSize);
    if (lastGlyphCount >= glyphBufferSize)
        return { 0, makeGlyphBufferAdvance() };

    GlyphBufferAdvance* advances = glyphBuffer.advances(0);
    float beforeWidth = 0;
    for (unsigned i = lastGlyphCount; i < glyphBufferSize; ++i)
        beforeWidth += width(advances[i]);

    auto initialAdvance = font.applyTransforms(glyphBuffer, lastGlyphCount, m_currentCharacterIndex, m_enableKerning, m_requiresShaping, m_font.fontDescription().computedLocale(), m_run.text(), m_run.direction());

#if USE(CTFONTSHAPEGLYPHS_WORKAROUND)
    // <rdar://problem/80798113>: If a character is not in BMP, and we don't have a glyph for it,
    // we'll end up with two 0 glyphs in a row for the two surrogates of the character.
    // We need to make sure that, after shaping, these double-0-glyphs aren't preserved.
    if (&font == &m_font.primaryFont() && !m_run.text().is8Bit()) {
        for (unsigned i = 0; i < glyphBuffer.size() - 1; ++i) {
            if (!glyphBuffer.glyphAt(i) && !glyphBuffer.glyphAt(i + 1)) {
                if (const auto& firstStringOffset = glyphBuffer.checkedStringOffsetAt(i, m_run.length())) {
                    if (const auto& secondStringOffset = glyphBuffer.checkedStringOffsetAt(i + 1, m_run.length())) {
                        if (secondStringOffset.value() == firstStringOffset.value() + 1
                            && U_IS_LEAD(m_run.text()[firstStringOffset.value()]) && U_IS_TRAIL(m_run.text()[secondStringOffset.value()])) {
                            glyphBuffer.remove(i + 1, 1);
                        }
                    }
                }
            }
        }
    }
#endif

    glyphBufferSize = glyphBuffer.size();
    advances = glyphBuffer.advances(0);

    GlyphBufferOrigin* origins = glyphBuffer.origins(0);
    for (unsigned i = lastGlyphCount; i < glyphBufferSize; ++i) {
        setHeight(advances[i], -height(advances[i]));
        setY(origins[i], -y(origins[i]));
    }

    for (unsigned i = lastGlyphCount; i < glyphBufferSize; ++i) {
        auto characterIndex = glyphBuffer.uncheckedStringOffsetAt(i);
        auto iterator = std::lower_bound(charactersTreatedAsSpace.begin(), charactersTreatedAsSpace.end(), characterIndex, [](const OriginalAdvancesForCharacterTreatedAsSpace& l, GlyphBufferStringOffset r) -> bool {
            return l.stringOffset < r;
        });
        if (iterator == charactersTreatedAsSpace.end() || iterator->stringOffset != characterIndex)
            continue;
        const auto& originalAdvances = *iterator;
        if (i && !originalAdvances.characterIsSpace)
            setWidth(*glyphBuffer.advances(i - 1), originalAdvances.advanceBeforeCharacter);
        setWidth(*glyphBuffer.advances(i), originalAdvances.advanceAtCharacter);
    }
    charactersTreatedAsSpace.clear();

    float afterWidth = 0;
    for (unsigned i = lastGlyphCount; i < glyphBufferSize; ++i)
        afterWidth += width(advances[i]);

    return { afterWidth - beforeWidth, initialAdvance };
}

static inline std::pair<bool, bool> expansionLocation(bool ideograph, bool treatAsSpace, bool ltr, bool isAfterExpansion, bool forbidLeftExpansion, bool forbidRightExpansion, bool forceLeftExpansion, bool forceRightExpansion)
{
    bool expandLeft = ideograph;
    bool expandRight = ideograph;
    if (treatAsSpace) {
        if (ltr)
            expandRight = true;
        else
            expandLeft = true;
    }
    if (isAfterExpansion) {
        if (ltr)
            expandLeft = false;
        else
            expandRight = false;
    }
    ASSERT(!forbidLeftExpansion || !forceLeftExpansion);
    ASSERT(!forbidRightExpansion || !forceRightExpansion);
    if (forbidLeftExpansion)
        expandLeft = false;
    if (forbidRightExpansion)
        expandRight = false;
    if (forceLeftExpansion)
        expandLeft = true;
    if (forceRightExpansion)
        expandRight = true;
    return std::make_pair(expandLeft, expandRight);
}

static void expandWithInitialAdvance(GlyphBufferAdvance& advanceToExpand, const GlyphBufferAdvance& initialAdvance)
{
    setWidth(advanceToExpand, width(advanceToExpand) + width(initialAdvance));
    setHeight(advanceToExpand, height(advanceToExpand) + height(initialAdvance));
}

void WidthIterator::applyInitialAdvance(GlyphBuffer& glyphBuffer, GlyphBufferAdvance initialAdvance, unsigned lastGlyphCount)
{
    ASSERT(glyphBuffer.size() >= lastGlyphCount);

    if (glyphBuffer.size() <= lastGlyphCount)
        return;

    ASSERT(lastGlyphCount || (!width(m_leftoverInitialAdvance) && !height(m_leftoverInitialAdvance)));

    if (m_run.direction() == TextDirection::RTL && lastGlyphCount) {
        auto& visuallyLastAdvance = *glyphBuffer.advances(lastGlyphCount);
        expandWithInitialAdvance(visuallyLastAdvance, m_leftoverInitialAdvance);
        m_runWidthSoFar += width(m_leftoverInitialAdvance);
        m_leftoverInitialAdvance = makeGlyphBufferAdvance();
    }

    if (m_run.direction() == TextDirection::RTL)
        m_leftoverInitialAdvance = initialAdvance;
    else {
        if (lastGlyphCount) {
            auto& visuallyPreviousAdvance = *glyphBuffer.advances(lastGlyphCount - 1);
            expandWithInitialAdvance(visuallyPreviousAdvance, initialAdvance);
            m_runWidthSoFar += width(initialAdvance);
        } else
            glyphBuffer.expandInitialAdvance(initialAdvance);
    }
}

void WidthIterator::commitCurrentFontRange(GlyphBuffer& glyphBuffer, unsigned& lastGlyphCount, unsigned currentCharacterIndex, const Font*& font, const Font& newFont, const Font& primaryFont, UChar32 character, float& widthOfCurrentFontRange, float nextCharacterWidth, CharactersTreatedAsSpace& charactersTreatedAsSpace)
{
#if ASSERT_ENABLED
    ASSERT(font);
    for (unsigned i = lastGlyphCount; i < glyphBuffer.size(); ++i)
        ASSERT(&glyphBuffer.fontAt(i) == font);
#endif

    auto applyFontTransformsResult = applyFontTransforms(glyphBuffer, lastGlyphCount, *font, charactersTreatedAsSpace);
    m_runWidthSoFar += applyFontTransformsResult.additionalAdvance;
    applyInitialAdvance(glyphBuffer, applyFontTransformsResult.initialAdvance, lastGlyphCount);
    m_currentCharacterIndex = currentCharacterIndex;

    if (widthOfCurrentFontRange && m_fallbackFonts && font != &primaryFont) {
        // FIXME: This does a little extra work that could be avoided if
        // glyphDataForCharacter() returned whether it chose to use a small caps font.
        if (!m_font.isSmallCaps() || character == u_toupper(character))
            m_fallbackFonts->add(font);
        else {
            auto glyphFont = m_font.glyphDataForCharacter(u_toupper(character), m_run.rtl()).font;
            if (glyphFont != &primaryFont)
                m_fallbackFonts->add(glyphFont);
        }
    }

    lastGlyphCount = glyphBuffer.size();
    font = &newFont;
    widthOfCurrentFontRange = nextCharacterWidth;
}

bool WidthIterator::hasExtraSpacing() const
{
    return (m_font.letterSpacing() || m_font.wordSpacing() || m_expansion) && !m_run.spacingDisabled();
}

static void addToGlyphBuffer(GlyphBuffer& glyphBuffer, Glyph glyph, const Font& font, float width, GlyphBufferStringOffset currentCharacterIndex, UChar32 character)
{
    glyphBuffer.add(glyph, font, width, currentCharacterIndex);
#if USE(CTFONTSHAPEGLYPHS)
    // These 0 glyphs are needed by shapers if the source text has surrogate pairs.
    // However, CTFontTransformGlyphs() can't delete these 0 glyphs from the shaped text,
    // so we shouldn't add them in the first place if we're using that shaping routine.
    // Any other shaping routine should delete these glyphs from the shaped text.
    if (!U_IS_BMP(character))
        glyphBuffer.add(0, font, 0, currentCharacterIndex + 1);
#else
    UNUSED_PARAM(character);
#endif
}

template <typename TextIterator>
inline void WidthIterator::advanceInternal(TextIterator& textIterator, GlyphBuffer& glyphBuffer)
{
    // The core logic here needs to match FontCascade::widthForSimpleText()
    bool rtl = m_run.rtl();

    FloatRect bounds;

    const Font& primaryFont = m_font.primaryFont();
    const Font* lastFontData = &primaryFont;
    unsigned lastGlyphCount = glyphBuffer.size();

    auto currentCharacterIndex = textIterator.currentIndex();
    UChar32 character = 0;
    float width = 0;
    float previousWidth = 0;
    unsigned clusterLength = 0;
    CharactersTreatedAsSpace charactersTreatedAsSpace;
    float widthOfCurrentFontRange = 0;
    // We are iterating in string order, not glyph order. Compare this to ComplexTextController::adjustGlyphsAndAdvances()
    while (textIterator.consume(character, clusterLength)) {
        // FIXME: Should we replace unpaired surrogates with the object replacement character?
        // Should we do this before or after shaping? What does a shaper do with an unpaired surrogate?
        m_containsTabs |= character == tabCharacter;
        currentCharacterIndex = textIterator.currentIndex();
        unsigned advanceLength = clusterLength;
        if (currentCharacterIndex + advanceLength == m_run.length())
            m_lastCharacterIndex = currentCharacterIndex;
        bool characterMustDrawSomething = !isDefaultIgnorableCodePoint(character);
#if USE(FREETYPE)
        // Freetype based ports only override the characters with Default_Ignorable unicode property when the font
        // doesn't support the code point. We should ignore them at this point to ensure they are not displayed.
        if (!characterMustDrawSomething) {
            textIterator.advance(advanceLength);
            currentCharacterIndex = textIterator.currentIndex();
            continue;
        }
#endif
        const GlyphData& glyphData = m_font.glyphDataForCharacter(character, rtl);
        Glyph glyph = glyphData.glyph;
        if (!glyph && !characterMustDrawSomething) {
            commitCurrentFontRange(glyphBuffer, lastGlyphCount, currentCharacterIndex, lastFontData, primaryFont, primaryFont, character, widthOfCurrentFontRange, width, charactersTreatedAsSpace);

            Glyph deletedGlyph = 0xFFFF;
            addToGlyphBuffer(glyphBuffer, deletedGlyph, primaryFont, 0, currentCharacterIndex, character);

            textIterator.advance(advanceLength);
            currentCharacterIndex = textIterator.currentIndex();
            continue;
        }
        const Font& font = glyphData.font ? *glyphData.font : primaryFont;

        previousWidth = width;
        width = font.widthForGlyph(glyph, Font::SyntheticBoldInclusion::Exclude); // We apply synthetic bold after shaping, in applyCSSVisibilityRules().

        if (&font != lastFontData)
            commitCurrentFontRange(glyphBuffer, lastGlyphCount, currentCharacterIndex, lastFontData, font, primaryFont, character, widthOfCurrentFontRange, width, charactersTreatedAsSpace);
        else
            widthOfCurrentFontRange += width;

        if (FontCascade::treatAsSpace(character))
            charactersTreatedAsSpace.constructAndAppend(currentCharacterIndex, character == space, previousWidth, character == tabCharacter ? width : font.spaceWidth(Font::SyntheticBoldInclusion::Exclude));

        if (m_accountForGlyphBounds) {
            bounds = font.boundsForGlyph(glyph);
            if (!currentCharacterIndex)
                m_firstGlyphOverflow = std::max<float>(0, -bounds.x());
        }

        if (m_forTextEmphasis && !FontCascade::canReceiveTextEmphasis(character))
            glyph = 0;

        addToGlyphBuffer(glyphBuffer, glyph, font, width, currentCharacterIndex, character);

        // Advance past the character we just dealt with.
        textIterator.advance(advanceLength);
        currentCharacterIndex = textIterator.currentIndex();

        m_runWidthSoFar += width;

        if (m_accountForGlyphBounds) {
            m_maxGlyphBoundingBoxY = std::max(m_maxGlyphBoundingBoxY, bounds.maxY());
            m_minGlyphBoundingBoxY = std::min(m_minGlyphBoundingBoxY, bounds.y());
            m_lastGlyphOverflow = std::max<float>(0, bounds.maxX() - width);
        }
    }

    commitCurrentFontRange(glyphBuffer, lastGlyphCount, currentCharacterIndex, lastFontData, primaryFont, primaryFont, character, widthOfCurrentFontRange, width, charactersTreatedAsSpace);
}

auto WidthIterator::calculateAdditionalWidth(GlyphBuffer& glyphBuffer, GlyphBufferStringOffset currentCharacterIndex, unsigned leadingGlyphIndex, unsigned trailingGlyphIndex, float position) const -> AdditionalWidth
{
    float leftAdditionalWidth = 0;
    float rightAdditionalWidth = 0;
    float leftExpansionAdditionalWidth = 0;
    float rightExpansionAdditionalWidth = 0;

    auto character = m_run[currentCharacterIndex];

    if (character == tabCharacter && m_run.allowTabs()) {
        auto& font = glyphBuffer.fontAt(trailingGlyphIndex);
        // Synthetic bold will be handled in applyCSSVisibilityRules() later.
        auto newWidth = m_font.tabWidth(font, m_run.tabSize(), position, Font::SyntheticBoldInclusion::Exclude);
        auto currentWidth = width(glyphBuffer.advanceAt(trailingGlyphIndex));
        rightAdditionalWidth += newWidth - currentWidth;
    }

    if (hasExtraSpacing()) {
        bool treatAsSpace = FontCascade::treatAsSpace(character);

        // This is a heuristic to determine if the character is non-visible. Non-visible characters don't get letter-spacing.
        float baseWidth = 0;
        for (unsigned i = leadingGlyphIndex; i <= trailingGlyphIndex; ++i)
            baseWidth += width(glyphBuffer.advanceAt(i));
        if (baseWidth)
            rightAdditionalWidth += m_font.letterSpacing();

        if (treatAsSpace && (character != tabCharacter || !m_run.allowTabs()) && (currentCharacterIndex || character == noBreakSpace) && m_font.wordSpacing())
            rightAdditionalWidth += m_font.wordSpacing();

        if (m_expansion > 0) {
            bool currentIsLastCharacter = m_lastCharacterIndex && currentCharacterIndex == static_cast<GlyphBufferStringOffset>(*m_lastCharacterIndex);

            bool isLeftmostCharacter = !currentCharacterIndex;
            bool isRightmostCharacter = currentIsLastCharacter;
            if (!m_run.ltr())
                std::swap(isLeftmostCharacter, isRightmostCharacter);

            bool forceLeftExpansion = isLeftmostCharacter && m_run.expansionBehavior().left == ExpansionBehavior::Behavior::Force;
            bool forceRightExpansion = isRightmostCharacter && m_run.expansionBehavior().right == ExpansionBehavior::Behavior::Force;
            bool forbidLeftExpansion = isLeftmostCharacter && m_run.expansionBehavior().left == ExpansionBehavior::Behavior::Forbid;
            bool forbidRightExpansion = isRightmostCharacter && m_run.expansionBehavior().right == ExpansionBehavior::Behavior::Forbid;

            bool isIdeograph = FontCascade::canExpandAroundIdeographsInComplexText() && FontCascade::isCJKIdeographOrSymbol(character);

            if (treatAsSpace || isIdeograph || forceLeftExpansion || forceRightExpansion) {
                auto [expandLeft, expandRight] = expansionLocation(isIdeograph, treatAsSpace, m_run.ltr(), m_isAfterExpansion, forbidLeftExpansion, forbidRightExpansion, forceLeftExpansion, forceRightExpansion);

                if (expandLeft)
                    leftExpansionAdditionalWidth += m_expansionPerOpportunity;
                if (expandRight)
                    rightExpansionAdditionalWidth += m_expansionPerOpportunity;
            }
        }
    }

    return { leftAdditionalWidth, rightAdditionalWidth, leftExpansionAdditionalWidth, rightExpansionAdditionalWidth };
}

struct GlyphIndexRange {
    // This means the character got expanded to glyphs inside the GlyphBuffer at indices [leadingGlyphIndex, trailingGlyphIndex].
    unsigned leadingGlyphIndex;
    unsigned trailingGlyphIndex;
};

void WidthIterator::applyAdditionalWidth(GlyphBuffer& glyphBuffer, GlyphIndexRange glyphIndexRange, float leftAdditionalWidth, float rightAdditionalWidth, float leftExpansionAdditionalWidth, float rightExpansionAdditionalWidth)
{
    m_expansion -= leftExpansionAdditionalWidth + rightExpansionAdditionalWidth;

    leftAdditionalWidth += leftExpansionAdditionalWidth;
    rightAdditionalWidth += rightExpansionAdditionalWidth;

    m_runWidthSoFar += leftAdditionalWidth;
    m_runWidthSoFar += rightAdditionalWidth;

    if (leftAdditionalWidth) {
        if (m_run.ltr()) {
            // Left additional width in LTR means the previous (leading) glyph's right side gets expanded.
            auto leadingGlyphIndex = glyphIndexRange.leadingGlyphIndex;
            if (leadingGlyphIndex)
                glyphBuffer.expandAdvance(leadingGlyphIndex - 1, leftAdditionalWidth);
            else
                glyphBuffer.expandInitialAdvance(leftAdditionalWidth);
        } else {
            // Left additional width in RTL means the next (trailing) glyph's right side gets expanded.
            auto trailingGlyphIndex = glyphIndexRange.trailingGlyphIndex;
            if (trailingGlyphIndex + 1 < glyphBuffer.size())
                glyphBuffer.expandAdvance(trailingGlyphIndex + 1, leftAdditionalWidth);
            else {
                m_leftoverJustificationWidth = leftAdditionalWidth;
                // We can't actually add in this width just yet.
                // Add it in when the client calls advance() again or finalize().
                m_runWidthSoFar -= m_leftoverJustificationWidth;
            }
        }
    }

    if (rightAdditionalWidth) {
        // Right additional width means the current glyph's right side gets expanded. This is true for both LTR and RTL.
        glyphBuffer.expandAdvance(glyphIndexRange.trailingGlyphIndex, rightAdditionalWidth);
    }
}

void WidthIterator::applyExtraSpacingAfterShaping(GlyphBuffer& glyphBuffer, unsigned characterStartIndex, unsigned glyphBufferStartIndex, unsigned characterDestinationIndex, float startingRunWidth)
{
    Vector<std::optional<GlyphIndexRange>> characterIndexToGlyphIndexRange(m_run.length(), std::nullopt);
    Vector<float> advanceWidths(m_run.length(), 0);
    for (unsigned i = glyphBufferStartIndex; i < glyphBuffer.size(); ++i) {
        auto stringOffset = glyphBuffer.checkedStringOffsetAt(i, m_run.length());
        if (!stringOffset)
            continue;
        advanceWidths[stringOffset.value()] += width(glyphBuffer.advanceAt(i));
        auto& glyphIndexRange = characterIndexToGlyphIndexRange[stringOffset.value()];
        if (glyphIndexRange)
            glyphIndexRange->trailingGlyphIndex = i;
        else
            glyphIndexRange = {{i, i}};
    }

    // SVG can stretch advances
    if (m_run.horizontalGlyphStretch() != 1) {
        for (unsigned i = glyphBufferStartIndex; i < glyphBuffer.size(); ++i) {
            // All characters' advances get stretched, except apparently tab characters...
            // This doesn't make much sense, because even tab characters get letter-spacing...
            auto stringOffset = glyphBuffer.checkedStringOffsetAt(i, m_run.length());
            if (stringOffset && m_run[stringOffset.value()] == tabCharacter)
                continue;

            auto currentAdvance = width(glyphBuffer.advanceAt(i));
            auto newAdvance = currentAdvance * m_run.horizontalGlyphStretch();
            glyphBuffer.expandAdvance(i, newAdvance - currentAdvance);
        }
    }

    float position = m_run.xPos() + startingRunWidth;
    for (auto i = characterStartIndex; i < characterDestinationIndex; ++i) {
        auto& glyphIndexRange = characterIndexToGlyphIndexRange[i];
        if (!glyphIndexRange)
            continue;

        auto width = calculateAdditionalWidth(glyphBuffer, i, glyphIndexRange->leadingGlyphIndex, glyphIndexRange->trailingGlyphIndex, position);
        applyAdditionalWidth(glyphBuffer, glyphIndexRange.value(), width.left, width.right, width.leftExpansion, width.rightExpansion);

        m_isAfterExpansion = (m_run.ltr() && width.rightExpansion) || (!m_run.ltr() && width.leftExpansion);

        // This isn't quite perfect, because we may come across a tab character in between two glyphs which both report to correspond to a previous character.
        // But, the fundamental concept of tabs isn't really compatible with complex text shaping, so this is probably okay.
        // We can probably just do the best we can here.
        // The only alternative, to calculate this position in glyph-space rather than character-space,
        // is O(n^2) because we're iterating across the string here, rather than glyphs, so we can't keep the calculation up-to-date,
        // which means calculateAdditionalWidth() would have to calculate the result from scratch whenever it's needed.
        // And we can't do some sort of prefix-sum thing because applyAdditionalWidth() would modify the values,
        // so updating the data structure each turn of this loop would also end up being O(n^2).
        // Unfortunately, strings with tabs are more likely to be long data-table kind of strings, which means O(n^2) is not acceptable.
        // Also, even if we did the O(n^2) thing, there would still be cases that wouldn't be perfect
        // (because the fundamental concept of tabs isn't really compatible with complex text shaping),
        // so let's choose the fast-wrong approach here instead of the slow-wrong approach.
        position += advanceWidths[i]
            + width.left
            + width.right
            + width.leftExpansion
            + width.rightExpansion;
    }
}

bool WidthIterator::characterCanUseSimplifiedTextMeasuring(UChar character, bool whitespaceIsCollapsed)
{
    // This function needs to be kept in sync with applyCSSVisibilityRules().

    switch (character) {
    case newlineCharacter:
    case carriageReturn:
    case zeroWidthNoBreakSpace:
    case zeroWidthNonJoiner:
    case zeroWidthJoiner:
        return true;
    case tabCharacter:
        if (!whitespaceIsCollapsed)
            return false;
        break;
    case noBreakSpace:
    case softHyphen:
    case leftToRightMark:
    case rightToLeftMark:
    case leftToRightEmbed:
    case rightToLeftEmbed:
    case leftToRightOverride:
    case rightToLeftOverride:
    case leftToRightIsolate:
    case rightToLeftIsolate:
    case popDirectionalFormatting:
    case popDirectionalIsolate:
    case firstStrongIsolate:
    case objectReplacementCharacter:
        return false;
        break;
    }

    if (character >= HiraganaLetterSmallA
        || u_charType(character) == U_CONTROL_CHAR
        || (character >= nullCharacter && character < space)
        || (character >= deleteCharacter && character < noBreakSpace))
        return false;

    return true;
}

void WidthIterator::applyCSSVisibilityRules(GlyphBuffer& glyphBuffer, unsigned glyphBufferStartIndex)
{
    // This function needs to be kept in sync with characterCanUseSimplifiedTextMeasuring().

    Vector<unsigned> glyphsIndicesToBeDeleted;

    float yPosition = height(glyphBuffer.initialAdvance());

    auto adjustForSyntheticBold = [&](auto index) {
        auto glyph = glyphBuffer.glyphAt(index);
        static constexpr const GlyphBufferGlyph deletedGlyph = 0xFFFF;
        auto syntheticBoldOffset = glyph == deletedGlyph ? 0 : glyphBuffer.fontAt(index).syntheticBoldOffset();
        m_runWidthSoFar += syntheticBoldOffset;
        auto& advance = glyphBuffer.advances(index)[0];
        setWidth(advance, width(advance) + syntheticBoldOffset);
    };

    auto clobberGlyph = [&](auto index, auto newGlyph) {
        glyphBuffer.glyphs(index)[0] = newGlyph;
    };

    // FIXME: It's technically wrong to call clobberAdvance or deleteGlyph here, because this is after initialAdvances have been
    // applied. If the last glyph in a run needs to have its advance clobbered, but the next run has an initial advance, we need
    // to apply the initial advance on the new clobbered advance, rather than clobbering the initial advance entirely.

    auto clobberAdvance = [&](auto index, auto newAdvance) {
        auto advanceBeforeClobbering = glyphBuffer.advanceAt(index);
        glyphBuffer.advances(index)[0] = makeGlyphBufferAdvance(newAdvance, height(advanceBeforeClobbering));
        m_runWidthSoFar += width(glyphBuffer.advanceAt(index)) - width(advanceBeforeClobbering);
        glyphBuffer.origins(index)[0] = makeGlyphBufferOrigin(0, -yPosition);
    };

    auto deleteGlyph = [&](auto index) {
        m_runWidthSoFar -= width(glyphBuffer.advanceAt(index));
        glyphBuffer.deleteGlyphWithoutAffectingSize(index);
    };

    auto makeGlyphInvisible = [&](auto index) {
        glyphBuffer.makeGlyphInvisible(index);
    };

    for (unsigned i = glyphBufferStartIndex; i < glyphBuffer.size(); yPosition += height(glyphBuffer.advanceAt(i)), ++i) {
        auto stringOffset = glyphBuffer.checkedStringOffsetAt(i, m_run.length());
        if (!stringOffset)
            continue;
        auto characterResponsibleForThisGlyph = m_run[stringOffset.value()];

        switch (characterResponsibleForThisGlyph) {
        case newlineCharacter:
        case carriageReturn:
            ASSERT(glyphBuffer.fonts(i)[0]);
            // FIXME: It isn't quite right to use the space glyph here, because the space character may be supposed to render with a totally unrelated font (because of fallback).
            // Instead, we should probably somehow have the caller pass in a Font/glyph pair to use in this situation.
            if (auto spaceGlyph = glyphBuffer.fontAt(i).spaceGlyph())
                clobberGlyph(i, spaceGlyph);
            adjustForSyntheticBold(i);
            continue;
        case noBreakSpace:
            adjustForSyntheticBold(i);
            continue;
        case tabCharacter:
            makeGlyphInvisible(i);
            adjustForSyntheticBold(i);
            continue;
        }

        // https://www.w3.org/TR/css-text-3/#white-space-processing
        // "Control characters (Unicode category Cc)—other than tabs (U+0009), line feeds (U+000A), carriage returns (U+000D) and sequences that form a segment break—must be rendered as a visible glyph"
        // Also, we're omitting NULL (U+0000) from this set because Chrome and Firefox do so and it's needed for compat. See https://github.com/w3c/csswg-drafts/pull/6983.
        if (characterResponsibleForThisGlyph != nullCharacter
            && u_charType(characterResponsibleForThisGlyph) == U_CONTROL_CHAR) {
            // Let's assume that .notdef is visible.
            GlyphBufferGlyph visibleGlyph = 0;
            clobberGlyph(i, visibleGlyph);
            clobberAdvance(i, glyphBuffer.fontAt(i).widthForGlyph(visibleGlyph));
            continue;
        }

        adjustForSyntheticBold(i);

        // https://drafts.csswg.org/css-text-3/#white-space-processing
        // "Unsupported Default_ignorable characters must be ignored for text rendering."
        if (FontCascade::isCharacterWhoseGlyphsShouldBeDeletedForTextRendering(characterResponsibleForThisGlyph)) {
            deleteGlyph(i);
            continue;
        }
    }
}

void WidthIterator::finalize(GlyphBuffer& buffer)
{
    ASSERT(m_run.rtl() || !m_leftoverJustificationWidth);
    // In LTR these do nothing. In RTL, these add left width by moving the whole run to the right.
    buffer.expandInitialAdvance(m_leftoverInitialAdvance);
    m_runWidthSoFar += width(m_leftoverInitialAdvance);
    buffer.expandInitialAdvance(m_leftoverJustificationWidth);
    m_runWidthSoFar += m_leftoverJustificationWidth;
    m_leftoverJustificationWidth = 0;
}

void WidthIterator::advance(unsigned offset, GlyphBuffer& glyphBuffer)
{
    m_containsTabs = false;
    unsigned length = m_run.length();

    if (offset > length)
        offset = length;

    if (m_currentCharacterIndex >= offset)
        return;

    unsigned characterStartIndex = m_currentCharacterIndex;
    unsigned glyphBufferStartIndex = glyphBuffer.size();
    float startingRunWidth = m_runWidthSoFar;

    if (m_run.is8Bit()) {
        Latin1TextIterator textIterator(m_run.data8(m_currentCharacterIndex), m_currentCharacterIndex, offset, length);
        advanceInternal(textIterator, glyphBuffer);
    } else {
#if USE(CLUSTER_AWARE_WIDTH_ITERATOR)
        ComposedCharacterClusterTextIterator textIterator(m_run.data16(m_currentCharacterIndex), m_currentCharacterIndex, offset, length);
#else
        SurrogatePairAwareTextIterator textIterator(m_run.data16(m_currentCharacterIndex), m_currentCharacterIndex, offset, length);
#endif
        advanceInternal(textIterator, glyphBuffer);
    }

    // In general, we have to apply spacing after shaping, because shaping requires its input to be unperturbed (see https://bugs.webkit.org/show_bug.cgi?id=215052).
    // So, if there's extra spacing to add, do it here after shaping occurs.
    if (glyphBufferStartIndex < glyphBuffer.size()) {
        glyphBuffer.expandAdvance(glyphBufferStartIndex, m_leftoverJustificationWidth);
        m_runWidthSoFar += m_leftoverJustificationWidth;
        m_leftoverJustificationWidth = 0;
    }

    if (hasExtraSpacing() || m_containsTabs || m_run.horizontalGlyphStretch() != 1)
        applyExtraSpacingAfterShaping(glyphBuffer, characterStartIndex, glyphBufferStartIndex, offset, startingRunWidth);

    applyCSSVisibilityRules(glyphBuffer, glyphBufferStartIndex);
}

bool WidthIterator::advanceOneCharacter(float& width, GlyphBuffer& glyphBuffer)
{
    unsigned oldSize = glyphBuffer.size();
    advance(m_currentCharacterIndex + 1, glyphBuffer);
    float w = 0;
    for (unsigned i = oldSize; i < glyphBuffer.size(); ++i)
        w += WebCore::width(glyphBuffer.advanceAt(i));
    width = w;
    return glyphBuffer.size() > oldSize;
}

}