fenix/PDF4QT
1
0
Fork 0
mirror of https://github.com/JakubMelka/PDF4QT.git synced 2024-12-26 16:22:50 +01:00
Code Issues Packages Projects Releases Wiki Activity

DocDiff application: Diffing text finished

Browse Source
This commit is contained in:
Jakub Melka 2021-09-27 19:40:46 +02:00
parent 5cdb6cdab6
commit 8508fe9ef2
3 changed files with 262 additions and 47 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

191
Pdf4QtLib/sources/pdfdiff.cpp
View File

@ -69,6 +69,7 @@ public:
static std::vector<TextCompareItem> prepareTextCompareItems(const PDFDocumentTextFlow& textFlow,
bool isWordsComparingMode,
bool isLeft);
static void refineTextRectangles(PDFDiffResult::RectInfos& items);
};
PDFDiff::PDFDiff(QObject* parent) :
@ -637,8 +638,10 @@ void PDFDiff::performCompare(const std::vector<PDFDiffPageContext>& leftPrepared
}
}
QMutex mutex;
// Jakub Melka: try to compare text differences
auto compareTexts = [this](PDFDiffHelper::TextFlowDifferences& context)
auto compareTexts = [this, &mutex](PDFDiffHelper::TextFlowDifferences& context)
{
using TextCompareItem = PDFDiffHelper::TextCompareItem;
const bool isWordsComparingMode = m_options.testFlag(CompareWords);
@ -720,6 +723,12 @@ void PDFDiff::performCompare(const std::vector<PDFDiffPageContext>& leftPrepared
QStringList leftStrings;
QStringList rightStrings;
PDFDiffResult::RectInfos leftRectInfos;
PDFDiffResult::RectInfos rightRectInfos;
PDFInteger pageIndex1 = -1;
PDFInteger pageIndex2 = -1;
for (; it != itEnd; ++it)
{
const PDFAlgorithmLongestCommonSubsequenceBase::SequenceItem& item = *it;
@ -728,16 +737,50 @@ void PDFDiff::performCompare(const std::vector<PDFDiffPageContext>& leftPrepared
{
const TextCompareItem& textCompareItem = leftItems[item.index1];
const auto& textFlow = textCompareItem.left ? context.leftTextFlow : context.rightTextFlow;
QStringRef text(&textFlow.getItem(textCompareItem.index)->text, textCompareItem.charIndex, textCompareItem.charCount);
const PDFDocumentTextFlow::Item* textItem = textFlow.getItem(textCompareItem.index);
QStringRef text(&textItem->text, textCompareItem.charIndex, textCompareItem.charCount);
leftStrings << text.toString();
if (pageIndex1 == -1)
{
pageIndex1 = textItem->pageIndex;
}
if (textCompareItem.charIndex + textCompareItem.charCount <= textItem->characterBoundingRects.size())
{
const size_t startIndex = textCompareItem.charIndex;
const size_t endIndex = startIndex + textCompareItem.charCount;
for (size_t i = startIndex; i < endIndex; ++i)
{
leftRectInfos.emplace_back(textItem->pageIndex, textItem->characterBoundingRects[i]);
}
}
}
if (item.isRightValid())
{
const TextCompareItem& textCompareItem = rightItems[item.index2];
const auto& textFlow = textCompareItem.left ? context.leftTextFlow : context.rightTextFlow;
QStringRef text(&textFlow.getItem(textCompareItem.index)->text, textCompareItem.charIndex, textCompareItem.charCount);
const PDFDocumentTextFlow::Item* textItem = textFlow.getItem(textCompareItem.index);
QStringRef text(&textItem->text, textCompareItem.charIndex, textCompareItem.charCount);
rightStrings << text.toString();
if (pageIndex2 == -1)
{
pageIndex2 = textItem->pageIndex;
}
if (textCompareItem.charIndex + textCompareItem.charCount <= textItem->characterBoundingRects.size())
{
const size_t startIndex = textCompareItem.charIndex;
const size_t endIndex = startIndex + textCompareItem.charCount;
for (size_t i = startIndex; i < endIndex; ++i)
{
rightRectInfos.emplace_back(textItem->pageIndex, textItem->characterBoundingRects[i]);
}
}
}
}
@ -755,7 +798,26 @@ void PDFDiff::performCompare(const std::vector<PDFDiffPageContext>& leftPrepared
rightString = rightStrings.join(QString());
}
PDFDiffHelper::refineTextRectangles(leftRectInfos);
PDFDiffHelper::refineTextRectangles(rightRectInfos);
QMutexLocker locker(&mutex);
if (!leftString.isEmpty() && !rightString.isEmpty())
{
m_result.addTextReplaced(pageIndex1, pageIndex2, leftString, rightString, leftRectInfos, rightRectInfos);
}
else
{
if (!leftString.isEmpty())
{
m_result.addTextRemoved(pageIndex1, leftString, leftRectInfos);
}
if (!rightString.isEmpty())
{
m_result.addTextAdded(pageIndex2, rightString, rightRectInfos);
}
}
}
};
@ -929,6 +991,68 @@ void PDFDiffResult::addAddedShadingContent(PDFInteger pageIndex, QRectF rect)
addRightItem(Type::AddedShadingContent, pageIndex, rect);
}
void PDFDiffResult::addTextAdded(PDFInteger pageIndex,
QString text,
const RectInfos& rectInfos)
{
Difference difference;
difference.type = Type::TextAdded;
difference.pageIndex2 = pageIndex;
difference.textAddedIndex = m_strings.size();
m_strings << text;
difference.rightRectIndex = m_rects.size();
difference.rightRectCount = rectInfos.size();
m_rects.insert(m_rects.end(), rectInfos.cbegin(), rectInfos.cend());
m_differences.emplace_back(std::move(difference));
}
void PDFDiffResult::addTextRemoved(PDFInteger pageIndex,
QString text,
const RectInfos& rectInfos)
{
Difference difference;
difference.type = Type::TextRemoved;
difference.pageIndex1 = pageIndex;
difference.textRemovedIndex = m_strings.size();
m_strings << text;
difference.leftRectIndex = m_rects.size();
difference.leftRectCount = rectInfos.size();
m_rects.insert(m_rects.end(), rectInfos.cbegin(), rectInfos.cend());
m_differences.emplace_back(std::move(difference));
}
void PDFDiffResult::addTextReplaced(PDFInteger pageIndex1,
PDFInteger pageIndex2,
QString textRemoved,
QString textAdded,
const RectInfos& rectInfos1,
const RectInfos& rectInfos2)
{
Difference difference;
difference.type = Type::TextReplaced;
difference.pageIndex1 = pageIndex1;
difference.pageIndex2 = pageIndex2;
difference.textRemovedIndex = m_strings.size();
m_strings << textRemoved;
difference.textAddedIndex = m_strings.size();
m_strings << textAdded;
difference.leftRectIndex = m_rects.size();
difference.leftRectCount = rectInfos1.size();
m_rects.insert(m_rects.end(), rectInfos1.cbegin(), rectInfos1.cend());
difference.rightRectIndex = m_rects.size();
difference.rightRectCount = rectInfos2.size();
m_rects.insert(m_rects.end(), rectInfos2.cbegin(), rectInfos2.cend());
m_differences.emplace_back(std::move(difference));
}
QString PDFDiffResult::getMessage(size_t index) const
{
if (index >= m_differences.size())
@ -972,6 +1096,15 @@ QString PDFDiffResult::getMessage(size_t index) const
case Type::AddedShadingContent:
return PDFDiff::tr("Added shading from page %1.").arg(difference.pageIndex2 + 1);
case Type::TextAdded:
return PDFDiff::tr("Text '%1' has been added to page %2.").arg(m_strings[difference.textAddedIndex]).arg(difference.pageIndex2 + 1);
case Type::TextRemoved:
return PDFDiff::tr("Text '%1' has been removed from page %2.").arg(m_strings[difference.textRemovedIndex]).arg(difference.pageIndex1 + 1);
case Type::TextReplaced:
return PDFDiff::tr("Text '%1' on page %2 has been replaced by text '%3' on page %4.").arg(m_strings[difference.textRemovedIndex]).arg(difference.pageIndex1 + 1).arg(m_strings[difference.textAddedIndex]).arg(difference.pageIndex2 + 1);
default:
Q_ASSERT(false);
break;
@ -984,14 +1117,14 @@ void PDFDiffResult::addRectLeft(Difference& difference, QRectF rect)
{
difference.leftRectIndex = m_rects.size();
difference.leftRectCount = 1;
m_rects.emplace_back(rect);
m_rects.emplace_back(difference.pageIndex1, rect);
}
void PDFDiffResult::addRectRight(Difference& difference, QRectF rect)
{
difference.rightRectIndex = m_rects.size();
difference.rightRectCount = 1;
m_rects.emplace_back(rect);
m_rects.emplace_back(difference.pageIndex2, rect);
}
PDFDiffHelper::Differences PDFDiffHelper::calculateDifferences(const GraphicPieceInfos& left,
@ -1195,4 +1328,52 @@ std::vector<PDFDiffHelper::TextCompareItem> PDFDiffHelper::prepareTextCompareIte
return items;
}
void PDFDiffHelper::refineTextRectangles(PDFDiffResult::RectInfos& items)
{
PDFDiffResult::RectInfos refinedItems;
auto it = items.cbegin();
auto itEnd = items.cend();
while (it != itEnd)
{
// Jakub Melka: find range which can be merged into one
// rectangle (it must be on a single page and rectangles must go
// in right direction).
auto itNext = std::next(it);
while (itNext != itEnd)
{
const std::pair<PDFInteger, QRectF>& currentItem = *std::prev(itNext);
const std::pair<PDFInteger, QRectF>& nextItem = *itNext;
if (nextItem.first != currentItem.first)
{
// Page index has changed...
break;
}
const QRectF& left = currentItem.second;
const QRectF& right = nextItem.second;
if (left.center().x() >= right.center().x())
{
break;
}
++itNext;
}
// Merge range [it, itNext) into one new sequence
QRectF unifiedRect;
for (auto cit = it; cit != itNext; ++cit)
{
unifiedRect = unifiedRect.united((*cit).second);
}
refinedItems.emplace_back((*it).first, unifiedRect);
it = itNext;
}
items = std::move(refinedItems);
}
} // namespace pdf

64
Pdf4QtLib/sources/pdfdiff.h
View File

@ -54,24 +54,32 @@ public:
AddedVectorGraphicContent,
AddedImageContent,
AddedShadingContent,
TextReplaced,
TextAdded,
TextRemoved,
};
struct Difference
{
Type type = Type::Invalid;
PDFInteger pageIndex1 = -1;
PDFInteger pageIndex2 = -1;
size_t leftRectIndex = 0;
size_t leftRectCount = 0;
size_t rightRectIndex = 0;
size_t rightRectCount = 0;
};
using Differences = std::vector<Difference>;
using RectInfos = std::vector<std::pair<PDFInteger, QRectF>>;
void setResult(PDFOperationResult result) { m_result = std::move(result); }
const PDFOperationResult& getResult() const { return m_result; }
/// Returns true, if some difference was found
bool isChanged() const { return getDifferencesCount() > 0; }
/// Returns true, if no difference was found
bool isSame() const { return !isChanged(); }
/// Returns number of detected changes
size_t getDifferencesCount() const { return m_differences.size(); }
/// Returns message describing difference in a page content
/// \param index Index
QString getMessage(size_t index) const;
private:
friend class PDFDiff;
void addPageMoved(PDFInteger pageIndex1, PDFInteger pageIndex2);
void addPageAdded(PDFInteger pageIndex);
void addPageRemoved(PDFInteger pageIndex);
@ -85,9 +93,34 @@ public:
void addAddedImageContent(PDFInteger pageIndex, QRectF rect);
void addAddedShadingContent(PDFInteger pageIndex, QRectF rect);
QString getMessage(size_t index) const;
void addTextAdded(PDFInteger pageIndex, QString text, const RectInfos& rectInfos);
void addTextRemoved(PDFInteger pageIndex, QString text, const RectInfos& rectInfos);
void addTextReplaced(PDFInteger pageIndex1,
PDFInteger pageIndex2,
QString textRemoved,
QString textAdded,
const RectInfos& rectInfos1,
const RectInfos& rectInfos2);
/// Single content difference descriptor. It describes type
/// of difference (such as graphics, image, text change) on a page
/// or on a list of multiple pages.
struct Difference
{
Type type = Type::Invalid;
PDFInteger pageIndex1 = -1;
PDFInteger pageIndex2 = -1;
size_t leftRectIndex = 0;
size_t leftRectCount = 0;
size_t rightRectIndex = 0;
size_t rightRectCount = 0;
int textAddedIndex = -1;
int textRemovedIndex = -1;
};
using Differences = std::vector<Difference>;
private:
void addLeftItem(Type type, PDFInteger pageIndex, QRectF rect);
void addRightItem(Type type, PDFInteger pageIndex, QRectF rect);
@ -95,8 +128,9 @@ private:
void addRectRight(Difference& difference, QRectF rect);
Differences m_differences;
std::vector<QRectF> m_rects;
RectInfos m_rects; ///< Rectangles with page indices
PDFOperationResult m_result;
QStringList m_strings;
};
/// Diff engine for comparing two pdf documents.

54
Pdf4QtLib/sources/pdfdocumenttextflow.cpp
View File

@ -206,7 +206,6 @@ protected:
virtual void performOutputCharacter(const PDFTextCharacterInfo& info) override;
virtual void performMarkedContentBegin(const QByteArray& tag, const PDFObject& properties) override;
virtual void performMarkedContentEnd() override;
virtual void performPathPainting(const QPainterPath& path, bool stroke, bool fill, bool text, Qt::FillRule fillRule) override;
private:
const PDFStructureItem* getStructureTreeItemFromMCID(PDFInteger mcid) const;
@ -237,32 +236,22 @@ private:
std::vector<QRectF> m_characterBoundingRects;
};
void PDFStructureTreeTextContentProcessor::performPathPainting(const QPainterPath& path, bool stroke, bool fill, bool text, Qt::FillRule fillRule)
{
if (!text)
{
// Jakub Melka: This should not occur
return;
}
if (!m_extractorOptions.testFlag(PDFStructureTreeTextExtractor::BoundingBoxes))
{
return;
}
Q_UNUSED(stroke);
Q_UNUSED(fill);
Q_UNUSED(fillRule);
QMatrix matrix = getCurrentWorldMatrix();
QPainterPath worldPath = matrix.map(path);
m_currentBoundingBox = m_currentBoundingBox.united(worldPath.controlPointRect());
m_characterBoundingRects.push_back(worldPath.controlPointRect());
}
void PDFStructureTreeTextContentProcessor::finishText()
{
m_currentText = m_currentText.trimmed();
QString trimmedText = m_currentText.trimmed();
const int index = m_currentText.indexOf(trimmedText);
Q_ASSERT(index != -1);
if (trimmedText.size() < m_currentText.size())
{
// Fix character bounding boxes...
if (m_characterBoundingRects.size() == m_currentText.size())
{
std::vector<QRectF> boundingRects(std::next(m_characterBoundingRects.cbegin(), index), std::next(m_characterBoundingRects.cbegin(), index + trimmedText.length()));
m_characterBoundingRects = std::move(boundingRects);
}
m_currentText = std::move(trimmedText);
}
if (!m_currentText.isEmpty() && (!m_extractorOptions.testFlag(PDFStructureTreeTextExtractor::SkipArtifact) || !isArtifact()))
{
if (m_extractorOptions.testFlag(PDFStructureTreeTextExtractor::AdjustReversedText) && isReversedText())
@ -276,6 +265,7 @@ void PDFStructureTreeTextContentProcessor::finishText()
m_currentText = qMove(reversed);
std::reverse(m_characterBoundingRects.begin(), m_characterBoundingRects.end());
}
Q_ASSERT(m_currentText.size() == m_characterBoundingRects.size() || m_characterBoundingRects.empty());
m_textSequence.emplace_back(PDFStructureTreeTextItem::createText(std::move(m_currentText), m_pageIndex, m_currentBoundingBox, std::move(m_characterBoundingRects)));
}
m_currentText = QString();
@ -381,8 +371,6 @@ bool PDFStructureTreeTextContentProcessor::isContentKindSuppressed(ContentKind k
switch (kind)
{
case ContentKind::Text:
return !m_extractorOptions.testFlag(PDFStructureTreeTextExtractor::BoundingBoxes);
case ContentKind::Shapes:
case ContentKind::Images:
case ContentKind::Shading:
@ -408,6 +396,18 @@ void PDFStructureTreeTextContentProcessor::performOutputCharacter(const PDFTextC
if (!info.character.isNull() && info.character != QChar(QChar::SoftHyphen))
{
m_currentText.push_back(info.character);
QPainterPath worldPath = info.matrix.map(info.outline);
if (!worldPath.isEmpty())
{
QRectF boundingRect = worldPath.controlPointRect();
m_currentBoundingBox = m_currentBoundingBox.united(boundingRect);
m_characterBoundingRects.push_back(boundingRect);
}
else
{
m_characterBoundingRects.push_back(QRectF());
}
}
}
}