UTF-8 support (PDF 2.0 compliance) and ProcedureSets (for compatibility)

PdfForQtLib/sources/pdfencoding.cpp

@@ -2193,6 +2193,10 @@ QString PDFEncoding::convertTextString(const QByteArray& stream)
     {
         return convertFromUnicode(stream);
     }
+    else if (hasUTF8LeadMarkings(stream))
+    {
+        return QString::fromUtf8(stream);
+    }
     else
     {
         return convert(stream, Encoding::PDFDoc);
@@ -2348,4 +2352,20 @@ bool PDFEncoding::hasUnicodeLeadMarkings(const QByteArray& stream)
     return false;
 }
 
+bool PDFEncoding::hasUTF8LeadMarkings(const QByteArray& stream)
+{
+    if (stream.size() >= 3)
+    {
+        if (static_cast<unsigned char>(stream[0]) == 239 &&
+            static_cast<unsigned char>(stream[1]) == 187 &&
+            static_cast<unsigned char>(stream[2]) == 191)
+        {
+            // UTF-8
+            return true;
+        }
+    }
+
+    return false;
+}
+
 } // namespace pdf

PdfForQtLib/sources/pdfencoding.h

@@ -105,11 +105,18 @@ public:
     static const encoding::EncodingTable* getTableForEncoding(Encoding encoding);
 
 private:
-    /// Returns true, if byte array has UTF-16BE unicode marking bytes at the
+    /// Returns true, if byte array has UTF-16BE/LE unicode marking bytes at the
     /// stream start. If they are present, then byte stream is probably encoded
     /// as unicode.
     /// \param stream Stream to be tested
     static bool hasUnicodeLeadMarkings(const QByteArray& stream);
+
+    /// Returns true, if byte array has UTF-8 unicode marking bytes at the stream
+    /// start. If they are present, then byte stream is probably encoded
+    /// as UTF-8 string.
+    /// \note UTF-8 strings were added in PDF 2.0 specification
+    /// \param stream Stream to be tested
+    static bool hasUTF8LeadMarkings(const QByteArray& stream);
 };
 
 }   // namespace pdf

PdfForQtLib/sources/pdfpagecontentprocessor.cpp

@@ -156,10 +156,10 @@ static constexpr const std::pair<const char*, PDFPageContentProcessor::Operator>
 };
 
 void PDFPageContentProcessor::initDictionaries(const PDFObject& resourcesObject)
-{
-    auto getDictionary = [this, &resourcesObject](const char* resourceName) -> const pdf::PDFDictionary*
 {
     const PDFObject& resources = m_document->getObject(resourcesObject);
+    auto getDictionary = [this, &resources](const char* resourceName) -> const pdf::PDFDictionary*
+    {
         if (resources.isDictionary() && resources.getDictionary()->hasKey(resourceName))
         {
             const PDFObject& resourceDictionary = m_document->getObject(resources.getDictionary()->get(resourceName));
@@ -179,6 +179,43 @@ void PDFPageContentProcessor::initDictionaries(const PDFObject& resourcesObject)
     m_propertiesDictionary = getDictionary("Properties");
     m_shadingDictionary = getDictionary("Shading");
     m_patternDictionary = getDictionary("Pattern");
+    m_procedureSets = NoProcSet;
+
+    if (resources.isDictionary() && resources.getDictionary()->hasKey("ProcSet"))
+    {
+        PDFDocumentDataLoaderDecorator loader(m_document);
+        std::vector<QByteArray> procedureSetNames = loader.readNameArrayFromDictionary(resources.getDictionary(), "ProcSet");
+
+        ProcedureSets newProcSet = EmptyProcSet;
+        for (const QByteArray& procedureSetName : procedureSetNames)
+        {
+           if (procedureSetName == "PDF")
+           {
+               newProcSet.setFlag(PDF);
+           }
+           else if (procedureSetName == "Text")
+           {
+               newProcSet.setFlag(Text);
+           }
+           else if (procedureSetName == "ImageB")
+           {
+               newProcSet.setFlag(ImageB);
+           }
+           else if (procedureSetName == "ImageC")
+           {
+               newProcSet.setFlag(ImageC);
+           }
+           else if (procedureSetName == "ImageI")
+           {
+               newProcSet.setFlag(ImageI);
+           }
+        }
+
+        if (newProcSet)
+        {
+            m_procedureSets = newProcSet;
+        }
+    }
 }
 
 PDFPageContentProcessor::PDFPageContentProcessor(const PDFPage* page,
@@ -200,6 +237,7 @@ PDFPageContentProcessor::PDFPageContentProcessor(const PDFPage* page,
     m_propertiesDictionary(nullptr),
     m_shadingDictionary(nullptr),
     m_patternDictionary(nullptr),
+    m_procedureSets(NoProcSet),
     m_textBeginEndState(0),
     m_compatibilityBeginEndState(0),
     m_drawingUncoloredTilingPatternState(0),
@@ -3501,7 +3539,8 @@ PDFPageContentProcessor::PDFPageContentProcessorStateGuard::PDFPageContentProces
     m_extendedGraphicStateDictionary(processor->m_extendedGraphicStateDictionary),
     m_propertiesDictionary(processor->m_propertiesDictionary),
     m_shadingDictionary(processor->m_shadingDictionary),
-    m_patternDictionary(processor->m_patternDictionary)
+    m_patternDictionary(processor->m_patternDictionary),
+    m_procedureSets(processor->m_procedureSets)
 {
     m_processor->operatorSaveGraphicState();
 }
@@ -3516,6 +3555,7 @@ PDFPageContentProcessor::PDFPageContentProcessorStateGuard::~PDFPageContentProce
     m_processor->m_propertiesDictionary = m_propertiesDictionary;
     m_processor->m_shadingDictionary = m_shadingDictionary;
     m_processor->m_patternDictionary = m_patternDictionary;
+    m_processor->m_procedureSets = m_procedureSets;
 
     m_processor->operatorRestoreGraphicState();
 }

PdfForQtLib/sources/pdfpagecontentprocessor.h

@@ -197,6 +197,18 @@ public:
         Invalid                             ///< Invalid operator, use for error reporting
     };
 
+    enum ProcedureSet
+    {
+        EmptyProcSet    = 0x0000,
+        NoProcSet       = 0x0001,
+        PDF             = 0x0002,
+        Text            = 0x0004,
+        ImageB          = 0x0008,
+        ImageC          = 0x0010,
+        ImageI          = 0x0020
+    };
+    Q_DECLARE_FLAGS(ProcedureSets, ProcedureSet)
+
     /// Process the contents of the page
     QList<PDFRenderError> processContents();
 
@@ -547,6 +559,10 @@ protected:
     /// Returns page bounding rectangle in device space
     const QRectF& getPageBoundingRectDeviceSpace() const { return m_pageBoundingRectDeviceSpace; }
 
+    /// Returns current procedure sets. Procedure sets are deprecated in PDF 2.0 and are here
+    /// only for compatibility purposes. See chapter 14.2 in PDF 2.0 specification.
+    ProcedureSets getProcedureSets() const { return m_procedureSets; }
+
 private:
     /// Initializes the resources dictionaries
     void initDictionaries(const PDFObject& resourcesObject);
@@ -621,6 +637,7 @@ private:
         const PDFDictionary* m_propertiesDictionary;
         const PDFDictionary* m_shadingDictionary;
         const PDFDictionary* m_patternDictionary;
+        ProcedureSets m_procedureSets;
     };
 
     class PDFPageContentProcessorGraphicStateSaveRestoreGuard
@@ -878,6 +895,7 @@ private:
     const PDFDictionary* m_propertiesDictionary;
     const PDFDictionary* m_shadingDictionary;
     const PDFDictionary* m_patternDictionary;
+    ProcedureSets m_procedureSets;
 
     // Default color spaces
     PDFColorSpacePointer m_deviceGrayColorSpace;