20 extract font metadata during text extraction

Makefile

@@ -16,7 +16,7 @@ FOFI_DIR = $(XDIR)fofi/
 GOO_DIR = $(XDIR)goo/
 XPDF_DIR = $(XDIR)xpdf/
 
-PY_SRC = src/xpydf/PdfLoader.cc src/xpydf/ImageInfoDev.cc src/xpydf/ImageDataDev.cc
+PY_SRC = src/xpydf/PdfLoader.cc src/xpydf/ImageInfoDev.cc src/xpydf/ImageDataDev.cc src/xpydf/FontOutputDev.cc
 SPLASH_SRC = $(wildcard $(SPLASH_DIR)*.cc)
 FOFI_SRC = $(wildcard $(FOFI_DIR)*.cc)
 

src/xpdf-4.04/xpdf/GfxFont.cc

@@ -1283,15 +1283,17 @@ Gfx8BitFont::Gfx8BitFont(XRef *xref, const char *tagA, Ref idA, GString *nameA,
       }
     }
   }
-
+  
   // construct the char code -> Unicode mapping object
   ctu = CharCodeToUnicode::make8BitToUnicode(toUnicode);
 
   // merge in a ToUnicode CMap, if there is one -- this overwrites
   // existing entries in ctu, i.e., the ToUnicode CMap takes
   // precedence, but the other encoding info is allowed to fill in any
   // holes
-  readToUnicodeCMap(fontDict, 8, ctu);
+  if (globalParams->getReadUnicodeCMap()) {
+    readToUnicodeCMap(fontDict, 8, ctu);
+  }
 
   // look for a Unicode-to-Unicode mapping
   if (name && (utu = globalParams->getUnicodeToUnicode(name))) {

src/xpdf-4.04/xpdf/GlobalParams.cc

@@ -749,6 +749,7 @@ GlobalParams::GlobalParams(const char *cfgFileName) {
   defaultPrinter = NULL;
   mapNumericCharNames = gTrue;
   mapUnknownCharNames = gFalse;
+  readUnicodeCMap = gTrue;
   mapExtTrueTypeFontsViaUnicode = gTrue;
   useTrueTypeUnicodeMapping = gFalse;
   droppedFonts = new GHash(gTrue);
@@ -3268,6 +3269,16 @@ GBool GlobalParams::getMapUnknownCharNames() {
   return map;
 }
 
+GBool GlobalParams::getReadUnicodeCMap() {
+  GBool read;
+
+  lockGlobalParams;
+  read = readUnicodeCMap;
+  unlockGlobalParams;
+
+  return read;
+}
+
 GBool GlobalParams::getMapExtTrueTypeFontsViaUnicode() {
   GBool map;
 
@@ -3793,6 +3804,12 @@ void GlobalParams::setMapUnknownCharNames(GBool map) {
   unlockGlobalParams;
 }
 
+void GlobalParams::setReadUnicodeCMap(GBool read) {
+  lockGlobalParams;
+  readUnicodeCMap = read;
+  unlockGlobalParams;
+}
+
 void GlobalParams::setMapExtTrueTypeFontsViaUnicode(GBool map) {
   lockGlobalParams;
   mapExtTrueTypeFontsViaUnicode = map;

src/xpdf-4.04/xpdf/GlobalParams.h

@@ -338,6 +338,7 @@ class GlobalParams {
   GString *getDefaultPrinter();
   GBool getMapNumericCharNames();
   GBool getMapUnknownCharNames();
+  GBool getReadUnicodeCMap();
   GBool getMapExtTrueTypeFontsViaUnicode();
   GBool getUseTrueTypeUnicodeMapping();
   GBool isDroppedFont(const char *fontName);
@@ -400,6 +401,7 @@ class GlobalParams {
   void setOverprintPreview(GBool preview);
   void setMapNumericCharNames(GBool map);
   void setMapUnknownCharNames(GBool map);
+  void setReadUnicodeCMap(GBool map);
   void setMapExtTrueTypeFontsViaUnicode(GBool map);
   void setTabStateFile(char *tabStateFileA);
   void setPrintCommands(GBool printCommandsA);
@@ -589,6 +591,7 @@ class GlobalParams {
 				//   from the viewer)
   GBool mapNumericCharNames;	// map numeric char names (from font subsets)?
   GBool mapUnknownCharNames;	// map unknown char names?
+  GBool readUnicodeCMap;	// Read the unicode c map?
   GBool mapExtTrueTypeFontsViaUnicode;  // map char codes to GID via Unicode
 				        //   for external TrueType fonts?
   GBool useTrueTypeUnicodeMapping;	// use the Unicode cmaps in TrueType

src/xpdf-4.04/xpdf/TextOutputDev.cc

@@ -203,38 +203,6 @@ static inline double dmax(double x, double y) {
 // TextChar
 //------------------------------------------------------------------------
 
-class TextChar {
-public:
-
-  TextChar(Unicode cA, int charPosA, int charLenA,
-	   double xMinA, double yMinA, double xMaxA, double yMaxA,
-	   int rotA, GBool rotatedA, GBool clippedA, GBool invisibleA,
-	   TextFontInfo *fontA, double fontSizeA,
-	   double colorRA, double colorGA, double colorBA);
-
-  static int cmpX(const void *p1, const void *p2);
-  static int cmpY(const void *p1, const void *p2);
-  static int cmpCharPos(const void *p1, const void *p2);
-
-  Unicode c;
-  int charPos;
-  int charLen;
-  double xMin, yMin, xMax, yMax;
-  TextFontInfo *font;
-  double fontSize;
-  double colorR,
-         colorG,
-         colorB;
-
-  // group the byte-size fields to minimize object size
-  Guchar rot;
-  char rotated;
-  char clipped;
-  char invisible;
-  char spaceAfter;
-  char overlap;
-};
-
 TextChar::TextChar(Unicode cA, int charPosA, int charLenA,
 		   double xMinA, double yMinA, double xMaxA, double yMaxA,
 		   int rotA, GBool rotatedA, GBool clippedA, GBool invisibleA,
@@ -1559,13 +1527,13 @@ void TextPage::addChar(GfxState *state, double x, double y,
       } else {
 	j = i;
       }
-      chars->append(new TextChar(uBuf[j], charPos, nBytes,
-				 xMin, yMin, xMax, yMax,
-				 curRot, rotated, clipped,
-				 state->getRender() == 3 || alpha < 0.001,
-				 curFont, curFontSize,
-				 colToDbl(rgb.r), colToDbl(rgb.g),
-				 colToDbl(rgb.b)));
+      chars->append(textCharType(uBuf[j], charPos, nBytes,
+        xMin, yMin, xMax, yMax,
+        curRot, rotated, clipped,
+        state->getRender() == 3 || alpha < 0.001,
+        curFont, curFontSize,
+        colToDbl(rgb.r), colToDbl(rgb.g),
+        colToDbl(rgb.b)));
     }
   }
 

src/xpdf-4.04/xpdf/TextOutputDev.h

@@ -137,6 +137,42 @@ class TextFontInfo {
   friend class TextWord;
 };
 
+//------------------------------------------------------------------------
+// TextChar
+//------------------------------------------------------------------------
+
+class TextChar {
+public:
+
+  TextChar(Unicode cA, int charPosA, int charLenA,
+    double xMinA, double yMinA, double xMaxA, double yMaxA,
+    int rotA, GBool rotatedA, GBool clippedA, GBool invisibleA,
+    TextFontInfo *fontA, double fontSizeA,
+    double colorRA, double colorGA, double colorBA);
+
+  static int cmpX(const void *p1, const void *p2);
+  static int cmpY(const void *p1, const void *p2);
+  static int cmpCharPos(const void *p1, const void *p2);
+
+  Unicode c;
+  int charPos;
+  int charLen;
+  double xMin, yMin, xMax, yMax;
+  TextFontInfo *font;
+  double fontSize;
+  double colorR,
+         colorG,
+         colorB;
+
+  // group the byte-size fields to minimize object size
+  Guchar rot;
+  char rotated;
+  char clipped;
+  char invisible;
+  char spaceAfter;
+  char overlap;
+};
+
 //------------------------------------------------------------------------
 // TextWord
 //------------------------------------------------------------------------
@@ -235,7 +271,7 @@ class TextLine {
   double getEdge(int idx) { return edge[idx]; }
   GBool getHyphenated() { return hyphenated; }
 
-private:
+public:
 
   static int cmpX(const void *p1, const void *p2);
 
@@ -386,7 +422,7 @@ class TextPage {
 public:
 
   TextPage(TextOutputControl *controlA);
-  ~TextPage();
+  virtual ~TextPage();
 
   // Write contents of page to a stream.
   void write(void *outputStream, TextOutputFunc outputFunc);
@@ -492,20 +528,25 @@ class TextPage {
   void removeChars(double xMin, double yMin, double xMax, double yMax,
 		   double xOverlapThresh, double yOverlapThresh);
 
-private:
+public:
+  virtual TextChar *textCharType(Unicode cA, int charPosA, int charLenA,
+    double xMinA, double yMinA, double xMaxA, double yMaxA,
+    int rotA, GBool rotatedA, GBool clippedA, GBool invisibleA,
+    TextFontInfo *fontA, double fontSizeA,
+    double colorRA, double colorGA, double colorBA
+  ) {
+    return new TextChar(cA, charPosA, charLenA, xMinA, yMinA, xMaxA, yMaxA,
+      rotA, rotatedA, clippedA, invisibleA, fontA, fontSizeA,
+      colorRA, colorGA, colorBA);
+  }
+  virtual void encodeFragment(Unicode *text, int len, UnicodeMap *uMap,
+		      GBool primaryLR, GString *s);
+
+  virtual void computeLinePhysWidth(TextLine *line, UnicodeMap *uMap);
 
-  void startPage(GfxState *state);
-  void clear();
-  void updateFont(GfxState *state);
   void addChar(GfxState *state, double x, double y,
 	       double dx, double dy,
 	       CharCode c, int nBytes, Unicode *u, int uLen);
-  void incCharCount(int nChars);
-  void beginActualText(GfxState *state, Unicode *u, int uLen);
-  void endActualText(GfxState *state);
-  void addUnderline(double x0, double y0, double x1, double y1);
-  void addLink(double xMin, double yMin, double xMax, double yMax,
-	       Link *link);
 
   // output
   void writeReadingOrder(void *outputStream,
@@ -538,8 +579,18 @@ class TextPage {
 		UnicodeMap *uMap,
 		char *space, int spaceLen,
 		char *eol, int eolLen);
-  void encodeFragment(Unicode *text, int len, UnicodeMap *uMap,
-		      GBool primaryLR, GString *s);
+
+private:
+
+  void startPage(GfxState *state);
+  void clear();
+  void updateFont(GfxState *state);
+  void incCharCount(int nChars);
+  void beginActualText(GfxState *state, Unicode *u, int uLen);
+  void endActualText(GfxState *state);
+  void addUnderline(double x0, double y0, double x1, double y1);
+  void addLink(double xMin, double yMin, double xMax, double yMax,
+	       Link *link);
   GBool unicodeEffectiveTypeLOrNum(Unicode u, Unicode left, Unicode right);
   GBool unicodeEffectiveTypeR(Unicode u, Unicode left, Unicode right);
 
@@ -593,7 +644,6 @@ class TextPage {
   int getCharDirection(TextChar *ch, TextChar *left, TextChar *right);
   int assignPhysLayoutPositions(GList *columns);
   void assignLinePhysPositions(GList *columns);
-  void computeLinePhysWidth(TextLine *line, UnicodeMap *uMap);
   int assignColumnPhysPositions(GList *columns);
   void buildSuperLines(TextBlock *blk, GList *superLines);
   void assignSimpleLayoutPositions(GList *superLines, UnicodeMap *uMap);
@@ -784,6 +834,10 @@ class TextOutputDev: public OutputDev {
   // Turn extra processing for HTML conversion on or off.
   void enableHTMLExtras(GBool html) { control.html = html; }
 
+protected:
+  TextPage *text;		// text for the current page
+  TextOutputControl control;	// formatting parameters
+
 private:
 
   void generateBOM();
@@ -792,8 +846,6 @@ class TextOutputDev: public OutputDev {
   void *outputStream;		// output stream
   GBool needClose;		// need to close the output file?
 				//   (only if outputStream is a FILE*)
-  TextPage *text;		// text for the current page
-  TextOutputControl control;	// formatting parameters
   GBool ok;			// set up ok?
 };