Reflow paragraphs when converting to html.

From: Erik Hovland <erik@hovland.org>

Based on the patch posted here:
http://lists.freedesktop.org/archives/poppler/2008-September/004126.html

It adds the capability to reflow paragraphs when converting to html.
---

 utils/HtmlFonts.cc     |    3 +
 utils/HtmlOutputDev.cc |  157 ++++++++++++++++++++++++++----------------------
 utils/pdftohtml.1      |    5 ++
 utils/pdftohtml.cc     |    6 +-
 4 files changed, 97 insertions(+), 74 deletions(-)


diff --git a/utils/HtmlFonts.cc b/utils/HtmlFonts.cc
index d2cbfd5..8570fe2 100644
--- a/utils/HtmlFonts.cc
+++ b/utils/HtmlFonts.cc
@@ -57,6 +57,7 @@ static Fonts fonts[font_num+1]={
 
 #define xoutRound(x) ((int)(x + 0.5))
 extern GBool xml;
+extern GBool reFlow;
 
 GooString* HtmlFont::DefaultFont=new GooString("Times"); // Arial,Helvetica,sans-serif
 
@@ -234,7 +235,7 @@ GooString* HtmlFont::HtmlFilter(Unicode* u, int uLen) {
 	case '&': tmp->append("&amp;");  break;
 	case '<': tmp->append("&lt;");  break;
 	case '>': tmp->append("&gt;");  break;
-	case ' ': tmp->append( !xml && ( i+1 >= uLen || !tmp->getLength() || tmp->getChar( tmp->getLength()-1 ) == ' ' ) ? "&nbsp;" : " " );
+	case ' ': tmp->append( !xml && !reFlow && ( i+1 >= uLen || !tmp->getLength() || tmp->getChar( tmp->getLength()-1 ) == ' ' ) ? "&nbsp;" : " " );
 	          break;
 	default:  
 	  {
diff --git a/utils/HtmlOutputDev.cc b/utils/HtmlOutputDev.cc
index 008090b..78439bf 100644
--- a/utils/HtmlOutputDev.cc
+++ b/utils/HtmlOutputDev.cc
@@ -72,6 +72,7 @@ extern GBool stout;
 extern GBool xml;
 extern GBool showHidden;
 extern GBool noMerge;
+extern GBool reFlow;
 
 static GooString* basename(GooString* str){
   
@@ -401,9 +402,9 @@ void HtmlPage::coalesce() {
   HtmlString *str1, *str2;
   HtmlFont *hfont1, *hfont2;
   double space, horSpace, vertSpace;
-  GBool addSpace, addLineBreak;
+  GBool addSpace, addLineBreak, nextLine;
   int n, i;
-  double curX, curY;
+  double curX, curY, lineStartX = 0.0;
 
 #if 0 //~ for debugging
   for (str1 = yxStrings; str1; str1 = str1->yxNext) {
@@ -471,53 +472,78 @@ void HtmlPage::coalesce() {
     str1->htext->insert(0, ls);
     delete ls;
   }
-  curX = str1->xMin; curY = str1->yMin;
+  curX = str1->xMin; curY = str1->yMin; lineStartX = curX;
 
   while (str1 && (str2 = str1->yxNext)) {
     hfont2 = getFont(str2);
     space = str1->yMax - str1->yMin;
     horSpace = str2->xMin - str1->xMax;
-    addLineBreak = !noMerge && (fabs(str1->xMin - str2->xMin) < 0.4);
+    addLineBreak = (nextLine = !noMerge && (fabs(str1->xMin - str2->xMin) < 0.4));
     vertSpace = str2->yMin - str1->yMax;
 
 //printf("coalesce %d %d %f? ", str1->dir, str2->dir, d);
 
-    if (str2->yMin >= str1->yMin && str2->yMin <= str1->yMax)
+    // Heuristic: If the last character in str1 is a hyphen then
+    // turn off addLineBreak. This will re-merge hyphenated words
+    // that have been split over multiple lines.
+    if (reFlow && str1->text[str1->len - 1] == '-')
     {
-	vertOverlap = str1->yMax - str2->yMin;
-    } else
-    if (str2->yMax >= str1->yMin && str2->yMax <= str1->yMax)
-    {
-	vertOverlap = str2->yMax - str1->yMin;
-    } else
-    {
-    	vertOverlap = 0;
-    } 
-    
-    if (
-	(
-	 (
-	  (
-	   (rawOrder && vertOverlap > 0.5 * space) 
-	   ||
-	   (!rawOrder && str2->yMin < str1->yMax)
-	  ) &&
-	  (horSpace > -0.5 * space && horSpace < space)
-	 ) ||
-       	 (vertSpace >= 0 && vertSpace < 0.5 * space && addLineBreak)
-	) &&
-	(!complexMode || (hfont1->isEqualIgnoreBold(*hfont2))) && // in complex mode fonts must be the same, in other modes fonts do not metter
-	str1->dir == str2->dir // text direction the same
-       ) 
+      addLineBreak = false;
+      str1->len--;
+      str1->htext->del(str1->htext->getLength() -1, 1);
+    }
+
+    // Is str2 a new paragraph?
+    if (nextLine
+	&&
+	((str2->xMin > lineStartX + 3.0)
+	 || (vertSpace > 0.5 * space)
+	 || xml
+	)
+       )
     {
-//      printf("yes\n");
+//      printf("new paragraph\n");
+      GBool finishA = str1->getLink() != NULL;
+      GBool finishBold = hfont1->isBold();
+      GBool finishItalic = hfont1->isItalic();
+      CloseTags( str1->htext, finishA, finishItalic, finishBold );
+
+      str1->xMin = curX; str1->yMin = curY;
+      str1 = str2;
+      curX = str1->xMin; curY = str1->yMin;
+      lineStartX= str1->xMin;
+      hfont1 = hfont2;
+      if( hfont1->isBold() )
+        str1->htext->insert(0,"<b>",3);
+      if( hfont1->isItalic() )
+        str1->htext->insert(0,"<i>",3);
+      if( str1->getLink() != NULL ) {
+        GooString *ls = str1->getLink()->getLinkStart();
+        str1->htext->insert(0, ls);
+        delete ls;
+      }
+    } else {
+//    printf("same paragraph\n");
+        
       n = str1->len + str2->len;
-      if ((addSpace = horSpace > 0.1 * space)) {
-        ++n;
+      if ((addSpace =
+           // if reflow is on then if horSpace is positive do regular check
+	   (reFlow && ((horSpace > 0.0 && horSpace > 0.1 * space)
+		       ||
+                       // but if horSpace is negative, check the Y directions
+		       (horSpace < 0.0 && str1->yMax < str2->yMax)
+		       )
+	   )
+	   ||
+           // otherwise reflow is off so do the usual check
+	   horSpace > 0.1 * space
+	  )
+	 )
+      {
+	++n;
       }
-      if (addLineBreak) {
+      if (nextLine)
         ++n;
-      }
   
       str1->size = (n + 15) & ~15;
       str1->text = (Unicode *)grealloc(str1->text,
@@ -525,19 +551,23 @@ void HtmlPage::coalesce() {
       str1->xRight = (double *)grealloc(str1->xRight,
 					str1->size * sizeof(double));
       if (addSpace) {
-		  str1->text[str1->len] = 0x20;
-		  str1->htext->append(xml?" ":"&nbsp;");
-		  str1->xRight[str1->len] = str2->xMin;
-		  ++str1->len;
+	str1->text[str1->len] = 0x20;
+	str1->htext->append((xml || reFlow) ? " " : "&nbsp;");
+	str1->xRight[str1->len] = str2->xMin;
+	++str1->len;
       }
-      if (addLineBreak) {
-	  str1->text[str1->len] = '\n';
-	  str1->htext->append("<br>");
-	  str1->xRight[str1->len] = str2->xMin;
-	  ++str1->len;
+
+      if (nextLine) {
+          if (addLineBreak) {
+	    str1->text[str1->len] = '\n';
+	    str1->htext->append(reFlow ? "\n" : "<br>");
+	    str1->xRight[str1->len] = str2->xMin;
+	    ++str1->len;
+          }
 	  str1->yMin = str2->yMin;
 	  str1->yMax = str2->yMax;
 	  str1->xMax = str2->xMax;
+	  lineStartX = str2->xMin;
 	  int fontLineSize = hfont1->getLineSize();
 	  int curLineSize = (int)(vertSpace + space); 
 	  if( curLineSize != fontLineSize )
@@ -589,26 +619,6 @@ void HtmlPage::coalesce() {
       }
       str1->yxNext = str2->yxNext;
       delete str2;
-    } else { // keep strings separate
-//      printf("no\n"); 
-      GBool finish_a = str1->getLink() != NULL;
-      GBool finish_bold   = hfont1->isBold();
-      GBool finish_italic = hfont1->isItalic();
-      CloseTags( str1->htext, finish_a, finish_italic, finish_bold );
-     
-      str1->xMin = curX; str1->yMin = curY; 
-      str1 = str2;
-      curX = str1->xMin; curY = str1->yMin;
-      hfont1 = hfont2;
-      if( hfont1->isBold() )
-	str1->htext->insert(0,"<b>",3);
-      if( hfont1->isItalic() )
-	str1->htext->insert(0,"<i>",3);
-      if( str1->getLink() != NULL ) {
-	GooString *ls = str1->getLink()->getLinkStart();
-	str1->htext->insert(0, ls);
-	delete ls;
-      }
     }
   }
   str1->xMin = curX; str1->yMin = curY;
@@ -777,13 +787,18 @@ void HtmlPage::dump(FILE *f, int pageNum)
     GooString* str;
     for(HtmlString *tmp=yxStrings;tmp;tmp=tmp->yxNext){
       if (tmp->htext){
-		str=new GooString(tmp->htext); 
-		fputs(str->getCString(),f);
-		delete str;      
-		fputs("<br>\n",f);  
+        str=new GooString(tmp->htext); 
+        if (reFlow)
+	  fputs("<p>\n", f);
+        fputs(str->getCString(),f);
+        delete str;      
+        if (reFlow)
+	  fputs("</p>\n", f);
+	else
+          fputs("<br>\n", f);  
       }
     }
-	fputs("<hr>\n",f);  
+    if (!reFlow) fputs("<hr>\n",f);  
   }
 }
 
@@ -1593,7 +1608,7 @@ GBool HtmlOutputDev::dumpDocOutline(Catalog* catalog)
 		if (noframes)
 		{
 			output = page; 
-			fputs("<hr>\n", output);
+			if (!reFlow) fputs("<hr>\n", output);
 		}
 		else
 		{
@@ -1610,7 +1625,7 @@ GBool HtmlOutputDev::dumpDocOutline(Catalog* catalog)
  
   	GBool done = newOutlineLevel(output, outlines, catalog);
   	if (done && !complexMode)
-    	fputs("<hr>\n", output);
+    	if (!reFlow) fputs("<hr>\n", output);
 	
 	if (bClose)
 	{
diff --git a/utils/pdftohtml.1 b/utils/pdftohtml.1
index 850aa84..0d30504 100644
--- a/utils/pdftohtml.1
+++ b/utils/pdftohtml.1
@@ -52,6 +52,11 @@ use standard output
 .B \-zoom <fp>
 zoom the pdf document (default 1.5)
 .TP
+.B \-reflow
+join paragraph lines together and separate paragraphs with a <p> tag. With
+this flag off, paragraph lines are separated by <br> tags and paragraphs are
+also separated by <br> tags.
+.TP
 .B \-xml
 output for XML post-processing
 .TP
diff --git a/utils/pdftohtml.cc b/utils/pdftohtml.cc
index 41312de..8be7974 100644
--- a/utils/pdftohtml.cc
+++ b/utils/pdftohtml.cc
@@ -69,6 +69,7 @@ static GBool errQuiet=gFalse;
 static GBool noDrm=gFalse;
 
 GBool showHidden = gFalse;
+GBool reFlow = gFalse;
 GBool noMerge = gFalse;
 static char ownerPassword[33] = "";
 static char userPassword[33] = "";
@@ -107,12 +108,14 @@ static const ArgDesc argDesc[] = {
    "zoom the pdf document (default 1.5)"},
   {"-xml",    argFlag,    &xml,         0,
    "output for XML post-processing"},
+  {"-reflow", argFlag,   &reFlow,   0,
+   "output reflow paragraphs"},
   {"-hidden", argFlag,   &showHidden,   0,
    "output hidden text"},
   {"-nomerge", argFlag, &noMerge, 0,
    "do not merge paragraphs"},   
   {"-enc",    argString,   textEncName,    sizeof(textEncName),
-   "output text encoding name"},
+   "output text encoding name (UTF-8, Latin1 etc"},
   {"-dev",    argString,   gsDevice,       sizeof(gsDevice),
    "output device name for Ghostscript (png16m, jpeg etc)"},
   {"-v",      argFlag,     &printVersion,  0,
@@ -250,7 +253,6 @@ int main(int argc, char *argv[]) {
    { 
        complexMode = gTrue;
        noframes = gTrue;
-       noMerge = gTrue;
    }
 
   // get page range