From ceea8193d851c76fbb53ef997844e532550b2806 Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Sat, 7 Jun 2014 00:51:19 +0800
Subject: [PATCH 01/10] Fix XHR failure for file: protocol.

---
 share/pdf2htmlEX.js.in | 5 ++---
 1 file changed, 2 insertions(+), 3 deletions(-)

diff --git a/share/pdf2htmlEX.js.in b/share/pdf2htmlEX.js.in
index 25352fe..f21d3df 100644
--- a/share/pdf2htmlEX.js.in
+++ b/share/pdf2htmlEX.js.in
@@ -372,9 +372,8 @@ Viewer.prototype = {
         var _idx = idx;
         var xhr = new XMLHttpRequest();
         xhr.open('GET', url, true);
-        xhr.onreadystatechange = function(){
-          if (xhr.readyState != 4) return;
-          if (xhr.status === 200) {
+        xhr.onload = function(){
+          if (xhr.status === 200 || xhr.status === 0) {
             // find the page element in the data
             var div = document.createElement('div');
             div.innerHTML = xhr.responseText;

From 4865d3aa5c7ad2b67d9c2da0e55121316962320c Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Sat, 7 Jun 2014 16:20:20 +0800
Subject: [PATCH 02/10] Add Duan Yao to AUTHORS

---
 AUTHORS | 1 +
 1 file changed, 1 insertion(+)

diff --git a/AUTHORS b/AUTHORS
index 95f0be3..ed7cc51 100644
--- a/AUTHORS
+++ b/AUTHORS
@@ -6,6 +6,7 @@ Chris Cinelli <chris@allestelle.com>
 Daniel Bonniot de Ruisselet <dbonniot@chemaxon.com>
 Deepak <iapain@gmail.com>
 Denis Sablic <denis.sablic@gmail.com>
+Duan Yao <duanyao@ustc.edu>
 filodej <philode@gmail.com>
 hasufell <julian.ospald@googlemail.com>
 Herbert Jones <herbert@mediafire.com>

From 82b0569c06d1d0dd9fe8e6ab2801cff1dde4bd8c Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Sat, 7 Jun 2014 02:01:10 +0800
Subject: [PATCH 03/10] Add svg-nodes-limit option to avoid overly complicated
 svg background.

---
 .../CairoBackgroundRenderer.cc                | 57 +++++++++++++++++--
 .../CairoBackgroundRenderer.h                 | 11 ++--
 .../SplashBackgroundRenderer.cc               | 47 +++++++++------
 .../SplashBackgroundRenderer.h                |  7 +--
 src/Param.h                                   |  1 +
 src/pdf2htmlEX.cc                             |  3 +-
 6 files changed, 92 insertions(+), 34 deletions(-)

diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.cc b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
index 074e4c1..c92703b 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
@@ -15,12 +15,28 @@
 #if ENABLE_SVG
 
 #include "CairoBackgroundRenderer.h"
+#include "SplashBackgroundRenderer.h"
 
 namespace pdf2htmlEX {
 
 using std::string;
 using std::ifstream;
 
+CairoBackgroundRenderer::CairoBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
+      : CairoOutputDev()
+      , html_renderer(html_renderer)
+      , param(param)
+      , surface(nullptr)
+      , use_bitmap(false)
+{
+    if (param.svg_nodes_limit > 0)
+    {
+        this->bitmap_renderer = new SplashBackgroundRenderer(html_renderer, param);
+    }
+    else
+        this->bitmap_renderer = nullptr;
+}
+
 void CairoBackgroundRenderer::drawChar(GfxState *state, double x, double y,
         double dx, double dy,
         double originX, double originY,
@@ -46,6 +62,8 @@ void CairoBackgroundRenderer::drawChar(GfxState *state, double x, double y,
 void CairoBackgroundRenderer::init(PDFDoc * doc)
 {
     startDoc(doc);
+    if (this->bitmap_renderer != nullptr)
+          this->bitmap_renderer->init(doc);
 }
 
 static GBool annot_cb(Annot *, void * pflag) {
@@ -67,13 +85,11 @@ void CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
         page_height = doc->getPageMediaHeight(pageno);
     }
 
-    {
-        auto fn = html_renderer->str_fmt("%s/bg%x.svg", (param.embed_image ? param.tmp_dir : param.dest_dir).c_str(), pageno);
-        if(param.embed_image)
-            html_renderer->tmp_files.add((char*)fn);
+    auto fn = html_renderer->str_fmt("%s/bg%x.svg", (param.embed_image ? param.tmp_dir : param.dest_dir).c_str(), pageno);
+    if(param.embed_image)
+        html_renderer->tmp_files.add((char*)fn);
 
-        surface = cairo_svg_surface_create((char*)fn, page_width * param.h_dpi / DEFAULT_DPI, page_height * param.v_dpi / DEFAULT_DPI);
-    }
+    surface = cairo_svg_surface_create((char*)fn, page_width * param.h_dpi / DEFAULT_DPI, page_height * param.v_dpi / DEFAULT_DPI);
     cairo_svg_surface_restrict_to_version(surface, CAIRO_SVG_VERSION_1_2);
     cairo_surface_set_fallback_resolution(surface, param.h_dpi, param.v_dpi);
 
@@ -105,10 +121,39 @@ void CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
         if(status)
             throw string("Error in cairo: ") + cairo_status_to_string(status);
     }
+
+    //check node count in the svg file, fall back to bitmap_renderer if necessary.
+    if (param.svg_nodes_limit > 0)
+    {
+        int n = 0;
+        char c;
+        ifstream svgfile((char*)fn);
+        //count of '<' in the file should be an approximation of node count.
+        while(svgfile >> c)
+        {
+            if (c == '<')
+                ++n;
+        }
+        svgfile.close();
+        if (n > param.svg_nodes_limit)
+        {
+            html_renderer->tmp_files.add((char*)fn);
+            use_bitmap = true;
+            bitmap_renderer->render_page(doc, pageno);
+        }
+        else
+            use_bitmap = false;
+    }
 }
 
 void CairoBackgroundRenderer::embed_image(int pageno)
 {
+    if (use_bitmap)
+    {
+        bitmap_renderer->embed_image(pageno);
+        return;
+    }
+
     auto & f_page = *(html_renderer->f_curpage);
     
     f_page << "<img class=\"" << CSS::FULL_BACKGROUND_IMAGE_CN 
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.h b/src/BackgroundRenderer/CairoBackgroundRenderer.h
index 37372f1..76e7e6d 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.h
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.h
@@ -20,16 +20,13 @@
 
 namespace pdf2htmlEX {
 
+class SplashBackgroundRenderer;
+
 // Based on BackgroundRenderer from poppler
 class CairoBackgroundRenderer : public BackgroundRenderer, CairoOutputDev 
 {
 public:
-  CairoBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
-      : CairoOutputDev()
-      , html_renderer(html_renderer)
-      , param(param)
-      , surface(nullptr)
-  { }
+  CairoBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param);
 
   virtual ~CairoBackgroundRenderer() { }
 
@@ -50,6 +47,8 @@ protected:
   HTMLRenderer * html_renderer;
   const Param & param;
   cairo_surface_t * surface;
+  SplashBackgroundRenderer * bitmap_renderer;
+  bool use_bitmap;
 };
 
 }
diff --git a/src/BackgroundRenderer/SplashBackgroundRenderer.cc b/src/BackgroundRenderer/SplashBackgroundRenderer.cc
index 0e42b32..dadad78 100644
--- a/src/BackgroundRenderer/SplashBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/SplashBackgroundRenderer.cc
@@ -28,6 +28,30 @@ using std::unique_ptr;
 
 const SplashColor SplashBackgroundRenderer::white = {255,255,255};
 
+SplashBackgroundRenderer::SplashBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
+    : SplashOutputDev(splashModeRGB8, 4, gFalse, (SplashColorPtr)(&white), gTrue, gTrue)
+    , html_renderer(html_renderer)
+    , param(param)
+{
+    if(false) { }
+#ifdef ENABLE_LIBPNG
+    else if(param.bg_format == "png" || param.bg_format == "svg")
+    {
+        format = "png";
+    }
+#endif
+#ifdef ENABLE_LIBJPEG
+    else if(param.bg_format == "jpg" || param.bg_format == "svg")
+    {
+        format = "jpg";
+    }
+#endif
+    else
+    {
+        throw string("Image format not supported: ") + param.bg_format;
+    }
+}
+
 /*
  * SplashOutputDev::startPage would paint the whole page with the background color
  * And thus have modified region set to the whole page area
@@ -96,7 +120,7 @@ void SplashBackgroundRenderer::embed_image(int pageno)
     if((xmin <= xmax) && (ymin <= ymax))
     {
         {
-            auto fn = html_renderer->str_fmt("%s/bg%x.%s", (param.embed_image ? param.tmp_dir : param.dest_dir).c_str(), pageno, param.bg_format.c_str());
+            auto fn = html_renderer->str_fmt("%s/bg%x.%s", (param.embed_image ? param.tmp_dir : param.dest_dir).c_str(), pageno, format.c_str());
             if(param.embed_image)
                 html_renderer->tmp_files.add((char*)fn);
 
@@ -118,21 +142,21 @@ void SplashBackgroundRenderer::embed_image(int pageno)
 
         if(param.embed_image)
         {
-            auto path = html_renderer->str_fmt("%s/bg%x.%s", param.tmp_dir.c_str(), pageno, param.bg_format.c_str());
+            auto path = html_renderer->str_fmt("%s/bg%x.%s", param.tmp_dir.c_str(), pageno, format.c_str());
             ifstream fin((char*)path, ifstream::binary);
             if(!fin)
                 throw string("Cannot read background image ") + (char*)path;
 
-            auto iter = FORMAT_MIME_TYPE_MAP.find(param.bg_format);
+            auto iter = FORMAT_MIME_TYPE_MAP.find(format);
             if(iter == FORMAT_MIME_TYPE_MAP.end())
-                throw string("Image format not supported: ") + param.bg_format;
+                throw string("Image format not supported: ") + format;
 
             string mime_type = iter->second;
             f_page << "data:" << mime_type << ";base64," << Base64Stream(fin);
         }
         else
         {
-            f_page << (char*)html_renderer->str_fmt("bg%x.%s", pageno, param.bg_format.c_str());
+            f_page << (char*)html_renderer->str_fmt("bg%x.%s", pageno, format.c_str());
         }
         f_page << "\"/>";
     }
@@ -153,23 +177,14 @@ void SplashBackgroundRenderer::dump_image(const char * filename, int x1, int y1,
     // use unique_ptr to auto delete the object upon exception
     unique_ptr<ImgWriter> writer;
 
-    if(false) { }
-#ifdef ENABLE_LIBPNG
-    else if(param.bg_format == "png")
+    if(format == "png")
     {
         writer = unique_ptr<ImgWriter>(new PNGWriter);
     }
-#endif
-#ifdef ENABLE_LIBJPEG
-    else if(param.bg_format == "jpg")
+    else if(format == "jpg")
     {
         writer = unique_ptr<ImgWriter>(new JpegWriter);
     }
-#endif
-    else
-    {
-        throw string("Image format not supported: ") + param.bg_format;
-    }
 
     if(!writer->init(f, width, height, param.h_dpi, param.v_dpi))
         throw "Cannot initialize PNGWriter";
diff --git a/src/BackgroundRenderer/SplashBackgroundRenderer.h b/src/BackgroundRenderer/SplashBackgroundRenderer.h
index e999a10..55b9a97 100644
--- a/src/BackgroundRenderer/SplashBackgroundRenderer.h
+++ b/src/BackgroundRenderer/SplashBackgroundRenderer.h
@@ -27,11 +27,7 @@ class SplashBackgroundRenderer : public BackgroundRenderer, SplashOutputDev
 public:
   static const SplashColor white;
 
-  SplashBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
-      : SplashOutputDev(splashModeRGB8, 4, gFalse, (SplashColorPtr)(&white), gTrue, gTrue)
-      , html_renderer(html_renderer)
-      , param(param)
-  { }
+  SplashBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param);
 
   virtual ~SplashBackgroundRenderer() { }
 
@@ -68,6 +64,7 @@ protected:
   void dump_image(const char * filename, int x1, int y1, int x2, int y2);
   HTMLRenderer * html_renderer;
   const Param & param;
+  std::string format;
 };
 
 } // namespace pdf2htmlEX
diff --git a/src/Param.h b/src/Param.h
index 9d42620..806f837 100644
--- a/src/Param.h
+++ b/src/Param.h
@@ -63,6 +63,7 @@ struct Param
 
     // background image
     std::string bg_format;
+    int svg_nodes_limit = 0;
 
     // encryption
     std::string owner_password, user_password;
diff --git a/src/pdf2htmlEX.cc b/src/pdf2htmlEX.cc
index 23e8d73..0df3d30 100644
--- a/src/pdf2htmlEX.cc
+++ b/src/pdf2htmlEX.cc
@@ -190,7 +190,8 @@ void parse_options (int argc, char **argv)
 
         // background image
         .add("bg-format", &param.bg_format, "png", "specify background image format")
-
+        .add("svg-nodes-limit", &param.svg_nodes_limit, 0, "if node count in a svg background image exceeds this limit,"
+                " fall back to bitmap background. 0 or negative means no limit.")
         // encryption
         .add("owner-password,o", &param.owner_password, "", "owner password (for encrypted files)", true)
         .add("user-password,u", &param.user_password, "", "user password (for encrypted files)", true)

From 1764c1c33439ed02b8740705b81ab074b55bcd51 Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Sat, 7 Jun 2014 09:02:51 +0800
Subject: [PATCH 04/10] fix travis build

---
 src/Param.h | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/src/Param.h b/src/Param.h
index 806f837..8d19361 100644
--- a/src/Param.h
+++ b/src/Param.h
@@ -63,7 +63,7 @@ struct Param
 
     // background image
     std::string bg_format;
-    int svg_nodes_limit = 0;
+    int svg_nodes_limit;
 
     // encryption
     std::string owner_password, user_password;

From 4add9da6e4f06c793c4177b2f8a48bd28b53e1ec Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Sun, 8 Jun 2014 12:43:20 +0800
Subject: [PATCH 05/10] Move BackgroundRenderer fallback logic to HTMLRenderer.

---
 src/BackgroundRenderer/BackgroundRenderer.cc  | 22 ++++----
 src/BackgroundRenderer/BackgroundRenderer.h   |  6 ++-
 .../CairoBackgroundRenderer.cc                | 52 ++++++-------------
 .../CairoBackgroundRenderer.h                 |  6 +--
 .../SplashBackgroundRenderer.cc               | 22 ++++----
 .../SplashBackgroundRenderer.h                |  6 +--
 src/HTMLRenderer/HTMLRenderer.h               |  3 +-
 src/HTMLRenderer/general.cc                   | 19 ++++++-
 8 files changed, 66 insertions(+), 70 deletions(-)

diff --git a/src/BackgroundRenderer/BackgroundRenderer.cc b/src/BackgroundRenderer/BackgroundRenderer.cc
index e6cf59c..164e766 100644
--- a/src/BackgroundRenderer/BackgroundRenderer.cc
+++ b/src/BackgroundRenderer/BackgroundRenderer.cc
@@ -20,18 +20,9 @@ namespace pdf2htmlEX {
 
 BackgroundRenderer * BackgroundRenderer::getBackgroundRenderer(const std::string & format, HTMLRenderer * html_renderer, const Param & param)
 {
-#ifdef ENABLE_LIBPNG
-    if(format == "png")
-    {
-        return new SplashBackgroundRenderer(html_renderer, param);
-    }
-#endif
-#ifdef ENABLE_LIBJPEG
-    if(format == "jpg")
-    {
-        return new SplashBackgroundRenderer(html_renderer, param);
-    }
-#endif
+    if (format == "png" || format == "jpg")
+        return new SplashBackgroundRenderer(format, html_renderer, param);
+
 #if ENABLE_SVG
     if (format == "svg")
     {
@@ -42,4 +33,11 @@ BackgroundRenderer * BackgroundRenderer::getBackgroundRenderer(const std::string
     return nullptr;
 }
 
+BackgroundRenderer * BackgroundRenderer::getFallbackBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
+{
+    if (param.bg_format == "svg" && param.svg_nodes_limit > 0)
+        return new SplashBackgroundRenderer("", html_renderer, param);
+    return nullptr;
+}
+
 } // namespace pdf2htmlEX
diff --git a/src/BackgroundRenderer/BackgroundRenderer.h b/src/BackgroundRenderer/BackgroundRenderer.h
index f6d898e..29e03b6 100644
--- a/src/BackgroundRenderer/BackgroundRenderer.h
+++ b/src/BackgroundRenderer/BackgroundRenderer.h
@@ -22,12 +22,16 @@ class BackgroundRenderer
 public:
     // return nullptr upon failure
     static BackgroundRenderer * getBackgroundRenderer(const std::string & format, HTMLRenderer * html_renderer, const Param & param);
+    // Return a fallback bg renderer according to param.bg_format.
+    // Currently only svg bg format might need a bitmap fallback.
+    static BackgroundRenderer * getFallbackBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param);
 
     BackgroundRenderer() {}
     virtual ~BackgroundRenderer() {}
 
     virtual void init(PDFDoc * doc) = 0;
-    virtual void render_page(PDFDoc * doc, int pageno) = 0;
+    //return true on success, false otherwise (e.g. need a fallback)
+    virtual bool render_page(PDFDoc * doc, int pageno) = 0;
     virtual void embed_image(int pageno) = 0;
 
 };
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.cc b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
index c92703b..4f30305 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
@@ -23,19 +23,11 @@ using std::string;
 using std::ifstream;
 
 CairoBackgroundRenderer::CairoBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
-      : CairoOutputDev()
-      , html_renderer(html_renderer)
-      , param(param)
-      , surface(nullptr)
-      , use_bitmap(false)
-{
-    if (param.svg_nodes_limit > 0)
-    {
-        this->bitmap_renderer = new SplashBackgroundRenderer(html_renderer, param);
-    }
-    else
-        this->bitmap_renderer = nullptr;
-}
+    : CairoOutputDev()
+    , html_renderer(html_renderer)
+    , param(param)
+    , surface(nullptr)
+{ }
 
 void CairoBackgroundRenderer::drawChar(GfxState *state, double x, double y,
         double dx, double dy,
@@ -62,15 +54,13 @@ void CairoBackgroundRenderer::drawChar(GfxState *state, double x, double y,
 void CairoBackgroundRenderer::init(PDFDoc * doc)
 {
     startDoc(doc);
-    if (this->bitmap_renderer != nullptr)
-          this->bitmap_renderer->init(doc);
 }
 
 static GBool annot_cb(Annot *, void * pflag) {
     return (*((bool*)pflag)) ? gTrue : gFalse;
 };
 
-void CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
+bool CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
 {
     double page_width;
     double page_height;
@@ -85,11 +75,11 @@ void CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
         page_height = doc->getPageMediaHeight(pageno);
     }
 
-    auto fn = html_renderer->str_fmt("%s/bg%x.svg", (param.embed_image ? param.tmp_dir : param.dest_dir).c_str(), pageno);
+    string fn = (char*)html_renderer->str_fmt("%s/bg%x.svg", (param.embed_image ? param.tmp_dir : param.dest_dir).c_str(), pageno);
     if(param.embed_image)
-        html_renderer->tmp_files.add((char*)fn);
+        html_renderer->tmp_files.add(fn);
 
-    surface = cairo_svg_surface_create((char*)fn, page_width * param.h_dpi / DEFAULT_DPI, page_height * param.v_dpi / DEFAULT_DPI);
+    surface = cairo_svg_surface_create(fn.c_str(), page_width * param.h_dpi / DEFAULT_DPI, page_height * param.v_dpi / DEFAULT_DPI);
     cairo_svg_surface_restrict_to_version(surface, CAIRO_SVG_VERSION_1_2);
     cairo_surface_set_fallback_resolution(surface, param.h_dpi, param.v_dpi);
 
@@ -127,33 +117,25 @@ void CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
     {
         int n = 0;
         char c;
-        ifstream svgfile((char*)fn);
+        ifstream svgfile(fn);
         //count of '<' in the file should be an approximation of node count.
         while(svgfile >> c)
         {
             if (c == '<')
                 ++n;
+            if (n > param.svg_nodes_limit)
+            {
+                html_renderer->tmp_files.add(fn);
+                return false;
+            }
         }
-        svgfile.close();
-        if (n > param.svg_nodes_limit)
-        {
-            html_renderer->tmp_files.add((char*)fn);
-            use_bitmap = true;
-            bitmap_renderer->render_page(doc, pageno);
-        }
-        else
-            use_bitmap = false;
     }
+
+    return true;
 }
 
 void CairoBackgroundRenderer::embed_image(int pageno)
 {
-    if (use_bitmap)
-    {
-        bitmap_renderer->embed_image(pageno);
-        return;
-    }
-
     auto & f_page = *(html_renderer->f_curpage);
     
     f_page << "<img class=\"" << CSS::FULL_BACKGROUND_IMAGE_CN 
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.h b/src/BackgroundRenderer/CairoBackgroundRenderer.h
index 76e7e6d..b24b05b 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.h
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.h
@@ -20,8 +20,6 @@
 
 namespace pdf2htmlEX {
 
-class SplashBackgroundRenderer;
-
 // Based on BackgroundRenderer from poppler
 class CairoBackgroundRenderer : public BackgroundRenderer, CairoOutputDev 
 {
@@ -31,7 +29,7 @@ public:
   virtual ~CairoBackgroundRenderer() { }
 
   virtual void init(PDFDoc * doc);
-  virtual void render_page(PDFDoc * doc, int pageno);
+  virtual bool render_page(PDFDoc * doc, int pageno);
   virtual void embed_image(int pageno);
 
   // Does this device use beginType3Char/endType3Char?  Otherwise,
@@ -47,8 +45,6 @@ protected:
   HTMLRenderer * html_renderer;
   const Param & param;
   cairo_surface_t * surface;
-  SplashBackgroundRenderer * bitmap_renderer;
-  bool use_bitmap;
 };
 
 }
diff --git a/src/BackgroundRenderer/SplashBackgroundRenderer.cc b/src/BackgroundRenderer/SplashBackgroundRenderer.cc
index dadad78..970d59e 100644
--- a/src/BackgroundRenderer/SplashBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/SplashBackgroundRenderer.cc
@@ -28,27 +28,26 @@ using std::unique_ptr;
 
 const SplashColor SplashBackgroundRenderer::white = {255,255,255};
 
-SplashBackgroundRenderer::SplashBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
+SplashBackgroundRenderer::SplashBackgroundRenderer(const string & imgFormat, HTMLRenderer * html_renderer, const Param & param)
     : SplashOutputDev(splashModeRGB8, 4, gFalse, (SplashColorPtr)(&white), gTrue, gTrue)
     , html_renderer(html_renderer)
     , param(param)
+    , format(imgFormat)
 {
-    if(false) { }
+    bool supported = false;
 #ifdef ENABLE_LIBPNG
-    else if(param.bg_format == "png" || param.bg_format == "svg")
-    {
+    if (format.empty())
         format = "png";
-    }
+    supported = supported || format == "png";
 #endif
 #ifdef ENABLE_LIBJPEG
-    else if(param.bg_format == "jpg" || param.bg_format == "svg")
-    {
+    if (format.empty())
         format = "jpg";
-    }
+    supported = supported || format == "jpg";
 #endif
-    else
+    if (!supported)
     {
-        throw string("Image format not supported: ") + param.bg_format;
+        throw string("Image format not supported: ") + format;
     }
 }
 
@@ -100,7 +99,7 @@ static GBool annot_cb(Annot *, void * pflag) {
     return (*((bool*)pflag)) ? gTrue : gFalse;
 };
 
-void SplashBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
+bool SplashBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
 {
     bool process_annotation = param.process_annotation;
     doc->displayPage(this, pageno, param.h_dpi, param.v_dpi,
@@ -108,6 +107,7 @@ void SplashBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
             (!(param.use_cropbox)),
             false, false,
             nullptr, nullptr, &annot_cb, &process_annotation);
+    return true;
 }
 
 void SplashBackgroundRenderer::embed_image(int pageno)
diff --git a/src/BackgroundRenderer/SplashBackgroundRenderer.h b/src/BackgroundRenderer/SplashBackgroundRenderer.h
index 55b9a97..9ec8de9 100644
--- a/src/BackgroundRenderer/SplashBackgroundRenderer.h
+++ b/src/BackgroundRenderer/SplashBackgroundRenderer.h
@@ -26,13 +26,13 @@ class SplashBackgroundRenderer : public BackgroundRenderer, SplashOutputDev
 {
 public:
   static const SplashColor white;
-
-  SplashBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param);
+  //format: "png" or "jpg", or "" for a default format
+  SplashBackgroundRenderer(const std::string & format, HTMLRenderer * html_renderer, const Param & param);
 
   virtual ~SplashBackgroundRenderer() { }
 
   virtual void init(PDFDoc * doc);
-  virtual void render_page(PDFDoc * doc, int pageno);
+  virtual bool render_page(PDFDoc * doc, int pageno);
   virtual void embed_image(int pageno);
 
   // Does this device use beginType3Char/endType3Char?  Otherwise,
diff --git a/src/HTMLRenderer/HTMLRenderer.h b/src/HTMLRenderer/HTMLRenderer.h
index 7d67f70..73929ab 100644
--- a/src/HTMLRenderer/HTMLRenderer.h
+++ b/src/HTMLRenderer/HTMLRenderer.h
@@ -327,7 +327,8 @@ protected:
     friend class CairoBackgroundRenderer; // ugly!
 #endif
     BackgroundRenderer * bg_renderer;
-
+    BackgroundRenderer * fallback_bg_renderer;
+    bool fallback_bg_required;
 
     struct {
         std::ofstream fs;
diff --git a/src/HTMLRenderer/general.cc b/src/HTMLRenderer/general.cc
index 9c85a97..803bc2d 100644
--- a/src/HTMLRenderer/general.cc
+++ b/src/HTMLRenderer/general.cc
@@ -98,12 +98,17 @@ void HTMLRenderer::process(PDFDoc *doc)
     // Process pages
 
     bg_renderer = nullptr;
+    fallback_bg_renderer = nullptr;
     if(param.process_nontext)
     {
         bg_renderer = BackgroundRenderer::getBackgroundRenderer(param.bg_format, this, param);
         if(!bg_renderer)
             throw "Cannot initialize background renderer, unsupported format";
         bg_renderer->init(doc);
+
+        fallback_bg_renderer = BackgroundRenderer::getFallbackBackgroundRenderer(this, param);
+        if (fallback_bg_renderer)
+            fallback_bg_renderer->init(doc);
     }
 
     int page_count = (param.last_page - param.first_page + 1);
@@ -130,7 +135,9 @@ void HTMLRenderer::process(PDFDoc *doc)
 
         if(param.process_nontext)
         {
-            bg_renderer->render_page(doc, i);
+            fallback_bg_required = !bg_renderer->render_page(doc, i);
+            if (fallback_bg_required && fallback_bg_renderer != nullptr)
+                fallback_bg_renderer->render_page(doc, i);
         }
 
         doc->displayPage(this, i,
@@ -163,6 +170,11 @@ void HTMLRenderer::process(PDFDoc *doc)
         delete bg_renderer;
         bg_renderer = nullptr;
     }
+    if(fallback_bg_renderer)
+    {
+        delete fallback_bg_renderer;
+        fallback_bg_renderer = nullptr;
+    }
 
     cerr << endl;
 }
@@ -219,7 +231,10 @@ void HTMLRenderer::startPage(int pageNum, GfxState *state, XRef * xref)
 
     if(param.process_nontext)
     {
-        bg_renderer->embed_image(pageNum);
+        if (!fallback_bg_required)
+            bg_renderer->embed_image(pageNum);
+        else if (fallback_bg_renderer != nullptr)
+            fallback_bg_renderer->embed_image(pageNum);
     }
 
     reset_state();

From 0586c88d26444fc88ece3295f9354778e3573e2a Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Sun, 8 Jun 2014 20:53:26 +0800
Subject: [PATCH 06/10] Implement --svg-embed-bitmap: saving bitmaps in svg as
 external files.

---
 3rdparty/poppler/git/CairoOutputDev.h         |  2 +-
 .../CairoBackgroundRenderer.cc                | 82 ++++++++++++++++++-
 .../CairoBackgroundRenderer.h                 | 17 +++-
 src/Param.h                                   |  1 +
 src/pdf2htmlEX.cc                             |  3 +
 5 files changed, 100 insertions(+), 5 deletions(-)

diff --git a/3rdparty/poppler/git/CairoOutputDev.h b/3rdparty/poppler/git/CairoOutputDev.h
index c9ae33d..727c687 100644
--- a/3rdparty/poppler/git/CairoOutputDev.h
+++ b/3rdparty/poppler/git/CairoOutputDev.h
@@ -272,7 +272,7 @@ protected:
   cairo_filter_t getFilterForSurface(cairo_surface_t *image,
 				     GBool interpolate);
   GBool getStreamData (Stream *str, char **buffer, int *length);
-  void setMimeData(Stream *str, Object *ref, cairo_surface_t *image);
+  virtual void setMimeData(Stream *str, Object *ref, cairo_surface_t *image);
   void fillToStrokePathClip(GfxState *state);
   void alignStrokeCoords(GfxSubpath *subpath, int i, double *x, double *y);
 
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.cc b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
index 4f30305..795336a 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
@@ -19,8 +19,7 @@
 
 namespace pdf2htmlEX {
 
-using std::string;
-using std::ifstream;
+using namespace std;
 
 CairoBackgroundRenderer::CairoBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
     : CairoOutputDev()
@@ -29,6 +28,15 @@ CairoBackgroundRenderer::CairoBackgroundRenderer(HTMLRenderer * html_renderer, c
     , surface(nullptr)
 { }
 
+CairoBackgroundRenderer::~CairoBackgroundRenderer()
+{
+    for(auto i = bitmaps_ref_count.begin(); i != bitmaps_ref_count.end(); ++i)
+    {
+        if (i->second == 0)
+            html_renderer->tmp_files.add(this->get_bitmap_path(i->first));
+    }
+}
+
 void CairoBackgroundRenderer::drawChar(GfxState *state, double x, double y,
         double dx, double dy,
         double originX, double originY,
@@ -86,6 +94,8 @@ bool CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
     cairo_t * cr = cairo_create(surface);
     setCairo(cr);
 
+    bitmaps_in_current_page.resize(0);
+
     bool process_annotation = param.process_annotation;
     doc->displayPage(this, pageno, param.h_dpi, param.v_dpi,
             0, 
@@ -131,6 +141,10 @@ bool CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
         }
     }
 
+    // the svg file is actually used, so add its bitmaps' ref count.
+    for (auto i = bitmaps_in_current_page.begin(); i != bitmaps_in_current_page.end(); i++)
+        ++bitmaps_ref_count[*i];
+
     return true;
 }
 
@@ -138,7 +152,19 @@ void CairoBackgroundRenderer::embed_image(int pageno)
 {
     auto & f_page = *(html_renderer->f_curpage);
     
-    f_page << "<img class=\"" << CSS::FULL_BACKGROUND_IMAGE_CN 
+    // SVGs introduced by <img> or background-image can't have external resources;
+    // SVGs introduced by <embed> and <object> can, but they are more expensive for browsers.
+    // So we use <img> if the SVG contains no external bitmaps, and use <embed> otherwise.
+    // See also:
+    //   https://developer.mozilla.org/en-US/docs/Web/SVG/SVG_as_an_Image
+    //   http://stackoverflow.com/questions/4476526/do-i-use-img-object-or-embed-for-svg-files
+
+    if (param.svg_embed_bitmap || bitmaps_in_current_page.empty())
+        f_page << "<img";
+    else
+        f_page << "<embed";
+
+    f_page << " class=\"" << CSS::FULL_BACKGROUND_IMAGE_CN
         << "\" alt=\"\" src=\"";
 
     if(param.embed_image)
@@ -156,6 +182,56 @@ void CairoBackgroundRenderer::embed_image(int pageno)
     f_page << "\"/>";
 }
 
+// use object number as bitmap file name, without pageno prefix,
+// because a bitmap may be shared by multiple pages.
+const char* CairoBackgroundRenderer::get_bitmap_path(int id)
+{
+    return html_renderer->str_fmt("%s/%d.jpg", param.dest_dir.c_str(), id);
+}
+// Override CairoOutputDev::setMimeData() and dump bitmaps in SVG to external files.
+void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surface_t *image)
+{
+    if (param.svg_embed_bitmap)
+    {
+        CairoOutputDev::setMimeData(str, ref, image);
+        return;
+    }
+
+    // TODO dump bitmaps in other formats.
+    if (str->getKind() != strDCT)
+        return;
+
+    // TODO inline image?
+    if (ref == nullptr || !ref->isRef())
+        return;
+
+    int imgId = ref->getRef().num;
+    auto uri = strdup((char*) html_renderer->str_fmt("%d.jpg", imgId));
+    auto st = cairo_surface_set_mime_data(image, CAIRO_MIME_TYPE_URI,
+        (unsigned char*) uri, strlen(uri), gfree, uri);
+    if (st)
+    {
+        gfree(uri);
+        return;
+    }
+    bitmaps_in_current_page.push_back(imgId);
+
+    if(bitmaps_ref_count.find(imgId) != bitmaps_ref_count.end())
+        return;
+
+    bitmaps_ref_count[imgId] = 0;
+
+    char *strBuffer;
+    int len;
+    if (getStreamData(str->getNextStream(), &strBuffer, &len))
+    {
+        string path = get_bitmap_path(imgId);
+        ofstream imgfile(path, ofstream::binary);
+        imgfile.write(strBuffer, len);
+        gfree(strBuffer);
+    }
+}
+
 } // namespace pdf2htmlEX
 
 #endif // ENABLE_SVG
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.h b/src/BackgroundRenderer/CairoBackgroundRenderer.h
index b24b05b..25d366a 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.h
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.h
@@ -12,6 +12,8 @@
 #include <CairoOutputDev.h>
 #include <cairo.h>
 #include <cairo-svg.h>
+#include <map>
+#include <vector>
 
 #include "pdf2htmlEX-config.h"
 
@@ -26,7 +28,7 @@ class CairoBackgroundRenderer : public BackgroundRenderer, CairoOutputDev
 public:
   CairoBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param);
 
-  virtual ~CairoBackgroundRenderer() { }
+  virtual ~CairoBackgroundRenderer();
 
   virtual void init(PDFDoc * doc);
   virtual bool render_page(PDFDoc * doc, int pageno);
@@ -41,10 +43,23 @@ public:
       double originX, double originY,
       CharCode code, int nBytes, Unicode *u, int uLen);
 
+protected:
+  virtual void setMimeData(Stream *str, Object *ref, cairo_surface_t *image);
+
 protected:
   HTMLRenderer * html_renderer;
   const Param & param;
   cairo_surface_t * surface;
+
+private:
+  // convert bitmap stream id to bitmap file name. No pageno prefix,
+  // because a bitmap may be shared by multiple pages.
+  const char* get_bitmap_path(int id);
+  // map<id_of_bitmap_stream, usage_count_in_all_svgs>
+  // note: if a svg bg fallbacks to bitmap bg, its bitmaps are not taken into account.
+  std::map<int, int> bitmaps_ref_count;
+  // id of bitmaps' stream used by current page
+  std::vector<int> bitmaps_in_current_page;
 };
 
 }
diff --git a/src/Param.h b/src/Param.h
index 8d19361..3f20eca 100644
--- a/src/Param.h
+++ b/src/Param.h
@@ -64,6 +64,7 @@ struct Param
     // background image
     std::string bg_format;
     int svg_nodes_limit;
+    int svg_embed_bitmap;
 
     // encryption
     std::string owner_password, user_password;
diff --git a/src/pdf2htmlEX.cc b/src/pdf2htmlEX.cc
index 0df3d30..ccb426d 100644
--- a/src/pdf2htmlEX.cc
+++ b/src/pdf2htmlEX.cc
@@ -192,6 +192,9 @@ void parse_options (int argc, char **argv)
         .add("bg-format", &param.bg_format, "png", "specify background image format")
         .add("svg-nodes-limit", &param.svg_nodes_limit, 0, "if node count in a svg background image exceeds this limit,"
                 " fall back to bitmap background. 0 or negative means no limit.")
+        .add("svg-embed-bitmap", &param.svg_embed_bitmap, 1, "embed bitmaps in svg files or save them as external files"
+                " (even if 0 is specified, some bitmaps may still be embedded)")
+
         // encryption
         .add("owner-password,o", &param.owner_password, "", "owner password (for encrypted files)", true)
         .add("user-password,u", &param.user_password, "", "user password (for encrypted files)", true)

From edb0ddacea9a05ea17508260edde0be5b5650bc7 Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Wed, 11 Jun 2014 21:59:50 +0800
Subject: [PATCH 07/10] Some coding style improvements

---
 src/BackgroundRenderer/BackgroundRenderer.cc  | 13 ++++++++++--
 .../CairoBackgroundRenderer.cc                | 20 ++++++++++---------
 .../CairoBackgroundRenderer.h                 |  4 ++--
 3 files changed, 24 insertions(+), 13 deletions(-)

diff --git a/src/BackgroundRenderer/BackgroundRenderer.cc b/src/BackgroundRenderer/BackgroundRenderer.cc
index 164e766..5c37fde 100644
--- a/src/BackgroundRenderer/BackgroundRenderer.cc
+++ b/src/BackgroundRenderer/BackgroundRenderer.cc
@@ -20,9 +20,18 @@ namespace pdf2htmlEX {
 
 BackgroundRenderer * BackgroundRenderer::getBackgroundRenderer(const std::string & format, HTMLRenderer * html_renderer, const Param & param)
 {
-    if (format == "png" || format == "jpg")
+#ifdef ENABLE_LIBPNG
+    if(format == "png")
+    {
         return new SplashBackgroundRenderer(format, html_renderer, param);
-
+    }
+#endif
+#ifdef ENABLE_LIBJPEG
+    if(format == "jpg")
+    {
+        return new SplashBackgroundRenderer(format, html_renderer, param);
+    }
+#endif
 #if ENABLE_SVG
     if (format == "svg")
     {
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.cc b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
index 795336a..c418933 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
@@ -19,7 +19,11 @@
 
 namespace pdf2htmlEX {
 
-using namespace std;
+using std::string;
+using std::ifstream;
+using std::ofstream;
+using std::vector;
+using std::unordered_map;
 
 CairoBackgroundRenderer::CairoBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
     : CairoOutputDev()
@@ -30,10 +34,10 @@ CairoBackgroundRenderer::CairoBackgroundRenderer(HTMLRenderer * html_renderer, c
 
 CairoBackgroundRenderer::~CairoBackgroundRenderer()
 {
-    for(auto i = bitmaps_ref_count.begin(); i != bitmaps_ref_count.end(); ++i)
+    for(auto itr = bitmaps_ref_count.begin(); itr != bitmaps_ref_count.end(); ++itr)
     {
-        if (i->second == 0)
-            html_renderer->tmp_files.add(this->get_bitmap_path(i->first));
+        if (itr->second == 0)
+            html_renderer->tmp_files.add(this->get_bitmap_path(itr->first));
     }
 }
 
@@ -182,8 +186,6 @@ void CairoBackgroundRenderer::embed_image(int pageno)
     f_page << "\"/>";
 }
 
-// use object number as bitmap file name, without pageno prefix,
-// because a bitmap may be shared by multiple pages.
 const char* CairoBackgroundRenderer::get_bitmap_path(int id)
 {
     return html_renderer->str_fmt("%s/%d.jpg", param.dest_dir.c_str(), id);
@@ -208,10 +210,10 @@ void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surfac
     int imgId = ref->getRef().num;
     auto uri = strdup((char*) html_renderer->str_fmt("%d.jpg", imgId));
     auto st = cairo_surface_set_mime_data(image, CAIRO_MIME_TYPE_URI,
-        (unsigned char*) uri, strlen(uri), gfree, uri);
+        (unsigned char*) uri, strlen(uri), free, uri);
     if (st)
     {
-        gfree(uri);
+        free(uri);
         return;
     }
     bitmaps_in_current_page.push_back(imgId);
@@ -228,7 +230,7 @@ void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surfac
         string path = get_bitmap_path(imgId);
         ofstream imgfile(path, ofstream::binary);
         imgfile.write(strBuffer, len);
-        gfree(strBuffer);
+        free(strBuffer);
     }
 }
 
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.h b/src/BackgroundRenderer/CairoBackgroundRenderer.h
index 25d366a..fc587a3 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.h
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.h
@@ -12,7 +12,7 @@
 #include <CairoOutputDev.h>
 #include <cairo.h>
 #include <cairo-svg.h>
-#include <map>
+#include <unordered_map>
 #include <vector>
 
 #include "pdf2htmlEX-config.h"
@@ -57,7 +57,7 @@ private:
   const char* get_bitmap_path(int id);
   // map<id_of_bitmap_stream, usage_count_in_all_svgs>
   // note: if a svg bg fallbacks to bitmap bg, its bitmaps are not taken into account.
-  std::map<int, int> bitmaps_ref_count;
+  std::unordered_map<int, int> bitmaps_ref_count;
   // id of bitmaps' stream used by current page
   std::vector<int> bitmaps_in_current_page;
 };

From 5681fe11c0367e959485ab2d06f864fd62efab0d Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Wed, 11 Jun 2014 22:19:17 +0800
Subject: [PATCH 08/10] Fix bad usage of html_renderer->str_fmt() in
 get_bitmap_path(); add prefix to bitmap file name.

---
 .../CairoBackgroundRenderer.cc                   | 16 ++++++++++------
 src/BackgroundRenderer/CairoBackgroundRenderer.h |  3 ++-
 2 files changed, 12 insertions(+), 7 deletions(-)

diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.cc b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
index c418933..56c8596 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
@@ -37,7 +37,10 @@ CairoBackgroundRenderer::~CairoBackgroundRenderer()
     for(auto itr = bitmaps_ref_count.begin(); itr != bitmaps_ref_count.end(); ++itr)
     {
         if (itr->second == 0)
-            html_renderer->tmp_files.add(this->get_bitmap_path(itr->first));
+        {
+            string path;
+            html_renderer->tmp_files.add(this->build_bitmap_path(itr->first, path));
+        }
     }
 }
 
@@ -186,9 +189,10 @@ void CairoBackgroundRenderer::embed_image(int pageno)
     f_page << "\"/>";
 }
 
-const char* CairoBackgroundRenderer::get_bitmap_path(int id)
+string & CairoBackgroundRenderer::build_bitmap_path(int id, string & path)
 {
-    return html_renderer->str_fmt("%s/%d.jpg", param.dest_dir.c_str(), id);
+    // "po" for "PDF Object"
+    return path = html_renderer->str_fmt("%s/po-%d.jpg", param.dest_dir.c_str(), id);
 }
 // Override CairoOutputDev::setMimeData() and dump bitmaps in SVG to external files.
 void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surface_t *image)
@@ -208,7 +212,7 @@ void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surfac
         return;
 
     int imgId = ref->getRef().num;
-    auto uri = strdup((char*) html_renderer->str_fmt("%d.jpg", imgId));
+    auto uri = strdup((char*) html_renderer->str_fmt("po-%d.jpg", imgId));
     auto st = cairo_surface_set_mime_data(image, CAIRO_MIME_TYPE_URI,
         (unsigned char*) uri, strlen(uri), free, uri);
     if (st)
@@ -227,8 +231,8 @@ void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surfac
     int len;
     if (getStreamData(str->getNextStream(), &strBuffer, &len))
     {
-        string path = get_bitmap_path(imgId);
-        ofstream imgfile(path, ofstream::binary);
+        string path;
+        ofstream imgfile(build_bitmap_path(imgId, path), ofstream::binary);
         imgfile.write(strBuffer, len);
         free(strBuffer);
     }
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.h b/src/BackgroundRenderer/CairoBackgroundRenderer.h
index fc587a3..65e4515 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.h
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.h
@@ -14,6 +14,7 @@
 #include <cairo-svg.h>
 #include <unordered_map>
 #include <vector>
+#include <string>
 
 #include "pdf2htmlEX-config.h"
 
@@ -54,7 +55,7 @@ protected:
 private:
   // convert bitmap stream id to bitmap file name. No pageno prefix,
   // because a bitmap may be shared by multiple pages.
-  const char* get_bitmap_path(int id);
+  std::string & build_bitmap_path(int id, std::string & path);
   // map<id_of_bitmap_stream, usage_count_in_all_svgs>
   // note: if a svg bg fallbacks to bitmap bg, its bitmaps are not taken into account.
   std::unordered_map<int, int> bitmaps_ref_count;

From fdbc70401f3d6ba4c21088256badae5ad4bc9b65 Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Wed, 11 Jun 2014 23:20:09 +0800
Subject: [PATCH 09/10] Improve help texts for --svg-node-count-limit and
 --svg-embed-bitmap; rename --svg-nodes-limit to --svg-node-count-limit.

---
 pdf2htmlEX.1.in                                   | 10 ++++++++++
 src/BackgroundRenderer/BackgroundRenderer.cc      |  2 +-
 src/BackgroundRenderer/CairoBackgroundRenderer.cc |  8 ++++----
 src/Param.h                                       |  2 +-
 src/pdf2htmlEX.cc                                 |  7 +++----
 5 files changed, 19 insertions(+), 10 deletions(-)

diff --git a/pdf2htmlEX.1.in b/pdf2htmlEX.1.in
index 8b01c34..b63b1d9 100644
--- a/pdf2htmlEX.1.in
+++ b/pdf2htmlEX.1.in
@@ -248,6 +248,16 @@ If set to 1, pdf2htmlEX will try to reduce the number of HTML elements used for
 .B --bg-format <format> (Default: png)
 Specify the background image format. Run `pdf2htmlEX -v` to check all supported formats.
 
+.TP
+.B --svg-node-count-limit <limit> (Default: -1)
+If node count in a svg background image exceeds this limit, fall back this page to bitmap background; negative value means no limit.
+This option is only useful when '--bg-format svg' is specified. Note that node count in svg is just calculated approximately.
+
+.TP
+.B --svg-embed-bitmap <0|1> (Default: 1)
+Whether embed bitmaps in svg background image. 1: embed bitmaps in svg background; 0: dump bitmaps to external files if possible.
+JPEG images in a PDF are most possibly dumped. This option is only useful when '--bg-format svg' is specified.
+
 .SS PDF Protection
 
 .TP
diff --git a/src/BackgroundRenderer/BackgroundRenderer.cc b/src/BackgroundRenderer/BackgroundRenderer.cc
index 5c37fde..1ae298c 100644
--- a/src/BackgroundRenderer/BackgroundRenderer.cc
+++ b/src/BackgroundRenderer/BackgroundRenderer.cc
@@ -44,7 +44,7 @@ BackgroundRenderer * BackgroundRenderer::getBackgroundRenderer(const std::string
 
 BackgroundRenderer * BackgroundRenderer::getFallbackBackgroundRenderer(HTMLRenderer * html_renderer, const Param & param)
 {
-    if (param.bg_format == "svg" && param.svg_nodes_limit > 0)
+    if (param.bg_format == "svg" && param.svg_node_count_limit >= 0)
         return new SplashBackgroundRenderer("", html_renderer, param);
     return nullptr;
 }
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.cc b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
index 56c8596..4279d3a 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
@@ -130,7 +130,7 @@ bool CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
     }
 
     //check node count in the svg file, fall back to bitmap_renderer if necessary.
-    if (param.svg_nodes_limit > 0)
+    if (param.svg_node_count_limit >= 0)
     {
         int n = 0;
         char c;
@@ -140,7 +140,7 @@ bool CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
         {
             if (c == '<')
                 ++n;
-            if (n > param.svg_nodes_limit)
+            if (n > param.svg_node_count_limit)
             {
                 html_renderer->tmp_files.add(fn);
                 return false;
@@ -149,8 +149,8 @@ bool CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
     }
 
     // the svg file is actually used, so add its bitmaps' ref count.
-    for (auto i = bitmaps_in_current_page.begin(); i != bitmaps_in_current_page.end(); i++)
-        ++bitmaps_ref_count[*i];
+    for (auto itr = bitmaps_in_current_page.begin(); itr != bitmaps_in_current_page.end(); itr++)
+        ++bitmaps_ref_count[*itr];
 
     return true;
 }
diff --git a/src/Param.h b/src/Param.h
index 3f20eca..8c16802 100644
--- a/src/Param.h
+++ b/src/Param.h
@@ -63,7 +63,7 @@ struct Param
 
     // background image
     std::string bg_format;
-    int svg_nodes_limit;
+    int svg_node_count_limit;
     int svg_embed_bitmap;
 
     // encryption
diff --git a/src/pdf2htmlEX.cc b/src/pdf2htmlEX.cc
index ccb426d..f20dc2b 100644
--- a/src/pdf2htmlEX.cc
+++ b/src/pdf2htmlEX.cc
@@ -190,10 +190,9 @@ void parse_options (int argc, char **argv)
 
         // background image
         .add("bg-format", &param.bg_format, "png", "specify background image format")
-        .add("svg-nodes-limit", &param.svg_nodes_limit, 0, "if node count in a svg background image exceeds this limit,"
-                " fall back to bitmap background. 0 or negative means no limit.")
-        .add("svg-embed-bitmap", &param.svg_embed_bitmap, 1, "embed bitmaps in svg files or save them as external files"
-                " (even if 0 is specified, some bitmaps may still be embedded)")
+        .add("svg-node-count-limit", &param.svg_node_count_limit, -1, "if node count in a svg background image exceeds this limit,"
+                " fall back this page to bitmap background; negative value means no limit.")
+        .add("svg-embed-bitmap", &param.svg_embed_bitmap, 1, "1: embed bitmaps in svg background; 0: dump bitmaps to external files if possible.")
 
         // encryption
         .add("owner-password,o", &param.owner_password, "", "owner password (for encrypted files)", true)

From ac8208a6938b83dc259b026f72404dfbddf7a269 Mon Sep 17 00:00:00 2001
From: Duan Yao <duanyao@ustc.edu>
Date: Tue, 17 Jun 2014 11:59:45 +0800
Subject: [PATCH 10/10] Code cleaning.

---
 .../CairoBackgroundRenderer.cc                   | 16 +++++++---------
 src/BackgroundRenderer/CairoBackgroundRenderer.h |  2 +-
 .../SplashBackgroundRenderer.cc                  | 11 ++++++++++-
 3 files changed, 18 insertions(+), 11 deletions(-)

diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.cc b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
index 4279d3a..63a6a81 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.cc
@@ -38,8 +38,7 @@ CairoBackgroundRenderer::~CairoBackgroundRenderer()
     {
         if (itr->second == 0)
         {
-            string path;
-            html_renderer->tmp_files.add(this->build_bitmap_path(itr->first, path));
+            html_renderer->tmp_files.add(this->build_bitmap_path(itr->first));
         }
     }
 }
@@ -101,7 +100,7 @@ bool CairoBackgroundRenderer::render_page(PDFDoc * doc, int pageno)
     cairo_t * cr = cairo_create(surface);
     setCairo(cr);
 
-    bitmaps_in_current_page.resize(0);
+    bitmaps_in_current_page.clear();
 
     bool process_annotation = param.process_annotation;
     doc->displayPage(this, pageno, param.h_dpi, param.v_dpi,
@@ -189,10 +188,10 @@ void CairoBackgroundRenderer::embed_image(int pageno)
     f_page << "\"/>";
 }
 
-string & CairoBackgroundRenderer::build_bitmap_path(int id, string & path)
+string CairoBackgroundRenderer::build_bitmap_path(int id)
 {
-    // "po" for "PDF Object"
-    return path = html_renderer->str_fmt("%s/po-%d.jpg", param.dest_dir.c_str(), id);
+    // "o" for "PDF Object"
+    return string(html_renderer->str_fmt("%s/o%d.jpg", param.dest_dir.c_str(), id));
 }
 // Override CairoOutputDev::setMimeData() and dump bitmaps in SVG to external files.
 void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surface_t *image)
@@ -212,7 +211,7 @@ void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surfac
         return;
 
     int imgId = ref->getRef().num;
-    auto uri = strdup((char*) html_renderer->str_fmt("po-%d.jpg", imgId));
+    auto uri = strdup((char*) html_renderer->str_fmt("o%d.jpg", imgId));
     auto st = cairo_surface_set_mime_data(image, CAIRO_MIME_TYPE_URI,
         (unsigned char*) uri, strlen(uri), free, uri);
     if (st)
@@ -231,8 +230,7 @@ void CairoBackgroundRenderer::setMimeData(Stream *str, Object *ref, cairo_surfac
     int len;
     if (getStreamData(str->getNextStream(), &strBuffer, &len))
     {
-        string path;
-        ofstream imgfile(build_bitmap_path(imgId, path), ofstream::binary);
+        ofstream imgfile(build_bitmap_path(imgId), ofstream::binary);
         imgfile.write(strBuffer, len);
         free(strBuffer);
     }
diff --git a/src/BackgroundRenderer/CairoBackgroundRenderer.h b/src/BackgroundRenderer/CairoBackgroundRenderer.h
index 65e4515..8abe5f3 100644
--- a/src/BackgroundRenderer/CairoBackgroundRenderer.h
+++ b/src/BackgroundRenderer/CairoBackgroundRenderer.h
@@ -55,7 +55,7 @@ protected:
 private:
   // convert bitmap stream id to bitmap file name. No pageno prefix,
   // because a bitmap may be shared by multiple pages.
-  std::string & build_bitmap_path(int id, std::string & path);
+  std::string build_bitmap_path(int id);
   // map<id_of_bitmap_stream, usage_count_in_all_svgs>
   // note: if a svg bg fallbacks to bitmap bg, its bitmaps are not taken into account.
   std::unordered_map<int, int> bitmaps_ref_count;
diff --git a/src/BackgroundRenderer/SplashBackgroundRenderer.cc b/src/BackgroundRenderer/SplashBackgroundRenderer.cc
index 970d59e..c596508 100644
--- a/src/BackgroundRenderer/SplashBackgroundRenderer.cc
+++ b/src/BackgroundRenderer/SplashBackgroundRenderer.cc
@@ -177,14 +177,23 @@ void SplashBackgroundRenderer::dump_image(const char * filename, int x1, int y1,
     // use unique_ptr to auto delete the object upon exception
     unique_ptr<ImgWriter> writer;
 
-    if(format == "png")
+    if(false) { }
+#ifdef ENABLE_LIBPNG
+    else if(format == "png")
     {
         writer = unique_ptr<ImgWriter>(new PNGWriter);
     }
+#endif
+#ifdef ENABLE_LIBJPEG
     else if(format == "jpg")
     {
         writer = unique_ptr<ImgWriter>(new JpegWriter);
     }
+#endif
+    else
+    {
+        throw string("Image format not supported: ") + format;
+    }
 
     if(!writer->init(f, width, height, param.h_dpi, param.v_dpi))
         throw "Cannot initialize PNGWriter";