better file structure

CMakeLists.txt

@@ -24,9 +24,12 @@ add_executable(pdf2htmlEX
     src/pdf2htmlEX.cc 
     src/HTMLRenderer.h 
     src/HTMLRenderer.cc 
+    src/HTMLRenderer/general.cc
     src/HTMLRenderer/state.cc
     src/HTMLRenderer/install.cc
     src/HTMLRenderer/export.cc
+    src/HTMLRenderer/text.cc
+    src/HTMLRenderer/image.cc
     src/BackgroundRenderer.h 
     src/BackgroundRenderer.cc 
     src/Consts.h 

README.md

@@ -30,7 +30,7 @@ Not supported yet
 Dependency
 ----------------------------
 * libpoppler with xpdf header >= 0.20.2
-* boost c++ library (format, program options, gil)
+* boost c++ library (format, program options, gil, filesystem)
 * fontforge **Please use [the lastest version](https://github.com/fontforge/fontforge)**
 
 HOW TO COMPILE

src/HTMLRenderer.cc

@@ -1,383 +0,0 @@
-/*
- * HTMLRenderer.cc
- *
- * Copyright (C) 2011 by Hongliang TIAN(tatetian@gmail.com)
- * Copyright (C) 2012 by Lu Wang coolwanglu<at>gmail.com
- */
-
-/*
- * TODO
- * font base64 embedding
- */
-
-#include <cmath>
-#include <cassert>
-#include <fstream>
-#include <algorithm>
-
-#include <boost/format.hpp>
-#include <boost/algorithm/string.hpp>
-#include <boost/filesystem/fstream.hpp>
-// for gil bug
-const int *int_p_NULL = nullptr;
-#include <boost/gil/gil_all.hpp>
-#include <boost/gil/extension/io/png_dynamic_io.hpp>
-
-#include <GfxFont.h>
-#include <CharCodeToUnicode.h>
-#include <fofi/FoFiType1C.h>
-#include <fofi/FoFiTrueType.h>
-#include <splash/SplashBitmap.h>
-
-#include "HTMLRenderer.h"
-#include "BackgroundRenderer.h"
-#include "Consts.h"
-#include "util.h"
-#include "config.h"
-
-/*
- * CSS classes
- *
- * p - Page
- * l - Line
- * w - White space
- * i - Image
- *
- * 
- * Reusable CSS classes
- *
- * f<hex> - Font (also for font names)
- * s<hex> - font Size
- * w<hex> - White space
- * t<hex> - Transform matrix
- * c<hex> - Color
- *
- */
-
-HTMLRenderer::HTMLRenderer(const Param * param)
-    :line_opened(false)
-    ,html_fout(param->output_filename.c_str(), ofstream::binary)
-    ,allcss_fout("all.css")
-    ,fontscript_fout(TMP_DIR+"/convert.pe")
-    ,image_count(0)
-    ,param(param)
-{
-    // install default font & size
-    install_font(nullptr);
-    install_font_size(0);
-
-    install_transform_matrix(id_matrix);
-
-    GfxRGB black;
-    black.r = black.g = black.b = 0;
-    install_color(&black);
-}
-
-HTMLRenderer::~HTMLRenderer()
-{
-}
-
-void HTMLRenderer::process(PDFDoc *doc)
-{
-    std::cerr << "Processing Text: ";
-    write_html_head();
-    xref = doc->getXRef();
-    for(int i = param->first_page; i <= param->last_page ; ++i) 
-    {
-        doc->displayPage(this, i, param->h_dpi, param->v_dpi,
-                0, true, false, false,
-                nullptr, nullptr, nullptr, nullptr);
-
-        std::cerr << ".";
-        std::cerr.flush();
-    }
-    write_html_tail();
-    std::cerr << std::endl;
-
-    if(param->process_nontext)
-    {
-        // Render non-text objects as image
-        std::cerr << "Processing Others: ";
-        // copied from poppler
-        SplashColor color;
-        color[0] = color[1] = color[2] = 255;
-
-        auto bg_renderer = new BackgroundRenderer(splashModeRGB8, 4, gFalse, color);
-        bg_renderer->startDoc(doc);
-
-        for(int i = param->first_page; i <= param->last_page ; ++i) 
-        {
-            doc->displayPage(bg_renderer, i, param->h_dpi2, param->v_dpi2,
-                    0, true, false, false,
-                    nullptr, nullptr, nullptr, nullptr);
-            bg_renderer->getBitmap()->writeImgFile(splashFormatPng, (char*)(boost::format("p%|1$x|.png")%i).str().c_str(), param->h_dpi2, param->v_dpi2);
-
-            std::cerr << ".";
-            std::cerr.flush();
-        }
-        delete bg_renderer;
-        std::cerr << std::endl;
-    }
-}
-
-void HTMLRenderer::write_html_head()
-{
-    html_fout << boost::filesystem::ifstream(PDF2HTMLEX_LIB_PATH / "head.html", ifstream::binary).rdbuf();
-}
-
-void HTMLRenderer::write_html_tail()
-{
-    html_fout << boost::filesystem::ifstream(PDF2HTMLEX_LIB_PATH / "tail.html", ifstream::binary).rdbuf();
-}
-
-void HTMLRenderer::startPage(int pageNum, GfxState *state) 
-{
-    this->pageNum = pageNum;
-    this->pageWidth = state->getPageWidth();
-    this->pageHeight = state->getPageHeight();
-
-    assert(!line_opened);
-
-    html_fout << boost::format("<div id=\"page-%3%\" class=\"p\" style=\"width:%1%px;height:%2%px;") % pageWidth % pageHeight % pageNum;
-
-    html_fout << boost::format("background-image:url(p%|3$x|.png);background-position:0 0;background-size:%1%px %2%px;background-repeat:no-repeat;") % pageWidth % pageHeight % pageNum;
-            
-    html_fout << "\">" << endl;
-
-    cur_fn_id = cur_fs_id = cur_tm_id = cur_color_id = 0;
-    cur_tx = cur_ty = 0;
-    cur_font_size = 0;
-
-    memcpy(cur_ctm, id_matrix, sizeof(cur_ctm));
-    memcpy(draw_ctm, id_matrix, sizeof(draw_ctm));
-    draw_font_size = 0;
-    draw_scale = 1.0;
-    draw_tx = draw_ty = 0;
-
-    cur_color.r = cur_color.g = cur_color.b = 0;
-    
-    reset_state_track();
-}
-
-void HTMLRenderer::endPage() {
-    close_cur_line();
-    // close page
-    html_fout << "</div>" << endl;
-}
-
-
-void HTMLRenderer::drawString(GfxState * state, GooString * s)
-{
-    if(s->getLength() == 0)
-        return;
-
-    auto font = state->getFont();
-    if((font == nullptr) || (font->getWMode()))
-    {
-        return;
-    }
-
-    //hidden
-    if((state->getRender() & 3) == 3)
-    {
-        return;
-    }
-
-    // see if the line has to be closed due to state change
-    check_state_change(state);
-    
-    // if the line is still open, try to merge with it
-    if(line_opened)
-    {
-        double target = (cur_tx - draw_tx) * draw_scale;
-        if(target > -param->h_eps)
-        {
-            if(target > param->h_eps)
-            {
-                double w;
-                auto wid = install_whitespace(target, w);
-                html_fout << boost::format("<span class=\"w w%|1$x|\"> </span>") % wid;
-                draw_tx += w / draw_scale;
-            }
-        }
-        else
-        {
-            // or can we shift left using simple tags?
-            close_cur_line();
-        }
-    }
-
-    if(!line_opened)
-    {
-        // have to open a new line
-        
-        // classes
-        html_fout << "<div class=\"l "
-            << boost::format("f%|1$x| s%|2$x| c%|3$x|") % cur_fn_id % cur_fs_id % cur_color_id;
-        
-        // "t0" is the id_matrix
-        if(cur_tm_id != 0)
-            html_fout << boost::format(" t%|1$x|") % cur_tm_id;
-
-        {
-            double x,y; // in user space
-            state->transform(state->getCurX(), state->getCurY(), &x, &y);
-            // TODO: recheck descent/ascent
-            html_fout << "\" style=\""
-                << "bottom:" << (y + state->getFont()->getDescent() * draw_font_size) << "px;"
-                << "top:" << (pageHeight - y - state->getFont()->getAscent() * draw_font_size) << "px;"
-                << "left:" << x << "px;"
-                ;
-        }
-        
-        // TODO: tracking
-        // letter & word spacing
-        if(_is_positive(state->getCharSpace()))
-            html_fout << "letter-spacing:" << state->getCharSpace() << "px;";
-        if(_is_positive(state->getWordSpace()))
-            html_fout << "word-spacing:" << state->getWordSpace() << "px;";
-
-        //debug 
-        //real pos & hori_scale
-        if(0)
-        {
-#if 0
-            html_fout << "\"";
-            double x,y;
-            state->transform(state->getCurX(), state->getCurY(), &x, &y);
-            html_fout << boost::format("data-lx=\"%5%\" data-ly=\"%6%\" data-drawscale=\"%4%\" data-x=\"%1%\" data-y=\"%2%\" data-hs=\"%3%")
-                %x%y%(state->getHorizScaling())%draw_scale%state->getLineX()%state->getLineY();
-#endif
-        }
-
-        html_fout << "\">";
-
-        line_opened = true;
-
-        draw_tx = cur_tx;
-    }
-
-
-    // Now ready to output
-    // get the unicodes
-    char *p = s->getCString();
-    int len = s->getLength();
-
-    double dx = 0;
-    double dy = 0;
-    double dx1,dy1;
-    double ox, oy;
-
-    int nChars = 0;
-    int nSpaces = 0;
-    int uLen;
-
-    CharCode code;
-    Unicode *u = nullptr;
-
-    while (len > 0) {
-        auto n = font->getNextChar(p, len, &code, &u, &uLen, &dx1, &dy1, &ox, &oy);
-
-        if(!(_equal(ox, 0) && _equal(oy, 0)))
-        {
-            std::cerr << "TODO: non-zero origins" << std::endl;
-        }
-
-        if(uLen == 0)
-        {
-            // TODO
-#if 0
-            CharCode c = 0;
-            for(int i = 0; i < n; ++i)
-            {
-                c = (c<<8) | (code&0xff);
-                code >>= 8;
-            }
-            for(int i = 0; i < n; ++i)
-            {
-                Unicode u = (c&0xff);
-                c >>= 8;
-                outputUnicodes(html_fout, &u, 1);
-            }
-#endif
-        }
-        else
-        {
-            outputUnicodes(html_fout, u, uLen);
-        }
-
-        dx += dx1;
-        dy += dy1;
-
-        if (n == 1 && *p == ' ') 
-        {
-            ++nSpaces;
-        }
-
-        ++nChars;
-        p += n;
-        len -= n;
-    }
-
-    dx = (dx * state->getFontSize() 
-            + nChars * state->getCharSpace() 
-            + nSpaces * state->getWordSpace()) * state->getHorizScaling();
-    
-    dy *= state->getFontSize();
-
-    cur_tx += dx;
-    cur_ty += dy;
-        
-    draw_tx += dx;
-    draw_ty += dy;
-}
-
-void HTMLRenderer::drawImage(GfxState * state, Object * ref, Stream * str, int width, int height, GfxImageColorMap * colorMap, GBool interpolate, int *maskColors, GBool inlineImg)
-{
-    if(maskColors)
-        return;
-
-    boost::gil::rgb8_image_t img(width, height);
-    auto imgview = view(img);
-    auto loc = imgview.xy_at(0,0);
-
-    ImageStream * img_stream = new ImageStream(str, width, colorMap->getNumPixelComps(), colorMap->getBits());
-    img_stream->reset();
-
-    for(int i = 0; i < height; ++i)
-    {
-        auto p = img_stream->getLine();
-        for(int j = 0; j < width; ++j)
-        {
-            GfxRGB rgb;
-            colorMap->getRGB(p, &rgb);
-
-            *loc = boost::gil::rgb8_pixel_t(colToByte(rgb.r), colToByte(rgb.g), colToByte(rgb.b));
-
-            p += colorMap->getNumPixelComps();
-
-            ++ loc.x();
-        }
-
-        loc = imgview.xy_at(0, i+1);
-    }
-
-    boost::gil::png_write_view((boost::format("i%|1$x|.png")%image_count).str(), imgview);
-    
-    img_stream->close();
-    delete img_stream;
-
-    close_cur_line();
-
-    double * ctm = state->getCTM();
-    ctm[4] = ctm[5] = 0.0;
-    html_fout << boost::format("<img class=\"i t%2%\" style=\"left:%3%px;bottom:%4%px;width:%5%px;height:%6%px;\" src=\"i%|1$x|.png\" />") % image_count % install_transform_matrix(ctm) % state->getCurX() % state->getCurY() % width % height << endl;
-
-
-    ++ image_count;
-}
-
-
-
-
-

src/HTMLRenderer/export.cc

@@ -12,6 +12,26 @@
 #include <boost/algorithm/string.hpp>
 #include <boost/format.hpp>
 
+
+/*
+ * CSS classes
+ *
+ * p - Page
+ * l - Line
+ * w - White space
+ * i - Image
+ *
+ * 
+ * Reusable CSS classes
+ *
+ * f<hex> - Font (also for font names)
+ * s<hex> - font Size
+ * w<hex> - White space
+ * t<hex> - Transform matrix
+ * c<hex> - Color
+ *
+ */
+
 void HTMLRenderer::export_remote_font(long long fn_id, const string & suffix, const string & format, GfxFont * font)
 {
     allcss_fout << boost::format("@font-face{font-family:f%|1$x|;src:url(f%|1$x|%2%)format(\"%3%\");}.f%|1$x|{font-family:f%|1$x|;") % fn_id % suffix % format;

src/HTMLRenderer/general.cc

@@ -0,0 +1,139 @@
+/*
+ * general.cc
+ *
+ * Hanlding general stuffs
+ *
+ * TODO: better name for this file?
+ *
+ * by WangLu
+ * 2012.08.14
+ */
+
+#include <iostream>
+
+#include <boost/format.hpp>
+#include <boost/filesystem/fstream.hpp>
+
+#include <splash/SplashBitmap.h>
+
+#include "HTMLRenderer.h"
+#include "BackgroundRenderer.h"
+#include "config.h"
+
+HTMLRenderer::HTMLRenderer(const Param * param)
+    :line_opened(false)
+    ,html_fout(param->output_filename.c_str(), ofstream::binary)
+    ,allcss_fout("all.css")
+    ,fontscript_fout(TMP_DIR+"/convert.pe")
+    ,image_count(0)
+    ,param(param)
+{
+    // install default font & size
+    install_font(nullptr);
+    install_font_size(0);
+
+    install_transform_matrix(id_matrix);
+
+    GfxRGB black;
+    black.r = black.g = black.b = 0;
+    install_color(&black);
+}
+
+HTMLRenderer::~HTMLRenderer()
+{ }
+
+void HTMLRenderer::process(PDFDoc *doc)
+{
+    std::cerr << "Processing Text: ";
+    write_html_head();
+    xref = doc->getXRef();
+    for(int i = param->first_page; i <= param->last_page ; ++i) 
+    {
+        doc->displayPage(this, i, param->h_dpi, param->v_dpi,
+                0, true, false, false,
+                nullptr, nullptr, nullptr, nullptr);
+
+        std::cerr << ".";
+        std::cerr.flush();
+    }
+    write_html_tail();
+    std::cerr << std::endl;
+
+    if(param->process_nontext)
+    {
+        // Render non-text objects as image
+        std::cerr << "Processing Others: ";
+        // copied from poppler
+        SplashColor color;
+        color[0] = color[1] = color[2] = 255;
+
+        auto bg_renderer = new BackgroundRenderer(splashModeRGB8, 4, gFalse, color);
+        bg_renderer->startDoc(doc);
+
+        for(int i = param->first_page; i <= param->last_page ; ++i) 
+        {
+            doc->displayPage(bg_renderer, i, param->h_dpi2, param->v_dpi2,
+                    0, true, false, false,
+                    nullptr, nullptr, nullptr, nullptr);
+            bg_renderer->getBitmap()->writeImgFile(splashFormatPng, (char*)(boost::format("p%|1$x|.png")%i).str().c_str(), param->h_dpi2, param->v_dpi2);
+
+            std::cerr << ".";
+            std::cerr.flush();
+        }
+        delete bg_renderer;
+        std::cerr << std::endl;
+    }
+}
+
+void HTMLRenderer::write_html_head()
+{
+    html_fout << boost::filesystem::ifstream(PDF2HTMLEX_LIB_PATH / "head.html", ifstream::binary).rdbuf();
+}
+
+void HTMLRenderer::write_html_tail()
+{
+    html_fout << boost::filesystem::ifstream(PDF2HTMLEX_LIB_PATH / "tail.html", ifstream::binary).rdbuf();
+}
+
+void HTMLRenderer::startPage(int pageNum, GfxState *state) 
+{
+    this->pageNum = pageNum;
+    this->pageWidth = state->getPageWidth();
+    this->pageHeight = state->getPageHeight();
+
+    assert(!line_opened);
+
+    html_fout << boost::format("<div id=\"page-%3%\" class=\"p\" style=\"width:%1%px;height:%2%px;") % pageWidth % pageHeight % pageNum;
+
+    html_fout << boost::format("background-image:url(p%|3$x|.png);background-position:0 0;background-size:%1%px %2%px;background-repeat:no-repeat;") % pageWidth % pageHeight % pageNum;
+            
+    html_fout << "\">" << endl;
+
+    cur_fn_id = cur_fs_id = cur_tm_id = cur_color_id = 0;
+    cur_tx = cur_ty = 0;
+    cur_font_size = 0;
+
+    memcpy(cur_ctm, id_matrix, sizeof(cur_ctm));
+    memcpy(draw_ctm, id_matrix, sizeof(draw_ctm));
+    draw_font_size = 0;
+    draw_scale = 1.0;
+    draw_tx = draw_ty = 0;
+
+    cur_color.r = cur_color.g = cur_color.b = 0;
+    
+    reset_state_track();
+}
+
+void HTMLRenderer::endPage() {
+    close_cur_line();
+    // close page
+    html_fout << "</div>" << endl;
+}
+
+
+
+
+
+
+
+

src/HTMLRenderer/image.cc

@@ -0,0 +1,62 @@
+/*
+ * image.cc
+ *
+ * Handling images
+ *
+ * by WangLu
+ * 2012.08.14
+ */
+
+#include <boost/format.hpp>
+// for gil bug
+const int *int_p_NULL = nullptr;
+#include <boost/gil/gil_all.hpp>
+#include <boost/gil/extension/io/png_dynamic_io.hpp>
+
+#include "HTMLRenderer.h"
+
+
+void HTMLRenderer::drawImage(GfxState * state, Object * ref, Stream * str, int width, int height, GfxImageColorMap * colorMap, GBool interpolate, int *maskColors, GBool inlineImg)
+{
+    if(maskColors)
+        return;
+
+    boost::gil::rgb8_image_t img(width, height);
+    auto imgview = view(img);
+    auto loc = imgview.xy_at(0,0);
+
+    ImageStream * img_stream = new ImageStream(str, width, colorMap->getNumPixelComps(), colorMap->getBits());
+    img_stream->reset();
+
+    for(int i = 0; i < height; ++i)
+    {
+        auto p = img_stream->getLine();
+        for(int j = 0; j < width; ++j)
+        {
+            GfxRGB rgb;
+            colorMap->getRGB(p, &rgb);
+
+            *loc = boost::gil::rgb8_pixel_t(colToByte(rgb.r), colToByte(rgb.g), colToByte(rgb.b));
+
+            p += colorMap->getNumPixelComps();
+
+            ++ loc.x();
+        }
+
+        loc = imgview.xy_at(0, i+1);
+    }
+
+    boost::gil::png_write_view((boost::format("i%|1$x|.png")%image_count).str(), imgview);
+    
+    img_stream->close();
+    delete img_stream;
+
+    close_cur_line();
+
+    double * ctm = state->getCTM();
+    ctm[4] = ctm[5] = 0.0;
+    html_fout << boost::format("<img class=\"i t%2%\" style=\"left:%3%px;bottom:%4%px;width:%5%px;height:%6%px;\" src=\"i%|1$x|.png\" />") % image_count % install_transform_matrix(ctm) % state->getCurX() % state->getCurY() % width % height << endl;
+
+
+    ++ image_count;
+}

src/HTMLRenderer/install.cc

@@ -91,124 +91,8 @@ long long HTMLRenderer::install_font(GfxFont * font)
 
 }
 
-std::string HTMLRenderer::dump_embedded_font (GfxFont * font, long long fn_id)
+// TODO
-{
+// add a new function and move to text.cc
-    // mupdf consulted
-    
-    Object ref_obj, font_obj, font_obj2, fontdesc_obj;
-    Object obj, obj1, obj2;
-    Dict * dict = nullptr;
-
-    std::string suffix, subtype;
-
-    char buf[1024];
-    int len;
-
-    ofstream outf;
-
-    auto * id = font->getID();
-    ref_obj.initRef(id->num, id->gen);
-    ref_obj.fetch(xref, &font_obj);
-    ref_obj.free();
-
-    if(!font_obj.isDict())
-    {
-        std::cerr << "Font object is not a dictionary" << std::endl;
-        goto err;
-    }
-
-    dict = font_obj.getDict();
-    if(dict->lookup("DescendantFonts", &font_obj2)->isArray())
-    {
-        if(font_obj2.arrayGetLength() == 0)
-        {
-            std::cerr << "Warning: empty DescendantFonts array" << std::endl;
-        }
-        else
-        {
-            if(font_obj2.arrayGetLength() > 1)
-                std::cerr << "TODO: multiple entries in DescendantFonts array" << std::endl;
-
-            if(font_obj2.arrayGet(0, &obj2)->isDict())
-            {
-                dict = obj2.getDict();
-            }
-        }
-    }
-
-    if(!dict->lookup("FontDescriptor", &fontdesc_obj)->isDict())
-    {
-        std::cerr << "Cannot find FontDescriptor " << std::endl;
-        goto err;
-    }
-
-    dict = fontdesc_obj.getDict();
-    
-    if(dict->lookup("FontFile3", &obj)->isStream())
-    {
-        if(obj.streamGetDict()->lookup("Subtype", &obj1)->isName())
-        {
-            subtype = obj1.getName();
-            if(subtype == "Type1C")
-            {
-                suffix = ".cff";
-            }
-            else if (subtype == "CIDFontType0C")
-            {
-                suffix = ".cid";
-            }
-            else
-            {
-                std::cerr << "Unknown subtype: " << subtype << std::endl;
-                goto err;
-            }
-        }
-        else
-        {
-            std::cerr << "Invalid subtype in font descriptor" << std::endl;
-            goto err;
-        }
-    }
-    else if (dict->lookup("FontFile2", &obj)->isStream())
-    { 
-        suffix = ".ttf";
-    }
-    else if (dict->lookup("FontFile", &obj)->isStream())
-    {
-        suffix = ".ttf";
-    }
-    else
-    {
-        std::cerr << "Cannot find FontFile for dump" << std::endl;
-        goto err;
-    }
-
-    if(suffix == "")
-    {
-        std::cerr << "Font type unrecognized" << std::endl;
-        goto err;
-    }
-
-    obj.streamReset();
-    outf.open((boost::format("%1%/f%|2$x|%3%")%TMP_DIR%fn_id%suffix).str().c_str(), ofstream::binary);
-    while((len = obj.streamGetChars(1024, (Guchar*)buf)) > 0)
-    {
-        outf.write(buf, len);
-    }
-    outf.close();
-    obj.streamClose();
-
-err:
-    obj2.free();
-    obj1.free();
-    obj.free();
-
-    fontdesc_obj.free();
-    font_obj2.free();
-    font_obj.free();
-    return suffix;
-}
-
 void HTMLRenderer::install_embedded_font(GfxFont * font, const std::string & suffix, long long fn_id)
 {
     // TODO Should use standard way to handle CID fonts

src/HTMLRenderer/text.cc

@@ -0,0 +1,298 @@
+/*
+ * text.ccc
+ *
+ * Handling text and relative stuffs
+ *
+ * by WangLu
+ * 2012.08.14
+ */
+
+#include <iostream>
+
+#include <boost/format.hpp>
+
+#include "HTMLRenderer.h"
+
+std::string HTMLRenderer::dump_embedded_font (GfxFont * font, long long fn_id)
+{
+    // mupdf consulted
+    
+    Object ref_obj, font_obj, font_obj2, fontdesc_obj;
+    Object obj, obj1, obj2;
+    Dict * dict = nullptr;
+
+    std::string suffix, subtype;
+
+    char buf[1024];
+    int len;
+
+    ofstream outf;
+
+    auto * id = font->getID();
+    ref_obj.initRef(id->num, id->gen);
+    ref_obj.fetch(xref, &font_obj);
+    ref_obj.free();
+
+    if(!font_obj.isDict())
+    {
+        std::cerr << "Font object is not a dictionary" << std::endl;
+        goto err;
+    }
+
+    dict = font_obj.getDict();
+    if(dict->lookup("DescendantFonts", &font_obj2)->isArray())
+    {
+        if(font_obj2.arrayGetLength() == 0)
+        {
+            std::cerr << "Warning: empty DescendantFonts array" << std::endl;
+        }
+        else
+        {
+            if(font_obj2.arrayGetLength() > 1)
+                std::cerr << "TODO: multiple entries in DescendantFonts array" << std::endl;
+
+            if(font_obj2.arrayGet(0, &obj2)->isDict())
+            {
+                dict = obj2.getDict();
+            }
+        }
+    }
+
+    if(!dict->lookup("FontDescriptor", &fontdesc_obj)->isDict())
+    {
+        std::cerr << "Cannot find FontDescriptor " << std::endl;
+        goto err;
+    }
+
+    dict = fontdesc_obj.getDict();
+    
+    if(dict->lookup("FontFile3", &obj)->isStream())
+    {
+        if(obj.streamGetDict()->lookup("Subtype", &obj1)->isName())
+        {
+            subtype = obj1.getName();
+            if(subtype == "Type1C")
+            {
+                suffix = ".cff";
+            }
+            else if (subtype == "CIDFontType0C")
+            {
+                suffix = ".cid";
+            }
+            else
+            {
+                std::cerr << "Unknown subtype: " << subtype << std::endl;
+                goto err;
+            }
+        }
+        else
+        {
+            std::cerr << "Invalid subtype in font descriptor" << std::endl;
+            goto err;
+        }
+    }
+    else if (dict->lookup("FontFile2", &obj)->isStream())
+    { 
+        suffix = ".ttf";
+    }
+    else if (dict->lookup("FontFile", &obj)->isStream())
+    {
+        suffix = ".ttf";
+    }
+    else
+    {
+        std::cerr << "Cannot find FontFile for dump" << std::endl;
+        goto err;
+    }
+
+    if(suffix == "")
+    {
+        std::cerr << "Font type unrecognized" << std::endl;
+        goto err;
+    }
+
+    obj.streamReset();
+    outf.open((boost::format("%1%/f%|2$x|%3%")%TMP_DIR%fn_id%suffix).str().c_str(), ofstream::binary);
+    while((len = obj.streamGetChars(1024, (Guchar*)buf)) > 0)
+    {
+        outf.write(buf, len);
+    }
+    outf.close();
+    obj.streamClose();
+
+err:
+    obj2.free();
+    obj1.free();
+    obj.free();
+
+    fontdesc_obj.free();
+    font_obj2.free();
+    font_obj.free();
+    return suffix;
+}
+
+void HTMLRenderer::drawString(GfxState * state, GooString * s)
+{
+    if(s->getLength() == 0)
+        return;
+
+    auto font = state->getFont();
+    if((font == nullptr) || (font->getWMode()))
+    {
+        return;
+    }
+
+    //hidden
+    if((state->getRender() & 3) == 3)
+    {
+        return;
+    }
+
+    // see if the line has to be closed due to state change
+    check_state_change(state);
+    
+    // if the line is still open, try to merge with it
+    if(line_opened)
+    {
+        double target = (cur_tx - draw_tx) * draw_scale;
+        if(target > -param->h_eps)
+        {
+            if(target > param->h_eps)
+            {
+                double w;
+                auto wid = install_whitespace(target, w);
+                html_fout << boost::format("<span class=\"w w%|1$x|\"> </span>") % wid;
+                draw_tx += w / draw_scale;
+            }
+        }
+        else
+        {
+            // or can we shift left using simple tags?
+            close_cur_line();
+        }
+    }
+
+    if(!line_opened)
+    {
+        // have to open a new line
+        
+        // classes
+        html_fout << "<div class=\"l "
+            << boost::format("f%|1$x| s%|2$x| c%|3$x|") % cur_fn_id % cur_fs_id % cur_color_id;
+        
+        // "t0" is the id_matrix
+        if(cur_tm_id != 0)
+            html_fout << boost::format(" t%|1$x|") % cur_tm_id;
+
+        {
+            double x,y; // in user space
+            state->transform(state->getCurX(), state->getCurY(), &x, &y);
+            // TODO: recheck descent/ascent
+            html_fout << "\" style=\""
+                << "bottom:" << (y + state->getFont()->getDescent() * draw_font_size) << "px;"
+                << "top:" << (pageHeight - y - state->getFont()->getAscent() * draw_font_size) << "px;"
+                << "left:" << x << "px;"
+                ;
+        }
+        
+        // TODO: tracking
+        // letter & word spacing
+        if(_is_positive(state->getCharSpace()))
+            html_fout << "letter-spacing:" << state->getCharSpace() << "px;";
+        if(_is_positive(state->getWordSpace()))
+            html_fout << "word-spacing:" << state->getWordSpace() << "px;";
+
+        //debug 
+        //real pos & hori_scale
+        if(0)
+        {
+#if 0
+            html_fout << "\"";
+            double x,y;
+            state->transform(state->getCurX(), state->getCurY(), &x, &y);
+            html_fout << boost::format("data-lx=\"%5%\" data-ly=\"%6%\" data-drawscale=\"%4%\" data-x=\"%1%\" data-y=\"%2%\" data-hs=\"%3%")
+                %x%y%(state->getHorizScaling())%draw_scale%state->getLineX()%state->getLineY();
+#endif
+        }
+
+        html_fout << "\">";
+
+        line_opened = true;
+
+        draw_tx = cur_tx;
+    }
+
+
+    // Now ready to output
+    // get the unicodes
+    char *p = s->getCString();
+    int len = s->getLength();
+
+    double dx = 0;
+    double dy = 0;
+    double dx1,dy1;
+    double ox, oy;
+
+    int nChars = 0;
+    int nSpaces = 0;
+    int uLen;
+
+    CharCode code;
+    Unicode *u = nullptr;
+
+    while (len > 0) {
+        auto n = font->getNextChar(p, len, &code, &u, &uLen, &dx1, &dy1, &ox, &oy);
+
+        if(!(_equal(ox, 0) && _equal(oy, 0)))
+        {
+            std::cerr << "TODO: non-zero origins" << std::endl;
+        }
+
+        if(uLen == 0)
+        {
+            // TODO
+#if 0
+            CharCode c = 0;
+            for(int i = 0; i < n; ++i)
+            {
+                c = (c<<8) | (code&0xff);
+                code >>= 8;
+            }
+            for(int i = 0; i < n; ++i)
+            {
+                Unicode u = (c&0xff);
+                c >>= 8;
+                outputUnicodes(html_fout, &u, 1);
+            }
+#endif
+        }
+        else
+        {
+            outputUnicodes(html_fout, u, uLen);
+        }
+
+        dx += dx1;
+        dy += dy1;
+
+        if (n == 1 && *p == ' ') 
+        {
+            ++nSpaces;
+        }
+
+        ++nChars;
+        p += n;
+        len -= n;
+    }
+
+    dx = (dx * state->getFontSize() 
+            + nChars * state->getCharSpace() 
+            + nSpaces * state->getWordSpace()) * state->getHorizScaling();
+    
+    dy *= state->getFontSize();
+
+    cur_tx += dx;
+    cur_ty += dy;
+        
+    draw_tx += dx;
+    draw_ty += dy;
+}