fix: robustify page count on broken pdf's

pyproject.toml

@@ -64,6 +64,7 @@ build = [
 ]
 dev = [
     "pytest>=7.4.2,<8.0.0",
+    "pypdf>=5.0.0,<6.0.0",
     "black[jupyter]>=24.4.2,<26.0.0",
     "python-semantic-release>=7.32.2,<8.0.0",
     "pre-commit>=3.7.1,<4.0.0",

src/parse/pdf_decoders/document.h

@@ -46,7 +46,7 @@ namespace pdflib
     // Decode a single page and return the page decoder directly
     page_decoder_ptr decode_page(int page_number,
                                  const decode_config& config);
-    
+
     // New: Direct access to page decoders (typed API)
     bool has_page_decoder(int page_number);
     page_decoder_ptr get_page_decoder(int page_number);
@@ -181,27 +181,64 @@ namespace pdflib
 
   bool pdf_decoder<DOCUMENT>::process_document_components()
   {
+    LOG_S(INFO) << __FUNCTION__;
+
     utils::timer timer;
 
     if(qpdf_root.hasKey("/Pages"))
       {
         qpdf_pages = qpdf_root.getKey("/Pages");
 
+        int _number_of_pages = -1;
         if(qpdf_pages.hasKey("/Count"))
           {
-            number_of_pages = qpdf_pages.getKey("/Count").getIntValue();
+            _number_of_pages = qpdf_pages.getKey("/Count").getIntValue();
+            //LOG_S(WARNING) << "`/Count` (before): " << _number_of_pages;
           }
-        else
+
+	// Be aware that this operation does some normalization
+        number_of_pages = 0;
+        for(QPDFObjectHandle page : qpdf_document.getAllPages()) 
+          {
+            number_of_pages += 1;
+          }
+        LOG_S(INFO) << "#-pages (from `qpdf_document.getAllPages()`): " << number_of_pages;
+
+	if(number_of_pages!=_number_of_pages and qpdf_pages.hasKey("/Count"))
           {
-            LOG_S(WARNING) << "filename: " << filename << " has no `/Count`";
-            number_of_pages = 0;
-            for(QPDFObjectHandle page : qpdf_document.getAllPages())
-              {
-                number_of_pages += 1;
-              }
+	    LOG_S(WARNING) << "`/Count` before (=" << _number_of_pages << ") != "
+			   << " len(`/Pages`) (=" << number_of_pages << ")";
           }
+
+        /*
+          if(qpdf_pages.hasKey("/Count"))
+          {
+          int __number_of_pages = qpdf_pages.getKey("/Count").getIntValue();
+          LOG_S(WARNING) << "`/Count` (after): " << __number_of_pages;
 
-        LOG_S(INFO) << "#-pages: " << number_of_pages;
+          if(_number_of_pages!=__number_of_pages)
+          {
+          LOG_S(WARNING) << "`/Count` before (=" << _number_of_pages << ") != "
+          << "`/Count` after (=" << __number_of_pages << ")";
+          }
+
+          if(number_of_pages!=_number_of_pages)
+          {
+          LOG_S(WARNING) << "`/Count` before (=" << _number_of_pages << ") != "
+          << " len(`/Pages`) (=" << number_of_pages << ")";
+          }
+
+          if(number_of_pages!=__number_of_pages)
+          {
+          LOG_S(WARNING) << "`/Count` after (=" << __number_of_pages << ") != "
+          << " len(`/Pages`) (=" << number_of_pages << ")";
+          }
+          }
+          else
+          {
+          LOG_S(WARNING) << "filename: " << filename << " has no `/Count`";
+          }
+        */
       }
     else
       {
@@ -434,7 +471,7 @@ namespace pdflib
       if(config.do_thread_safe)
         {
           // creates its own QPDF document
-          page_decoder = std::make_shared<pdf_decoder<PAGE>>(buffer, password, page_number);	  
+          page_decoder = std::make_shared<pdf_decoder<PAGE>>(buffer, password, page_number);
         }
       else
         {

src/parse/pdf_resources/page_font/base_fonts.h

@@ -162,7 +162,7 @@ namespace pdflib
   {
     if(initialized)
       {
-	LOG_S(WARNING) << "skipping base_fonts::initialise, already initialized ...";
+	LOG_S(INFO) << "skipping base_fonts::initialise, already initialized ...";
 	return;
       }
 

src/parse/pdf_resources/page_font/encodings.h

@@ -43,7 +43,7 @@ namespace pdflib
   {
     if(initialized)
       {
-	LOG_S(WARNING) << "skipping font_encodings::initialise, already initialized ...";
+	LOG_S(INFO) << "skipping font_encodings::initialise, already initialized ...";
 	return;
       }
 

src/parse/pdf_resources/page_font/font_cids.h

@@ -87,7 +87,7 @@ namespace pdflib
   {
     if(initialized)
       {
-	LOG_S(WARNING) << "skipping font_cids::initialise, already initialized ...";
+	LOG_S(INFO) << "skipping font_cids::initialise, already initialized ...";
 	return;
       }
 

src/parse/pdf_resources/page_font/glyphs.h

@@ -101,14 +101,15 @@ namespace pdflib
     LOG_S(ERROR) << "could not find a glyph with name=" << key;
     unknown_glyphs.insert(key);
 
-    return "glyph["+key+"]";
+    // FIXME: we should not do this, especially if the decode_config does not allow this! 
+    return "GLYPH["+key+"]";
   }
 
   void font_glyphs::initialise(std::string dirname)
   {
     if(initialized)
       {
-	LOG_S(WARNING) << "skipping font_glyphs::initialise, already initialized ...";
+	LOG_S(INFO) << "skipping font_glyphs::initialise, already initialized ...";
 	return;
       }
 
@@ -215,7 +216,7 @@ namespace pdflib
 
   void font_glyphs::read_file_uni(std::string filename)
   {
-    LOG_S(WARNING) << __FUNCTION__ << ": " << filename;
+    LOG_S(INFO) << __FUNCTION__ << ": " << filename;
 
     std::ifstream file(filename.c_str());
 

src/pybind/docling_parser.h

@@ -77,7 +77,7 @@ namespace docling
     pdf_resources_dir(resource_utils::get_resources_dir(true).string()),
     key2doc({})
   {
-    LOG_S(WARNING) << "pdf_resources_dir: " << pdf_resources_dir;
+    LOG_S(INFO) << "pdf_resources_dir: " << pdf_resources_dir;
 
     auto RESOURCE_DIR_KEY = pdflib::pdf_resource<pdflib::PAGE_FONT>::RESOURCE_DIR_KEY;
 
@@ -95,7 +95,7 @@ namespace docling
   {
     set_loglevel_with_label(level);
 
-    LOG_S(WARNING) << "pdf_resources_dir: " << pdf_resources_dir;
+    LOG_S(INFO) << "pdf_resources_dir: " << pdf_resources_dir;
 
     auto RESOURCE_DIR_KEY = pdflib::pdf_resource<pdflib::PAGE_FONT>::RESOURCE_DIR_KEY;