feat: add the export of annotations and ToC (#58)

Signed-off-by: Peter Staar <[email protected]>

Author: PeterStaar-IBM

app/pybind_parse.cpp

@@ -85,6 +85,12 @@ PYBIND11_MODULE(docling_parse, m) {
     .def("unload_document", &docling::docling_parser_v2::unload_document)
 
     .def("number_of_pages", &docling::docling_parser_v2::number_of_pages)
+
+    .def("get_annotations", &docling::docling_parser_v2::get_annotations,
+	 "Get annotations at the top-level of the document")
+    
+    .def("get_table_of_contents", &docling::docling_parser_v2::get_table_of_contents,
+	 "Get the table-of-contents (None if not available)")
 
     .def("parse_pdf_from_key",
 	 pybind11::overload_cast<std::string>(&docling::docling_parser_v2::parse_pdf_from_key),

docling_parse/visualize.py

@@ -212,6 +212,24 @@ def visualise_v1(
             img.save(oname)
 
 
+def draw_annotations(draw, annot, H, W):
+
+    if "/Rect" in annot:
+        bbox = annot["/Rect"]
+
+        bl = (bbox[0], H - bbox[1])
+        br = (bbox[2], H - bbox[1])
+        tr = (bbox[2], H - bbox[3])
+        tl = (bbox[0], H - bbox[3])
+
+        # Draw the rectangle as a polygon
+        draw.polygon([bl, br, tr, tl], outline="white", fill="green")
+
+    if "/Kids" in annot:
+        for _ in annot["/Kids"]:
+            draw_annotations(draw, annot, H, W)
+
+
 def visualise_v2(
     log_level: str,
     pdf_path: str,
@@ -260,6 +278,8 @@ def visualise_v2(
 
             lines = page[_]["lines"]
 
+            annots = page["annotations"]
+
             if PIL_INSTALLED:
 
                 W = dimension["width"]
@@ -286,24 +306,6 @@ def visualise_v2(
                     # Draw the rectangle as a polygon
                     draw.polygon([bl, br, tr, tl], outline="green", fill="yellow")
 
-                # Draw each rectangle by connecting its four points
-                for line in lines:
-
-                    i = line["i"]
-                    x = line["x"]
-                    y = line["y"]
-
-                    for l in range(0, len(i), 2):
-                        i0 = i[l + 0]
-                        i1 = i[l + 1]
-
-                        for k in range(i0, i1 - 1):
-                            draw.line(
-                                (x[k], H - y[k], x[k + 1], H - y[k + 1]),
-                                fill="black",
-                                width=3,
-                            )
-
                 # Draw each rectangle by connecting its four points
                 for row in cells:
 
@@ -330,6 +332,28 @@ def visualise_v2(
                     # You can change the outline and fill color
                     draw.polygon(rect, outline="red", fill="blue")
 
+                # Draw widgets
+                for annot in annots:
+                    draw_annotations(draw, annot, H, W)
+
+                # Draw each rectangle by connecting its four points
+                for line in lines:
+
+                    i = line["i"]
+                    x = line["x"]
+                    y = line["y"]
+
+                    for l in range(0, len(i), 2):
+                        i0 = i[l + 0]
+                        i1 = i[l + 1]
+
+                        for k in range(i0, i1 - 1):
+                            draw.line(
+                                (x[k], H - y[k], x[k + 1], H - y[k + 1]),
+                                fill="black",
+                                width=1,
+                            )
+
                 # Show the image
                 if interactive:
                     img.show()

src/pybind/docling_parser_v2.h

@@ -35,6 +35,9 @@ namespace docling
 
     int number_of_pages(std::string key);
 
+    nlohmann::json get_annotations(std::string key);
+    nlohmann::json get_table_of_contents(std::string key);
+    
     nlohmann::json parse_pdf_from_key(std::string key);
 
     nlohmann::json parse_pdf_from_key_on_page(std::string key, int page);
@@ -230,10 +233,40 @@ namespace docling
 
     return -1;
   }
+
+  nlohmann::json docling_parser_v2::get_annotations(std::string key)
+  {
+    LOG_S(INFO) << __FUNCTION__;
+
+    auto itr = key2doc.find(key);
+
+    if(itr==key2doc.end())
+      {
+	LOG_S(ERROR) << "key not found: " << key;
+	return nlohmann::json::value_t::null;	
+      }
+
+    return (itr->second)->get_annotations();
+  }
+  
+  nlohmann::json docling_parser_v2::get_table_of_contents(std::string key)
+  {
+    LOG_S(INFO) << __FUNCTION__;
+
+    auto itr = key2doc.find(key);
+
+    if(itr==key2doc.end())
+      {
+	LOG_S(ERROR) << "key not found: " << key;
+	return nlohmann::json::value_t::null;	
+      }
+
+    return (itr->second)->get_table_of_contents();
+  }
 
   nlohmann::json docling_parser_v2::parse_pdf_from_key(std::string key)
   {
-    LOG_S(WARNING) << __FUNCTION__;
+    LOG_S(INFO) << __FUNCTION__;
 
     auto itr = key2doc.find(key);
 

src/v2/parser.h

@@ -24,8 +24,6 @@ namespace plib
     bool initialise(nlohmann::json& data);
 
   private:
-
-
 
     void execute_parse();
 

src/v2/pdf_decoders/document.h

@@ -21,6 +21,9 @@ namespace pdflib
     nlohmann::json get();
 
     int get_number_of_pages() { return number_of_pages; }
+
+    nlohmann::json get_annotations() { return json_annots; }
+    nlohmann::json get_table_of_contents() { return json_annots["table_of_contents"]; }
 
     bool process_document_from_file(std::string& _filename);
     bool process_document_from_bytesio(std::string& _buffer);
@@ -47,6 +50,8 @@ namespace pdflib
 
     int number_of_pages;    
 
+    //nlohmann::json json_toc; // table-of-contents
+    nlohmann::json json_annots;
     nlohmann::json json_document;
   };
 
@@ -56,15 +61,15 @@ namespace pdflib
 
     timings({}),
     qpdf_document(),
-
-    //qpdf_root(NULL),
-    //qpdf_pages(NULL),
 
     // have compatibulity between QPDF v10 and v11
     qpdf_root(),
     qpdf_pages(),
 
     number_of_pages(-1),
+
+    //json_toc(nlohmann::json::value_t::null),
+    json_annots(nlohmann::json::value_t::null),
     json_document(nlohmann::json::value_t::null)
   {}
 
@@ -75,14 +80,14 @@ namespace pdflib
     timings(timings_),
     qpdf_document(),
 
-    //qpdf_root(NULL),
-    //qpdf_pages(NULL),
-
     // have compatibulity between QPDF v10 and v11
     qpdf_root(),
     qpdf_pages(),
 
     number_of_pages(-1),
+
+    //json_toc(nlohmann::json::value_t::null),
+    json_annots(nlohmann::json::value_t::null),
     json_document(nlohmann::json::value_t::null)
   {}
 
@@ -91,6 +96,11 @@ namespace pdflib
 
   nlohmann::json pdf_decoder<DOCUMENT>::get()
   {
+    {
+      //json_document["table_of_contents"] = json_toc;      
+      json_document["annotations"] = json_annots;
+    }
+    
     {
       nlohmann::json& timings_ = json_document["timings"];
 
@@ -118,6 +128,9 @@ namespace pdflib
         qpdf_root  = qpdf_document.getRoot();
         qpdf_pages = qpdf_root.getKey("/Pages");
 
+	//json_toc = extract_toc_in_json(qpdf_root);
+	json_annots = extract_document_annotations_in_json(qpdf_document, qpdf_root);
+	
         number_of_pages = qpdf_pages.getKey("/Count").getIntValue();    
         LOG_S(INFO) << "#-pages: " << number_of_pages;
 
@@ -148,13 +161,17 @@ namespace pdflib
     try
       {
 	std::string description = "processing buffer";	
-        qpdf_document.processMemoryFile(description.c_str(), buffer.c_str(), buffer.size());
+        qpdf_document.processMemoryFile(description.c_str(),
+					buffer.c_str(), buffer.size());
 
         LOG_S(INFO) << "buffer processed by qpdf!";        
 
         qpdf_root  = qpdf_document.getRoot();
         qpdf_pages = qpdf_root.getKey("/Pages");
 
+	//json_toc = extract_toc_in_json(qpdf_root);
+	json_annots = extract_document_annotations_in_json(qpdf_document, qpdf_root);
+	
         number_of_pages = qpdf_pages.getKey("/Count").getIntValue();    
         LOG_S(INFO) << "#-pages: " << number_of_pages;
 
@@ -253,7 +270,8 @@ namespace pdflib
     timings[__FUNCTION__] = timer.get_time();
   }
 
-  void pdf_decoder<DOCUMENT>::update_timings(std::map<std::string, double>& timings_, bool set_timer)
+  void pdf_decoder<DOCUMENT>::update_timings(std::map<std::string, double>& timings_,
+					     bool set_timer)
   {
     for(auto itr=timings_.begin(); itr!=timings_.end(); itr++)
       {

src/v2/pdf_decoders/page.h

@@ -39,6 +39,8 @@ namespace pdflib
     // Contents
     void decode_contents();
 
+    void decode_annots();
+
     void sanitise_contents();
 
   private:
@@ -49,6 +51,8 @@ namespace pdflib
     QPDFObjectHandle qpdf_fonts;
     QPDFObjectHandle qpdf_xobjects;
 
+    nlohmann::json json_annots;
+    
     nlohmann::json json_page;
     nlohmann::json json_resources;
     nlohmann::json json_grphs;
@@ -85,6 +89,8 @@ namespace pdflib
   {
     nlohmann::json result;
     {
+      result["annotations"] = json_annots;
+      
       nlohmann::json& timings_ = result["timings"];
       {
 	for(auto itr=timings.begin(); itr!=timings.end(); itr++)
@@ -126,6 +132,8 @@ namespace pdflib
     utils::timer timer;
 
     json_page = to_json(qpdf_page);
+
+    json_annots = extract_annots_in_json(qpdf_page);
 
     try
       {
@@ -153,8 +161,10 @@ namespace pdflib
 
     decode_contents();
 
+    decode_annots();
+    
     sanitise_contents();
-
+    
     timings[__FUNCTION__] = timer.get_time();
 
     return timings;
@@ -264,6 +274,82 @@ namespace pdflib
           }
       }
 
+    timings[__FUNCTION__] = timer.get_time();
+  }
+
+  void pdf_decoder<PAGE>::decode_annots()
+  {
+    LOG_S(INFO) << __FUNCTION__;
+    utils::timer timer;
+    
+    //LOG_S(INFO) << "analyzing: " << json_annots.dump(2);
+    if(json_annots.is_array())
+      {
+	for(auto item:json_annots)
+	  {
+	    LOG_S(INFO) << "analyzing: " << item.dump(2);
+
+	    if(item.count("/Type")==1 and item["/Type"].get<std::string>()=="/Annot" and
+	       item.count("/Subtype")==1 and item["/Subtype"].get<std::string>()=="/Widget" and
+	       item.count("/Rect")==1 and
+	       item.count("/V")==1 and
+	       item.count("/T")==1 and true)
+	      {
+		std::array<double, 4> bbox = item["/Rect"].get<std::array<double, 4> >();
+		//LOG_S(INFO) << bbox[0] << ", "<< bbox[1] << ", "<< bbox[2] << ", "<< bbox[3];
+
+		std::string text = item["/V"].get<std::string>();
+		//LOG_S(INFO) << "text: " << text;
+		
+		pdf_resource<PAGE_CELL> cell;
+		{
+		  cell.widget = true;
+		  
+		  cell.x0 = bbox[0];
+		  cell.y0 = bbox[1];
+		  cell.x1 = bbox[2];
+		  cell.y1 = bbox[3];
+
+		  cell.r_x0 = bbox[0];
+		  cell.r_y0 = bbox[1];
+		  cell.r_x1 = bbox[2];
+		  cell.r_y1 = bbox[1];
+		  cell.r_x2 = bbox[2];
+		  cell.r_y2 = bbox[3];
+		  cell.r_x3 = bbox[0];
+		  cell.r_y3 = bbox[3];
+
+		  cell.text = text;
+		  cell.rendering_mode = 0;
+
+		  cell.space_width = 0;
+		  cell.chars  = {};//chars;
+		  cell.widths = {};//widths;
+		  
+		  cell.enc_name = "Form-font"; //font.get_encoding_name();
+		  
+		  cell.font_enc = "Form-font"; //to_string(font.get_encoding());
+		  cell.font_key = "Form-font"; //font.get_key();
+		  
+		  cell.font_name = "Form-font"; //font.get_name();
+		  cell.font_size = 0; //font_size/1000.0;
+		  
+		  cell.italic = false;
+		  cell.bold   = false;
+		  
+		  cell.ocr        = false;
+		  cell.confidence = -1.0;
+		  
+		  cell.stack_size  = -1;
+		  cell.block_count = -1;
+		  cell.instr_count = -1;
+		}
+
+		page_cells.push_back(cell);
+	      }
+	  }
+      }
+    
     timings[__FUNCTION__] = timer.get_time();    
   }