docling-parse 2.0.2__tar.gz → 2.0.3__tar.gz

{docling_parse-2.0.2 → docling_parse-2.0.3}/CHANGELOG.md

@@ -1,3 +1,9 @@
+## [v2.0.3](https://github.com/DS4SD/docling-parse/releases/tag/v2.0.3) - 2024-11-05
+
+### Fix
+
+* Replace all the FATAL with ERROR messages in the v2 parser ([#53](https://github.com/DS4SD/docling-parse/issues/53)) ([`cd15d00`](https://github.com/DS4SD/docling-parse/commit/cd15d00ddb6c67fada0056ec068caffc003d5edc))
+
 ## [v2.0.2](https://github.com/DS4SD/docling-parse/releases/tag/v2.0.2) - 2024-10-30
 
 ### Fix

{docling_parse-2.0.2 → docling_parse-2.0.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.1
 Name: docling-parse
-Version: 2.0.2
+Version: 2.0.3
 Summary: Simple package to extract text with coordinates from programmatic PDFs
 Home-page: https://github.com/DS4SD/docling-parse
 License: MIT

{docling_parse-2.0.2 → docling_parse-2.0.3}/docling_parse/visualize.py

@@ -53,6 +53,13 @@ def parse_args():
         help="Enable interactive mode (default: False)",
     )
 
+    # Add an optional boolean argument for interactive mode
+    parser.add_argument(
+        "--display-text",
+        action="store_true",
+        help="Enable interactive mode (default: False)",
+    )
+
     # Add an argument for the output directory, defaulting to "./tmp"
     parser.add_argument(
         "-o",
@@ -91,11 +98,17 @@ def parse_args():
         args.interactive,
         args.output_dir,
         int(args.page),
+        args.display_text,
     )
 
 
 def visualise_v1(
-    log_level: str, pdf_path: str, interactive: str, output_dir: str, page_num: int
+    log_level: str,
+    pdf_path: str,
+    interactive: str,
+    output_dir: str,
+    page_num: int,
+    display_text: bool,
 ):
 
     parser = pdf_parser_v1()
@@ -200,7 +213,12 @@ def visualise_v1(
 
 
 def visualise_v2(
-    log_level: str, pdf_path: str, interactive: str, output_dir: str, page_num: int
+    log_level: str,
+    pdf_path: str,
+    interactive: str,
+    output_dir: str,
+    page_num: int,
+    display_text: bool,
 ):
 
     parser = pdf_parser_v2(log_level)
@@ -214,10 +232,17 @@ def visualise_v2(
 
     doc = None
 
-    if page_num == -1:
-        doc = parser.parse_pdf_from_key(doc_key)
-    else:
-        doc = parser.parse_pdf_from_key_on_page(doc_key, page_num)
+    try:
+        if page_num == -1:
+            doc = parser.parse_pdf_from_key(doc_key)
+        else:
+            doc = parser.parse_pdf_from_key_on_page(doc_key, page_num)
+    except Exception as exc:
+        print(f"Could not parse pdf-document: {exc}")
+        doc = None
+
+    if doc == None:
+        return
 
     parser.unload_document(doc_key)
 
@@ -295,6 +320,9 @@ def visualise_v2(
                         (x[3], H - y[3]),
                     ]
 
+                    if display_text:
+                        print(row[cells_header.index("text")])
+
                     if "glyph" in row[cells_header.index("text")]:
                         print(f" skip cell -> {row}")
                         continue
@@ -328,12 +356,12 @@ def visualise_v2(
 
 def main():
 
-    log_level, version, pdf, interactive, output_dir, page = parse_args()
+    log_level, version, pdf, interactive, output_dir, page, display_text = parse_args()
 
     if version == "v1":
-        visualise_v1(log_level, pdf, interactive, output_dir, page)
+        visualise_v1(log_level, pdf, interactive, output_dir, page, display_text)
     elif version == "v2":
-        visualise_v2(log_level, pdf, interactive, output_dir, page)
+        visualise_v2(log_level, pdf, interactive, output_dir, page, display_text)
     else:
         return -1
 

{docling_parse-2.0.2 → docling_parse-2.0.3}/pyproject.toml

@@ -1,6 +1,6 @@
 [tool.poetry]
 name = "docling-parse"
-version = "2.0.2"
+version = "2.0.3"
 description = "Simple package to extract text with coordinates from programmatic PDFs"
 authors = ["Peter Staar <taa@zurich.ibm.com>"]
 maintainers = [

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/enums.h

@@ -36,7 +36,7 @@ namespace pdflib
     else if (name=="CID_FONT_TYPE_2" or name=="/CIDFontType2") { return CID_FONT_TYPE_2; }
     else
       {
-        LOG_S(FATAL) << "unknown subtype " << name;
+        LOG_S(ERROR) << "unknown subtype " << name;
         return NULL_TYPE; 
       }
   }
@@ -59,7 +59,7 @@ namespace pdflib
 
       default:
         {
-          LOG_S(FATAL) << "encountered a NULL_ENCODING";
+          LOG_S(ERROR) << "encountered a NULL_ENCODING";
           return "NULL_ENCODING";
         }
       }
@@ -90,7 +90,7 @@ namespace pdflib
     else if(name=="CMAP_RESOURCES"                          ) { return CMAP_RESOURCES; }
     else 
       {
-        LOG_S(FATAL) << __FILE__ << ":" << __LINE__ << " --> unknown encoding " << name;
+        LOG_S(ERROR) << __FILE__ << ":" << __LINE__ << " --> unknown encoding " << name;
         return NULL_ENCODING; 
       }
   }
@@ -109,7 +109,7 @@ namespace pdflib
 
       default:
         {
-          LOG_S(FATAL) << "encountered a NULL_ENCODING";
+          LOG_S(ERROR) << "encountered a NULL_ENCODING";
           return "NULL_ENCODING";
         }
       }

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_decoders/stream.h

@@ -253,7 +253,11 @@ namespace pdflib
   {
     if(stack.size()==0)
       {
-        LOG_S(FATAL) << "stack-size is zero!";
+	std::stringstream message;
+	message << "stack-size is zero in " << __FILE__ << ":" << __LINE__;
+
+	LOG_S(ERROR) << message.str();
+	throw std::logic_error(message.str());
       }
 
     pdf_state<GLOBAL>& state = stack.back();

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_dimension.h

@@ -110,7 +110,7 @@ namespace pdflib
       }
     else
       {
-        LOG_S(FATAL) << "The page is missing the required '/MediaBox'";
+        LOG_S(ERROR) << "The page is missing the required '/MediaBox'";
       }
 
     if(json_resources.count("/CropBox"))
@@ -173,8 +173,12 @@ namespace pdflib
       }
     else
       {
-        LOG_S(FATAL) << "could not find the page-dimensions: " 
-                     << json_resources.dump(4);
+	std::stringstream ss;
+	ss << "could not find the page-dimensions: " 
+	   << json_resources.dump(4);
+	
+        LOG_S(ERROR) << ss.str();
+	throw std::logic_error(ss.str());
       }
   }
 

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_font/base_font.h

@@ -157,9 +157,15 @@ namespace pdflib
 	return bbox[3];
       }
 
-    LOG_S(FATAL) << "properties does not have key 'Ascender': " 
-                 << properties.dump(2);
-
+    {
+      std::stringstream ss;
+      ss << "properties does not have key 'Ascender': " 
+	 << properties.dump(2);
+      
+      LOG_S(ERROR) << ss.str();
+      throw std::logic_error(ss.str());
+    }
+    
     return -1.;
   }
 
@@ -179,9 +185,15 @@ namespace pdflib
 	return bbox[1];
       }
 
-    LOG_S(FATAL) << "properties does not have key 'Descender': " 
-                 << properties.dump(2);
+    {
+      std::stringstream ss;
+      ss << "properties does not have key 'Descender': " 
+	 << properties.dump(2);
 
+      LOG_S(ERROR) << ss.str();
+      throw std::logic_error(ss.str());
+    }
+    
     return -1.;
   }
 
@@ -194,9 +206,15 @@ namespace pdflib
         return properties["FontBBox"].get<std::array<double, 4> >();
       }
 
-    LOG_S(FATAL) << "properties does not have key 'FontBBox': " 
-                 << properties.dump(2);
-
+    {
+      std::stringstream ss;
+      ss << "properties does not have key 'FontBBox': " 
+	 << properties.dump(2);
+      
+      LOG_S(ERROR) << ss.str();
+      throw std::logic_error(ss.str());
+    }
+    
     return {0.0, 0.0, 0.0, 0.0};
   }
 

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_font/base_fonts.h

@@ -121,8 +121,6 @@ namespace pdflib
       {
         if(norm_name.find(itr->first)!=std::string::npos)
           {	    
-            //return itr->first;
-
 	    // we have to be careful that "Helvetica" is not returned for Helvetice-Bold!
 	    if(result.size()<(itr->first).size())
 	      {
@@ -136,8 +134,8 @@ namespace pdflib
 	return result;
       }
 
-    LOG_S(FATAL) << "unkown " << font_name << "[norm_name=" << norm_name << "]";
-
+    LOG_S(ERROR) << "unkown " << font_name << "[norm_name=" << norm_name << "]";
+    
     return "Unknown";
   }
 
@@ -269,7 +267,11 @@ namespace pdflib
 
     if(fontname=="unknown")
       {
-        LOG_S(FATAL) << "no FontName found in " << filename;
+	std::stringstream ss;
+	ss << "no FontName found in " << filename;
+	
+        LOG_S(ERROR) << ss.str();
+	throw std::logic_error(ss.str());
       }
 
     return fontname;

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_font/cmap.h

@@ -434,7 +434,7 @@ namespace pdflib
 
                   if(_map.count(begin+i)==1)
                     {
-                      LOG_S(FATAL) << "overwriting number c=" << begin+i;
+                      LOG_S(WARNING) << "overwriting number c=" << begin+i;
                     }
 
                   _map[begin + i] = tmp;

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_font/font_cid.h

@@ -87,7 +87,8 @@ namespace pdflib
     if(file.fail())
       {
 	LOG_S(ERROR) << "filename does not exists: " << filename;	
-	LOG_S(FATAL) << "unknown data-file!";
+
+	
       }
 
     bool        cmap=false;
@@ -143,8 +144,11 @@ namespace pdflib
 
     if(file.fail())
       {
-	LOG_S(ERROR) << "filename does not exists: " << filename;	
-	LOG_S(FATAL) << "unknown data-file!";
+	std::stringstream ss;
+	ss << "filename does not exists: " << filename;
+
+	LOG_S(ERROR) << ss.str();
+	throw std::logic_error(ss.str());
       }
 
     std::vector<int> col_inds = {};
@@ -246,7 +250,7 @@ namespace pdflib
           }
         else
           {
-            LOG_S(FATAL) << "we should never arrive here!";
+            LOG_S(ERROR) << "all options exhausted for " << __FUNCTION__;
           }
       }
 

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_font/glyphs.h

@@ -152,8 +152,11 @@ namespace pdflib
 
     if(file.fail())
       {
-	LOG_S(ERROR) << "filename does not exists: " << filename;	
-	LOG_S(FATAL) << "unknown data-file!";
+	std::stringstream ss;
+	ss << "filename does not exists: " << filename;	
+	
+	LOG_S(ERROR) << ss.str();
+	throw std::logic_error(ss.str());
       }
 
     std::string line;
@@ -218,8 +221,11 @@ namespace pdflib
 
     if(file.fail())
       {
-	LOG_S(ERROR) << "filename does not exists: " << filename;	
-	LOG_S(FATAL) << "unknown data-file!";
+	std::stringstream ss;
+	ss << "filename does not exists: " << filename;	
+	
+	LOG_S(ERROR) << ss.str();
+	throw std::logic_error(ss.str());
       }
     
     std::string line;

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_font.h

@@ -180,8 +180,9 @@ namespace pdflib
       }
     else
       {
-	LOG_S(FATAL) << "no existing pdf_resources_dir: " 
-		     << pdf_resources_dir;
+	std::string message = "no existing pdf_resources_dir: " +  pdf_resources_dir; 
+	LOG_S(ERROR) << message;
+	throw std::logic_error(message);
       }
 
     utils::timer timer;
@@ -1148,7 +1149,11 @@ namespace pdflib
           }
         else
           {
-            LOG_S(FATAL) << "unknown type in " << __FUNCTION__;
+	    std::stringstream message;
+	    message <<  "unknown type in " << __FUNCTION__;	    
+
+	    LOG_S(ERROR) << message.str();
+	    throw std::logic_error(message.str());
           }
       }
   }
@@ -1166,7 +1171,12 @@ namespace pdflib
         if(not qpdf_font.hasKey("/ToUnicode"))
           {
             auto tmp = to_json(qpdf_font);
-            LOG_S(FATAL) << "qpdf-font: " << tmp.dump();
+
+	    std::stringstream ss;
+	    ss << "qpdf-font: " << tmp.dump();
+	    
+            LOG_S(ERROR) << ss.str();
+	    throw std::logic_error(ss.str());
           }
 
         auto qpdf_obj = qpdf_font.getKey("/ToUnicode");

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_fonts.h

@@ -77,15 +77,25 @@ namespace pdflib
   {
     if(page_fonts.count(font_name)==1)
       {
-        return page_fonts[font_name];
+        return page_fonts.at(font_name);
       }
     else
       {
         std::stringstream ss;
+	ss << "font_name [" << font_name << "] is not known: ";
         for(auto itr=page_fonts.begin(); itr!=page_fonts.end(); itr++)
-          ss << itr->first << ", ";
-
-        LOG_S(FATAL) << "font_name [" << font_name << "] is not known: " << ss.str();
+	  {
+	    if(itr==page_fonts.begin())
+	      {
+		ss << itr->first;
+	      }
+	    else
+	      {
+		ss << ", " << itr->first; 
+	      }
+	  }
+
+	throw std::logic_error(ss.str());
       }
 
     return (page_fonts.begin()->second);
@@ -102,17 +112,24 @@ namespace pdflib
         nlohmann::json& val = pair.value();
 
         LOG_S(INFO) << "decoding font: " << key;// << "\n" << val.dump(2);
-        assert(qpdf_fonts.hasKey(key));
-
-        pdf_resource<PAGE_FONT> page_font;
-        page_font.set(key, val, qpdf_fonts.getKey(key));
-
-        if(page_fonts.count(key)==1)
-          {
-            LOG_S(FATAL) << "We are overwriting a font!! BE CAREFUL!";
-          }
-
-        page_fonts[key] = page_font;
+        //assert(qpdf_fonts.hasKey(key));
+
+	if(qpdf_fonts.hasKey(key))
+	  {
+	    pdf_resource<PAGE_FONT> page_font;
+	    page_font.set(key, val, qpdf_fonts.getKey(key));
+	    
+	    if(page_fonts.count(key)==1)
+	      {
+		LOG_S(ERROR) << "We are overwriting a font!";
+	      }
+	    
+	    page_fonts[key] = page_font;
+	  }
+	else
+	  {
+	    LOG_S(ERROR) << "qpdf does not have key: " << key;
+	  }
       }
   }
 

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_grphs.h

@@ -83,9 +83,17 @@ namespace pdflib
       {
         std::stringstream ss;
         for(auto itr=page_grphs.begin(); itr!=page_grphs.end(); itr++)
-          ss << itr->first << ", ";
-
-        LOG_S(FATAL) << "graphics state with name '" << grph_name << "' is not known: " << ss.str();
+	  {
+	    ss << itr->first << ", ";
+	  }
+
+	{
+	  std::stringstream ss;
+	  ss << "graphics state with name '" << grph_name << "' is not known: " << ss.str();
+	  
+	  LOG_S(ERROR) << ss.str();
+	  throw std::logic_error(ss.str());
+	}
       }
 
     return (page_grphs.begin()->second);
@@ -109,7 +117,11 @@ namespace pdflib
 
         if(page_grphs.count(key)==1)
           {
-            LOG_S(FATAL) << "We are overwriting a grph!! BE CAREFUL!";
+	    std::stringstream ss;
+	    ss << "we are overwriting a grph!";
+
+	    LOG_S(ERROR) << ss.str();
+	    //throw std::logic_error(ss.str());
           }
 
         page_grphs[key] = page_grph;

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/pdf_resources/page_xobject.h

@@ -196,7 +196,11 @@ namespace pdflib
       }
     catch(const std::exception& exc)
       {
-        LOG_S(FATAL) << "encountered an error: " << exc.what();
+	std::stringstream ss;
+	ss << "encountered an error: " << exc.what();
+
+	LOG_S(ERROR) << ss.str();
+	throw std::logic_error(ss.str());
       }
 
     return stream;

{docling_parse-2.0.2 → docling_parse-2.0.3}/src/v2/qpdf/instruction.h

@@ -86,9 +86,13 @@ namespace pdflib
         return obj.getStringValue(); 
       }
     else
-      {        
-        LOG_S(FATAL) << "can not decode a string value for key: " << key 
-                     << " and value: " << val; 
+      {
+	std::stringstream ss;
+        ss << "can not decode a string value for key: " << key 
+	   << " and value: " << val;
+
+	LOG_S(ERROR) << ss.str();
+	throw std::logic_error(ss.str());
       }
 
     return "null";    
@@ -105,9 +109,13 @@ namespace pdflib
         return obj.getUTF8Value(); 
       }
     else
-      {        
-        LOG_S(FATAL) << "can not decode a string value for key: " << key 
-                     << " and value: " << val; 
+      {
+	std::stringstream ss;
+        ss << "can not decode a string value for key: " << key 
+	   << " and value: " << val;
+
+	LOG_S(ERROR) << ss.str();
+	throw std::logic_error(ss.str());
       }
 
     return "null";