temporary Solutions for annots attachments

1. CosObjects are exposed from PDDoc and PDPage
2. xref for overriden entries are ignored.
3. helper enahnced to extract FileAttachment annots.

Author: sambitdash

src/CosDoc.jl

@@ -275,7 +275,9 @@ function read_xref_table(ps::BufferedInputStream, doc::CosDocImpl)
             if (v[18] != LATIN_F)
                 ref = CosIndirectObjectRef(oid, parse(Int,String(v[12:16])))
 
-                doc.xref[ref] = CosObjectLoc(parse(Int,String(v[1:10])))
+                if !haskey(doc.xref,ref)
+                    doc.xref[ref] = CosObjectLoc(parse(Int,String(v[1:10])))
+                end
             end
 
             oid +=1

src/CosObjStream.jl

@@ -80,7 +80,9 @@ function read_xref_stream(xrefstm::CosObject,
       it += recsize
       if (record[1] != 0)
         count_record +=1
-        xref[ref]=CosObjectLoc(loc,stm)
+        if !haskey(xref,ref)
+          xref[ref]=CosObjectLoc(loc,stm)
+        end
       end
     end
   end

src/PDDoc.jl

@@ -1,6 +1,8 @@
 export PDDoc,
        pdDocOpen,
        pdDocClose,
+       pdDocGetCatalog,
+       pdDocGetCosDoc,
        pdDocGetPageCount,
        pdDocGetPage
 
@@ -9,6 +11,7 @@ export PDDoc,
 function pdDocOpen(fp::String)
   doc = PDDocImpl(fp)
   update_page_tree(doc)
+  update_structure_tree(doc)
   return doc
 end
 
@@ -24,6 +27,8 @@ function pdDocGetCatalog(doc::PDDoc)
   return doc.catalog
 end
 
+pdDocGetCosDoc(doc::PDDoc)= doc.cosDoc
+
 function pdDocGetPage(doc::PDDoc, num::Int)
   cosobj = find_page_from_treenode(doc.pages, num)
   return create_pdpage(doc, cosobj)
@@ -38,10 +43,11 @@ end
   cosDoc::CosDoc
   catalog::CosObject
   pages::CosObject
+  isTagged::Symbol #Valid values :tagged, :none and :suspect
   function PDDocImpl(fp::String)
     cosDoc = cosDocOpen(fp)
     catalog = cosDocGetRoot(cosDoc)
-    new(cosDoc,catalog,CosNull)
+    new(cosDoc,catalog,CosNull,:none)
   end
 end
 
@@ -118,3 +124,15 @@ function find_page_from_treenode(node::CosObject, pageno::Int)
     end
   end
 end
+
+function update_structure_tree(doc::PDDocImpl)
+  catalog = pdDocGetCatalog(doc)
+  marking = get(catalog, CosName("MarkInfo"))
+
+  if (marking !== CosNull)
+    tagged = get(marking, CosName("Marked"))
+    suspect = get(marking, CosName("Suspect"))
+    doc.isTagged = (suspect === CosTrue) ? (:suspect) :
+                   (tagged  === CosTrue) ? (:tagged)  : (:none)
+  end
+end

src/PDPage.jl

@@ -1,6 +1,7 @@
 export PDPage,
        pdPageGetContents,
        pdPageIsEmpty,
+       pdPageGetCosObject,
        pdPageGetContentObjects
 
 @compat abstract type PDPage end
@@ -14,6 +15,8 @@ export PDPage,
     new(doc, cospage, contents, Nullable{PDPageObjectGroup}())
 end
 
+pdPageGetCosObject(page)=page.cospage
+
 function pdPageGetContents(page::PDPageImpl)
   if (page.contents === CosNull)
     ref = get_page_content_ref(page)

test/helpers.jl

@@ -78,3 +78,45 @@ function pdfhlp_extract_doc_embedded_files(filename,dir=tempdir())
     pdDocClose(doc)
   end
 end
+
+function pdfhlp_extract_doc_attachment_files(filename,dir=tempdir())
+  file=rsplit(filename, '/',limit=2)
+  filenm=file[end]
+  dirpath=joinpath(dir,filenm)
+  if isdir(dirpath)
+    rm(dirpath; force=true, recursive=true)
+  end
+  mkdir(dirpath)
+  doc=pdDocOpen(filename)
+  cosDoc=pdDocGetCosDoc(doc)
+  try
+    npage= pdDocGetPageCount(doc)
+    for i=1:npage
+      page = pdDocGetPage(doc, i)
+      cospage = pdPageGetCosObject(page)
+      annots=cosDocGetObject(cosDoc, get(cospage, CosName("Annots")))
+      if (annots === CosNull)
+        continue
+      end
+      annotsarr=get(annots)
+      for annot in annotsarr
+        annotdict = cosDocGetObject(cosDoc, annot)
+        subtype = get(annotdict,CosName("Subtype"))
+        if (subtype == CosName("FileAttachment"))
+          filespec=cosDocGetObject(cosDoc, get(annotdict,CosName("FS")))
+          ef=get(filespec, CosName("EF"))
+          filename=get(filespec,CosName("F")) #UF could be there as well.
+          stmref=get(ef, CosName("F"))
+          stm=cosDocGetObject(cosDoc,stmref)
+          bufstm=decode(stm)
+          buf=read(bufstm)
+          close(bufstm)
+          path=joinpath(dirpath,get(filename))
+          write(path,buf)
+        end
+      end
+    end
+  finally
+    pdDocClose(doc)
+  end
+end