Added %ExtractedText

cpdf.ml

@@ -531,19 +531,16 @@ let print_pdf_objs pdf =
        Printf.printf "%s\n" (Pdfwrite.string_of_pdf obj))
     pdf
 
-
-
 (* Return page label at pdf page num, or page number in arabic if no label *) 
 let pagelabel pdf num =
   Pdfpagelabels.pagelabeltext_of_pagenumber
     num
     (Pdfpagelabels.complete (Pdfpagelabels.read pdf))
 
-
 let rec process_text text m =
   match m with
-  | ([] : (string * string) list) -> Cpdfstrftime.strftime text
+  | [] -> Cpdfstrftime.strftime text
-  | (s, r)::t -> process_text (string_replace_all s r text) t
+  | (s, r)::t -> process_text (string_replace_all_lazy s r text) t
 
 let expand_date = function
   | "now" -> Cpdfstrftime.strftime "D:%Y%m%d%H%M%S"
@@ -1661,28 +1658,79 @@ let make_font embed fontname =
              ("/Encoding", Pdf.Name "/WinAnsiEncoding");
              ("/BaseFont", Pdf.Name ("/" ^ fontname))]
 
+let extract_page_text only_fontsize pdf _ page =
+  let text_extractor = ref None in
+  let right_font_size = ref false in
+    fold_left ( ^ ) ""
+      (map
+        (function
+         | Pdfops.Op_Tf (fontname, fontsize) ->
+             right_font_size :=
+               begin match only_fontsize with
+                 Some x -> x = fontsize
+               | _ -> false
+               end;
+             let fontdict =
+               match Pdf.lookup_direct pdf "/Font" page.Pdfpage.resources with
+               | None -> raise (Pdf.PDFError "Missing /Font in text extraction")
+               | Some d ->
+                   match Pdf.lookup_direct pdf fontname d with
+                   | None -> raise (Pdf.PDFError "Missing font in text extraction")
+                   | Some d -> d
+             in
+               text_extractor := Some (Pdftext.text_extractor_of_font pdf fontdict);
+               ""
+         | Pdfops.Op_Tj text when !text_extractor <> None ->
+             if not !right_font_size then
+               ""
+             else
+               Pdftext.utf8_of_codepoints
+                 (Pdftext.codepoints_of_text (unopt !text_extractor) text)
+         | Pdfops.Op_TJ (Pdf.Array objs) when !text_extractor <> None ->
+             if not !right_font_size then
+               ""
+             else
+               fold_left ( ^ ) ""
+                 (option_map
+                    (function
+                     | Pdf.String text ->
+                         Some
+                           (Pdftext.utf8_of_codepoints
+                             (Pdftext.codepoints_of_text (unopt !text_extractor) text))
+                     | _ -> None)
+                    objs)
+         | _ -> "")
+        (Pdfops.parse_operators pdf page.Pdfpage.resources page.Pdfpage.content))
+
+(* For each page, extract all the ops with text in them, and concatenate it all together *)
+let extract_text extract_text_font_size pdf range =
+  fold_left (fun x y -> x ^ (if x <> "" && y <> "" then "\n" else "") ^ y) ""
+    (map_pages (extract_page_text extract_text_font_size pdf) pdf range)
+
 let addtext
   metrics lines linewidth outline fast colour fontname embed bates batespad fontsize font
   underneath position hoffset voffset text pages orientation cropbox opacity
-  justification filename pdf
+  justification filename extract_text_font_size pdf
 =
   let endpage = Pdfpage.endpage pdf in
-  let replace_pairs pdf filename bates batespad num =
+  let replace_pairs pdf filename bates batespad num page =
-      ["%Page", string_of_int num;
+      ["%Page", (fun () -> string_of_int num);
-       "%Roman", roman_upper num;
+       "%Roman", (fun () -> roman_upper num);
-       "%roman", roman_lower num;
+       "%roman", (fun () -> roman_lower num);
-       "%filename", filename;
+       "%filename", (fun () -> filename);
-       "%Label", pagelabel pdf num;
+       "%Label", (fun () -> pagelabel pdf num);
-       "%EndPage", string_of_int endpage;
+       "%EndPage", (fun () -> string_of_int endpage);
-       "%EndLabel", pagelabel pdf endpage;
+       "%EndLabel", (fun () -> pagelabel pdf endpage);
+       "%ExtractedText", (fun () -> extract_page_text extract_text_font_size pdf num page);
        "%Bates",
-          (let numstring = string_of_int (bates + num - 1) in
+          (fun () ->
+            (let numstring = string_of_int (bates + num - 1) in
              match batespad with
                None -> numstring
              | Some w ->
                  if String.length numstring >= w
                    then numstring
-                   else implode (many '0' (w - String.length numstring)) ^ numstring)]
+                   else implode (many '0' (w - String.length numstring)) ^ numstring))]
   in
   let addtext_page num page =
     let resources', unique_extgstatename =
@@ -1708,7 +1756,7 @@ let addtext
       in
         let unique_fontname = Pdf.unique_key "F" fontdict in
           let ops =
-            let text = process_text text (replace_pairs pdf filename bates batespad num) in
+            let text = process_text text (replace_pairs pdf filename bates batespad num page) in
               let calc_textwidth text =
                 match font with
                 | Some f ->
@@ -1734,8 +1782,10 @@ let addtext
                         (rawwidth *. fontsize) /. 1000.
               in
                 let expanded_lines =
-                  map (function text -> process_text text (replace_pairs pdf
+                  map
-                  filename bates batespad num)) lines
+                    (function text ->
+                       process_text text (replace_pairs pdf filename bates batespad num page))
+                    lines
                 in
                 let textwidth = calc_textwidth text
                 and allwidths = map calc_textwidth expanded_lines in
@@ -1807,7 +1857,7 @@ let unescape_string s =
 let
   addtexts metrics linewidth outline fast fontname font embed bates batespad colour position linespacing
   fontsize underneath text pages orientation cropbox opacity justification
-  midline topline filename pdf
+  midline topline filename extract_text_font_size pdf
 =
   (*flprint "addtexts:\n";
   iter (Printf.printf "%C ") (explode text);
@@ -1883,6 +1933,7 @@ let
                    addtext metrics lines linewidth outline fast colour fontname
                    embed bates batespad fontsize font underneath position hoff voff line
                    pages orientation cropbox opacity justification filename
+                   extract_text_font_size
                    !pdf;
                  voffset := !voffset +. (linespacing *. fontsize))
             lines;
@@ -3664,3 +3715,4 @@ let add_page_labels pdf style prefix startval range =
         ranges;
         Pdfpagelabels.write pdf !labels
 
+

cpdf.mli

@@ -287,7 +287,7 @@ type justification =
   | CentreJustify
   | RightJustify
 
-(** [calculate ignore_d w (xmin, ymin, xmax, ymax) orientation pos] calculates
+(** [calculate_position ignore_d w (xmin, ymin, xmax, ymax) orientation pos] calculates
 the absolute position of text given its width, bounding box, orientation and
 position. If [ignore_d] is true, the distance from the position (e.g 10 in
 TopLeft 10) is ignored (considered zero). *)
@@ -324,6 +324,7 @@ val addtexts :
     bool ->(*midline adjust?*)
     bool ->(*topline adjust?*)
     string ->(*filename*)
+    float option -> (*extract_text_font_size*)
     Pdf.t ->(*pdf*)
     Pdf.t
 
@@ -492,3 +493,4 @@ val call_cpdflin : string -> string -> string -> string -> int
 
 val debug : bool ref
 
+val extract_text : float option -> Pdf.t -> int list -> string 

cpdfcommand.ml

@@ -1109,15 +1109,11 @@ let setfont f =
     end;
   args.fontname <- f
 
+let setextracttextfontsize f =
+  args.extract_text_font_size <- Some f
+
 let setfontsize f =
-  if f > 0.
+  if f > 0.  then args.fontsize <- f else error "Negative font size specified" 
-    then
-      begin
-        args.fontsize <- f;
-        args.extract_text_font_size <- Some f
-      end
-    else
-      error "Negative font size specified" 
 
 let setaddtext s =
   setop (AddText s) ()
@@ -2089,6 +2085,7 @@ and specs =
    ("-debug-crypt", Arg.Unit setdebugcrypt, "");
    ("-fix-prince", Arg.Unit (setop RemoveUnusedResources), "");
    ("-extract-text", Arg.Unit (setop ExtractText), "");
+   ("-extract-text-font-size", Arg.Float setextracttextfontsize, "");
    (*("-change-font-size-to", Arg.Float setchangefontsizeto, "");
    ("-change-font-size-shift", Arg.String setchangefontsizeshift, "");
    ("-change-font-size-color", Arg.String setchangefontsizecolor, "")*)
@@ -3039,55 +3036,6 @@ let remove_unused_resources_page pdf n page =
 let remove_unused_resources pdf =
   Cpdf.process_pages (remove_unused_resources_page pdf) pdf (ilist 1 (Pdfpage.endpage pdf))
 
-let extract_page_text only_fontsize pdf _ page =
-  let text_extractor = ref None in
-  let right_font_size = ref false in
-    fold_left ( ^ ) ""
-      (map
-        (function
-         | Pdfops.Op_Tf (fontname, fontsize) ->
-             right_font_size :=
-               begin match only_fontsize with
-                 Some x -> x = fontsize
-               | _ -> false
-               end;
-             let fontdict =
-               match Pdf.lookup_direct pdf "/Font" page.Pdfpage.resources with
-               | None -> raise (Pdf.PDFError "Missing /Font in text extraction")
-               | Some d ->
-                   match Pdf.lookup_direct pdf fontname d with
-                   | None -> raise (Pdf.PDFError "Missing font in text extraction")
-                   | Some d -> d
-             in
-               text_extractor := Some (Pdftext.text_extractor_of_font pdf fontdict);
-               ""
-         | Pdfops.Op_Tj text when !text_extractor <> None ->
-             if not !right_font_size then
-               ""
-             else
-               Pdftext.utf8_of_codepoints
-                 (Pdftext.codepoints_of_text (unopt !text_extractor) text)
-         | Pdfops.Op_TJ (Pdf.Array objs) when !text_extractor <> None ->
-             if not !right_font_size then
-               ""
-             else
-               fold_left ( ^ ) ""
-                 (option_map
-                    (function
-                     | Pdf.String text ->
-                         Some
-                           (Pdftext.utf8_of_codepoints
-                             (Pdftext.codepoints_of_text (unopt !text_extractor) text))
-                     | _ -> None)
-                    objs)
-         | _ -> "")
-        (Pdfops.parse_operators pdf page.Pdfpage.resources page.Pdfpage.content))
-
-(* For each page, extract all the ops with text in them, and concatenate it all together *)
-let extract_text pdf range =
-  fold_left (fun x y -> x ^ (if x <> "" && y <> "" then "\n" else "") ^ y) ""
-    (Cpdf.map_pages (extract_page_text args.extract_text_font_size pdf) pdf range)
-
 (* Extracts font to font.dat in CWD. *)
 let extract_fontfile pagenumber fontname pdf =
   let resources = (select pagenumber (Pdfpage.pages_of_pagetree pdf)).Pdfpage.resources in
@@ -3831,7 +3779,8 @@ let go () =
                    font args.embedfonts args.bates args.batespad args.color args.position
                    args.linespacing args.fontsize args.underneath text range
                    args.orientation args.relative_to_cropbox args.opacity
-                   args.justification args.midline args.topline filename pdf)
+                   args.justification args.midline args.topline filename
+                   args.extract_text_font_size pdf)
   | Some RemoveText ->
       let pdf = get_single_pdf args.op false in
         let range = parse_pagespec pdf (get_pagespec ()) in
@@ -3951,7 +3900,7 @@ let go () =
   | Some ExtractText ->
       let pdf = get_single_pdf args.op true in
         let range = parse_pagespec pdf (get_pagespec ()) in
-          let text = extract_text pdf range in
+          let text = Cpdf.extract_text args.extract_text_font_size pdf range in
             begin match args.out with
             | File filename ->
                 let fh = open_out_bin filename in