cpdf-source/cpdfremovetext.ml

open Pdfutil

let removetext range pdf =
  (* Could fail on nesting, or other marked content inside our marked content.*)
  let rec remove_until_last_EMC level = function
    | [] -> []
    | Pdfops.Op_BMC "/CPDFSTAMP"::more ->
        remove_until_last_EMC (level + 1) more
    | Pdfops.Op_EMC::more ->
        if level = 1
          then more
          else remove_until_last_EMC (level - 1) more
    | _::more ->
        remove_until_last_EMC level more
  in
    let rec remove_stamps prev = function
      | [] -> rev prev
      | Pdfops.Op_BMC "/CPDFSTAMP"::more ->
          let rest = remove_until_last_EMC 1 more in
            remove_stamps prev rest
      | h::t -> remove_stamps (h::prev) t
    in
      let removetext_page _ page =
        {page with
           Pdfpage.content =
             let ops = Pdfops.parse_operators pdf page.Pdfpage.resources page.Pdfpage.content in
               [Pdfops.stream_of_ops (remove_stamps [] ops)]}
      in
        Cpdfpage.process_pages (Pdfpage.ppstub removetext_page) pdf range

let rec remove_all_text_ops pdf resources content =
  let is_textop = function
    Pdfops.Op_Tj _ | Pdfops.Op_' _ | Pdfops.Op_'' _ | Pdfops.Op_TJ _ -> true
  | _ -> false
  in
    let content' =
      let ops = Pdfops.parse_operators pdf resources content in
        Pdfops.stream_of_ops
          (option_map (function x -> if is_textop x then None else Some x) ops) 
    in
      [content']

let remove_all_text_page pdf p =
  let resources = p.Pdfpage.resources in
  let content = p.Pdfpage.content in
    Pdfpage.process_xobjects pdf p remove_all_text_ops;
    {p with Pdfpage.content = remove_all_text_ops pdf resources content}, pdf

let remove_all_text range pdf =
  let pages = Pdfpage.pages_of_pagetree pdf in
    let pagenums = indx pages in
    let pdf = ref pdf in
    let pages' = ref [] in
      iter2 
        (fun p pagenum ->
          let p', pdf' =
            if mem pagenum range
              then remove_all_text_page !pdf p
              else p, !pdf
          in
            pdf := pdf';
            pages' =| p')
        pages
        pagenums;
      Pdfpage.change_pages true !pdf (rev !pages')