dumper

dumper.ml

let encodev ?(encoding=`UTF_8) xs =
  let b = Buffer.create 16 in
  let e = Uutf.encoder encoding (`Buffer b) in
  List.iter (fun x -> Uutf.encode e (`Uchar x) |> ignore) xs;
  Uutf.encode e `End |> ignore;
  Buffer.contents b

module S = struct

  open Scanf

  let rec bscans sc =
    try bscanf sc "0x%x %r" bscans (fun i is -> i::is)
    with _ ->
      try bscanf sc "U+%x %r" bscans (fun i is -> i::is)
      with _ -> []

  let scans s = sscanf s "%r" bscans (fun is -> is)
end

module P = struct

  let list ?(brackets=("","")) ?(sep=" ") p oc xs =
    let rec go = function
      | [] -> ()
      | [x] -> p oc x
      | x::xs -> p oc x ; print_string sep ; go xs in
    let (l, r) = brackets in
    output_string oc l; go xs; output_string oc r

  let u oc = Printf.fprintf oc "U+%04X"

end

module Sequence_io = struct

  let rec read_lines ic f =
    match input_line ic with
    | line -> f line; read_lines ic f
    | exception End_of_file -> ()
    | exception Unix.Unix_error (Unix.EINTR, _, _) -> read_lines ic f

end

let xf ic oc =
  let open Sequence in
  Sequence_io.read_lines ic @@ fun line ->
    let us = S.scans line in
    Printf.fprintf oc "* %a: |%s|\n%!"
      P.(list ~brackets:("[", "]") ~sep:", " u) us (encodev us)

let () = xf stdin stdout

emoji.txt

U+1F600
U+1F602

paras.txt

U+2028
U+2029
U+000A
U+000D
U+0085