https://gitlab.com/nomadic-labs/data-encoding/
Raw File
Tip revision: 875b189e841bcb29e6e54d572d0ae8e829e9ea15 authored by Raphaƫl Proust on 12 February 2020, 21:24 UTC
Update CHANGES.md for release
Tip revision: 875b189
binary_writer.ml
(*****************************************************************************)
(*                                                                           *)
(* Open Source License                                                       *)
(* Copyright (c) 2018 Dynamic Ledger Solutions, Inc. <contact@tezos.com>     *)
(*                                                                           *)
(* Permission is hereby granted, free of charge, to any person obtaining a   *)
(* copy of this software and associated documentation files (the "Software"),*)
(* to deal in the Software without restriction, including without limitation *)
(* the rights to use, copy, modify, merge, publish, distribute, sublicense,  *)
(* and/or sell copies of the Software, and to permit persons to whom the     *)
(* Software is furnished to do so, subject to the following conditions:      *)
(*                                                                           *)
(* The above copyright notice and this permission notice shall be included   *)
(* in all copies or substantial portions of the Software.                    *)
(*                                                                           *)
(* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR*)
(* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,  *)
(* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL   *)
(* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER*)
(* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING   *)
(* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER       *)
(* DEALINGS IN THE SOFTWARE.                                                 *)
(*                                                                           *)
(*****************************************************************************)

open Binary_error

let raise error = raise (Write_error error)

(** Imperative state of the binary writer. *)
type state = {
  mutable buffer : Bytes.t;  (** The buffer where to write. *)
  mutable offset : int;
      (** The offset of the next byte to be written in [buffer]. *)
  mutable allowed_bytes : int option;
      (** Maximum number of bytes that are allowed to be write in [buffer]
      (after [offset]) before to fail (None = illimited). *)
}

let check_allowed_bytes state size =
  match state.allowed_bytes with
  | Some len when len < size ->
      raise Size_limit_exceeded
  | Some len ->
      state.allowed_bytes <- Some (len - size)
  | None ->
      ()

(** [may_resize state size] will first ensure there is enough
    space in [state.buffer] for writing [size] bytes (starting at
    [state.offset]).

    When the buffer does not have enough space for writing [size] bytes,
    but still has enough [allowed_bytes], it will replace the buffer
    with a buffer large enough.

    @raise [Binary_error.Write_error Size_limit_exceeded] when there is
           not enough allowed bytes to write [size] bytes. *)
let may_resize state size =
  check_allowed_bytes state size ;
  let buffer_len = Bytes.length state.buffer in
  if buffer_len - state.offset < size then (
    let new_buffer = Bytes.create (max (2 * buffer_len) (buffer_len + size)) in
    Bytes.blit state.buffer 0 new_buffer 0 state.offset ;
    state.buffer <- new_buffer ) ;
  state.offset <- state.offset + size

(** Writer for all the atomic types. *)
module Atom = struct
  let check_int_range min v max =
    if v < min || max < v then raise (Invalid_int {min; v; max})

  let check_float_range min v max =
    if v < min || max < v then raise (Invalid_float {min; v; max})

  let set_int kind buffer ofs v =
    match kind with
    | `Int31 | `Uint30 ->
        TzEndian.set_int32 buffer ofs (Int32.of_int v)
    | `Int16 | `Uint16 ->
        TzEndian.set_int16 buffer ofs v
    | `Int8 | `Uint8 ->
        TzEndian.set_int8 buffer ofs v

  let int kind state v =
    check_int_range (Binary_size.min_int kind) v (Binary_size.max_int kind) ;
    let ofs = state.offset in
    may_resize state (Binary_size.integer_to_size kind) ;
    set_int kind state.buffer ofs v

  let int8 = int `Int8

  let uint8 = int `Uint8

  let int16 = int `Int16

  let uint16 = int `Uint16

  let uint30 = int `Uint30

  let int31 = int `Int31

  let bool state v = uint8 state (if v then 255 else 0)

  let int32 state v =
    let ofs = state.offset in
    may_resize state Binary_size.int32 ;
    TzEndian.set_int32 state.buffer ofs v

  let int64 state v =
    let ofs = state.offset in
    may_resize state Binary_size.int64 ;
    TzEndian.set_int64 state.buffer ofs v

  let ranged_int ~minimum ~maximum state v =
    check_int_range minimum v maximum ;
    let v = if minimum >= 0 then v - minimum else v in
    match Binary_size.range_to_size ~minimum ~maximum with
    | `Uint8 ->
        uint8 state v
    | `Uint16 ->
        uint16 state v
    | `Uint30 ->
        uint30 state v
    | `Int8 ->
        int8 state v
    | `Int16 ->
        int16 state v
    | `Int31 ->
        int31 state v

  let n state v =
    if Z.sign v < 0 then raise Invalid_natural ;
    if Z.equal v Z.zero then uint8 state 0x00
    else
      let bits = Z.numbits v in
      let get_chunk pos len = Z.to_int (Z.extract v pos len) in
      let length = Binary_length.n_length v in
      let offset = state.offset in
      may_resize state length ;
      for i = 0 to length - 1 do
        let pos = i * 7 in
        let chunk_len = if i = length - 1 then bits - pos else 7 in
        TzEndian.set_int8
          state.buffer
          (offset + i)
          ((if i = length - 1 then 0x00 else 0x80) lor get_chunk pos chunk_len)
      done

  let z state v =
    let sign = Z.sign v < 0 in
    let bits = Z.numbits v in
    if Z.equal v Z.zero then uint8 state 0x00
    else
      let v = Z.abs v in
      let get_chunk pos len = Z.to_int (Z.extract v pos len) in
      let length = Binary_length.z_length v in
      let offset = state.offset in
      may_resize state length ;
      TzEndian.set_int8
        state.buffer
        offset
        ( (if sign then 0x40 else 0x00)
        lor (if bits > 6 then 0x80 else 0x00)
        lor get_chunk 0 6 ) ;
      for i = 1 to length - 1 do
        let pos = 6 + ((i - 1) * 7) in
        let chunk_len = if i = length - 1 then bits - pos else 7 in
        TzEndian.set_int8
          state.buffer
          (offset + i)
          ((if i = length - 1 then 0x00 else 0x80) lor get_chunk pos chunk_len)
      done

  let float state v =
    let ofs = state.offset in
    may_resize state Binary_size.float ;
    TzEndian.set_double state.buffer ofs v

  let ranged_float ~minimum ~maximum state v =
    check_float_range minimum v maximum ;
    float state v

  let string_enum tbl arr state v =
    let value =
      try snd (Hashtbl.find tbl v) with Not_found -> raise No_case_matched
    in
    match Binary_size.enum_size arr with
    | `Uint30 ->
        uint30 state value
    | `Uint16 ->
        uint16 state value
    | `Uint8 ->
        uint8 state value

  let fixed_kind_bytes length state s =
    if Bytes.length s <> length then
      raise (Invalid_bytes_length {expected = length; found = Bytes.length s}) ;
    let ofs = state.offset in
    may_resize state length ;
    Bytes.blit s 0 state.buffer ofs length

  let fixed_kind_string length state s =
    if String.length s <> length then
      raise
        (Invalid_string_length {expected = length; found = String.length s}) ;
    let ofs = state.offset in
    may_resize state length ;
    Bytes.blit_string s 0 state.buffer ofs length

  let tag = function `Uint8 -> uint8 | `Uint16 -> uint16
end

(** Main recursive writing function. *)
let rec write_rec : type a. a Encoding.t -> state -> a -> unit =
 fun e state value ->
  let open Encoding in
  match e.encoding with
  | Null ->
      ()
  | Empty ->
      ()
  | Constant _ ->
      ()
  | Ignore ->
      ()
  | Bool ->
      Atom.bool state value
  | Int8 ->
      Atom.int8 state value
  | Uint8 ->
      Atom.uint8 state value
  | Int16 ->
      Atom.int16 state value
  | Uint16 ->
      Atom.uint16 state value
  | Int31 ->
      Atom.int31 state value
  | Int32 ->
      Atom.int32 state value
  | Int64 ->
      Atom.int64 state value
  | N ->
      Atom.n state value
  | Z ->
      Atom.z state value
  | Float ->
      Atom.float state value
  | Bytes (`Fixed n) ->
      Atom.fixed_kind_bytes n state value
  | Bytes `Variable ->
      let length = Bytes.length value in
      Atom.fixed_kind_bytes length state value
  | String (`Fixed n) ->
      Atom.fixed_kind_string n state value
  | String `Variable ->
      let length = String.length value in
      Atom.fixed_kind_string length state value
  | Padded (e, n) ->
      write_rec e state value ;
      Atom.fixed_kind_string n state (String.make n '\000')
  | RangedInt {minimum; maximum} ->
      Atom.ranged_int ~minimum ~maximum state value
  | RangedFloat {minimum; maximum} ->
      Atom.ranged_float ~minimum ~maximum state value
  | String_enum (tbl, arr) ->
      Atom.string_enum tbl arr state value
  | Array (Some max_length, _e) when Array.length value > max_length ->
      raise Array_too_long
  | Array (_, e) ->
      Array.iter (write_rec e state) value
  | List (Some max_length, _e) when List.length value > max_length ->
      raise List_too_long
  | List (_, e) ->
      List.iter (write_rec e state) value
  | Obj (Req {encoding = e; _}) ->
      write_rec e state value
  | Obj (Opt {kind = `Dynamic; encoding = e; _}) -> (
    match value with
    | None ->
        Atom.bool state false
    | Some value ->
        Atom.bool state true ; write_rec e state value )
  | Obj (Opt {kind = `Variable; encoding = e; _}) -> (
    match value with None -> () | Some value -> write_rec e state value )
  | Obj (Dft {encoding = e; _}) ->
      write_rec e state value
  | Objs {left; right; _} ->
      let (v1, v2) = value in
      write_rec left state v1 ; write_rec right state v2
  | Tup e ->
      write_rec e state value
  | Tups {left; right; _} ->
      let (v1, v2) = value in
      write_rec left state v1 ; write_rec right state v2
  | Conv {encoding = e; proj; _} ->
      write_rec e state (proj value)
  | Union {tag_size; cases; _} ->
      let rec write_case = function
        | [] ->
            raise No_case_matched
        | Case {tag = Json_only; _} :: tl ->
            write_case tl
        | Case {encoding = e; proj; tag = Tag tag; _} :: tl -> (
          match proj value with
          | None ->
              write_case tl
          | Some value ->
              Atom.tag tag_size state tag ;
              write_rec e state value )
      in
      write_case cases
  | Dynamic_size {kind; encoding = e} ->
      let initial_offset = state.offset in
      Atom.int kind state 0 ;
      (* place holder for [size] *)
      write_with_limit (Binary_size.max_int kind) e state value ;
      (* patch the written [size] *)
      Atom.set_int
        kind
        state.buffer
        initial_offset
        (state.offset - initial_offset - Binary_size.integer_to_size kind)
  | Check_size {limit; encoding = e} ->
      write_with_limit limit e state value
  | Describe {encoding = e; _} ->
      write_rec e state value
  | Splitted {encoding = e; _} ->
      write_rec e state value
  | Mu {fix; _} ->
      write_rec (fix e) state value
  | Delayed f ->
      write_rec (f ()) state value

and write_with_limit : type a. int -> a Encoding.t -> state -> a -> unit =
 fun limit e state value ->
  (* backup the current limit *)
  let old_limit = state.allowed_bytes in
  (* install the new limit (only if smaller than the current limit) *)
  let limit =
    match state.allowed_bytes with
    | None ->
        limit
    | Some old_limit ->
        min old_limit limit
  in
  state.allowed_bytes <- Some limit ;
  write_rec e state value ;
  (* restore the previous limit (minus the read bytes) *)
  match old_limit with
  | None ->
      state.allowed_bytes <- None
  | Some old_limit ->
      let remaining =
        match state.allowed_bytes with None -> assert false | Some len -> len
      in
      let read = limit - remaining in
      state.allowed_bytes <- Some (old_limit - read)

(** ******************** *)

(** Various entry points *)

let write_exn e v buffer offset len =
  (* By harcoding [allowed_bytes] with the buffer length,
       we ensure that [write] will never reallocate the buffer. *)
  let state = {buffer; offset; allowed_bytes = Some len} in
  write_rec e state v ; state.offset

let write e v buffer offset len =
  try Ok (write_exn e v buffer offset len) with Write_error err -> Error err

let write_opt e v buffer offset len =
  try Some (write_exn e v buffer offset len) with Write_error _ -> None

let to_bytes_exn ?(buffer_size = 128) e v =
  match Encoding.classify e with
  | `Fixed n ->
      (* Preallocate the complete buffer *)
      let state =
        {buffer = Bytes.create n; offset = 0; allowed_bytes = Some n}
      in
      write_rec e state v ; state.buffer
  | `Dynamic | `Variable ->
      (* Preallocate a minimal buffer and let's not hardcode a
         limit to its extension. *)
      let state =
        {buffer = Bytes.create buffer_size; offset = 0; allowed_bytes = None}
      in
      write_rec e state v ;
      Bytes.sub state.buffer 0 state.offset

let to_bytes_opt ?buffer_size e v =
  try Some (to_bytes_exn ?buffer_size e v) with Write_error _ -> None

let to_bytes ?buffer_size e v =
  try Ok (to_bytes_exn ?buffer_size e v) with Write_error err -> Error err
back to top