bncode.js

/*jshint es5:false, asi:true, quotmark:false, eqeqeq:false, forin: false */

/*
 * (c) 2011-14 Tim Becker, see file LICENSE for details
 */

/*
 * Provides functionality for bencoding and decoding as use in
 * bittorrent and described in: http://www.bittorrent.org/beps/bep_0003.html
 *
 * Encoding is as follows:
 *
 *    var benc  = require('bncode'),
 *        exmp = {}
 *
 *    exmp.bla = "blup"
 *    exmp.foo = "bar"
 *    exmp.one = 1
 *    exmp.woah = {}
 *    exmp.woah.arr = []
 *    exmp.woah.arr.push(1)
 *    exmp.woah.arr.push(2)
 *    exmp.woah.arr.push(3)
 *    exmp.str = new Buffer("Buffers work too")
 *
 *    var bencBuffer = benc.encode(exmp) i
 *
 *    // d3:bla4:blup3:foo3:bar3:onei1e4:woahd3:arr \
 *    // li1ei2ei3eee3:str16:Buffers work tooe
 *
 *
 * Decoding will work in progressively, e.g. if you're receiving partial
 * bencoded strings on the network:
 *
 *    var benc = require("bncode"),
 *        buf  = null
 *
 *    decoder = new bncode.decoder()
 *    while (buf = receiveData()) {
 *      decoder.decode(buf)
 *    }
 *
 *    log(decoder.result())
 *
 *
 * Or "all in one"
 *
 *    var benc = require("bncode"),
 *        buf  = getBuffer(),
 *        dec  = benc.decode(buf)
 *
 *    log(dec.bla)
 *
 *
 * There are some subtleties concerning bencoded strings. These are
 * decoded as Buffer objects because they are just strings of raw bytes
 * and as such would wreak havoc with multi byte strings in javascript.
 *
 * The exception to this is strings that appear as keys in bencoded
 * dicts. These are decoded as Javascript Strings, as they should always
 * be strings of (ascii) characters and if they weren't decoded as JS
 * Strings, dict's would map to Javascript objects with properties.
 *
 */

exports.encode  = Bencode
exports.decoder = Bdecode
exports.decode  = decode
exports.Stream  = Stream

var inherits = require('util').inherits
var Transform = require('stream').Transform

var I     = 'i'.charCodeAt(0)
var L     = 'l'.charCodeAt(0)
var E     = 'e'.charCodeAt(0)
var D     = 'd'.charCodeAt(0)
var COLON = ':'.charCodeAt(0)
var DASH  = '-'.charCodeAt(0)

var STATE_INITIAL          = 0
var STATE_STATE_STRING_LEN = STATE_INITIAL          + 1
var STATE_STRING           = STATE_STATE_STRING_LEN + 1
var STATE_COLON            = STATE_STRING           + 1
var STATE_STATE_INTEGER    = STATE_COLON            + 1
var STATE_INTEGER          = STATE_STATE_INTEGER    + 1

/*
 * This is the internal state machine for taking apart bencoded strings,
 * it's not exposed in the eports.  It's constructed with four callbacks
 * that get fired when:
 *
 * cb: a value (string or number) is encountered
 * cb_list: a begin list element is encountered
 * cb_dict: a beginning of dictionary is encountered.
 * cd_end: an end element, wheter dict or list is encountered
 *
 * Once constructed, the machine may be fed with buffers containing
 * partial bencoded string. Call `consistent` to check whether the
 * current state is consistent, e.g. not smack-dap in the middle of
 * a string or a number and if the dict, list and end calls balance
 *
 *
 * The functionality being so rudimentary requires some more state and
 * logic in the code executing the machine, for this see Context, below.
 *
 */

function BdecodeSMachine (cb, cb_list, cb_dict, cb_end) {
  var depth = 0
  var state = STATE_INITIAL

  this.consistent = function () {
    return state === STATE_INITIAL && depth === 0
  }

  var strLen = 0
  var str    = ''
  var _int   = 0
  var neg    = false

  this.parse = function (buffer, encoding) {
    if (typeof buffer === 'string') {
      buffer = new Buffer(buffer, encoding || 'utf8')
    }

    for (var pos = 0; pos !== buffer.length; ++pos) {
      switch (state) {
        case STATE_INITIAL:
          switch (buffer[pos]) {
            case 0x30:
            case 0x31:
            case 0x32:
            case 0x33:
            case 0x34:
            case 0x35:
            case 0x36:
            case 0x37:
            case 0x38:
            case 0x39:
              state = STATE_STATE_STRING_LEN
              strLen = 0
              strLen += buffer[pos] - 0x30
              break
            case I:
              state = STATE_STATE_INTEGER
              _int  = 0
              neg   = false
              break
            case L:
              state = STATE_INITIAL
              depth += 1
              cb_list()
              break
            case D:
              state = STATE_INITIAL
              depth += 1
              cb_dict()
              break
            case E:
              state = STATE_INITIAL
              depth -= 1
              if (depth < 0) {
                throw new Error('end with no beginning: ' + pos)
              } else {
                cb_end()
              }
              break
          }
          break
        case STATE_STATE_STRING_LEN:
          if (integer(buffer[pos])) {
            strLen *= 10
            strLen += buffer[pos] - 0x30
          } else {
            str = new Buffer(strLen)
            pos -=1
            state = STATE_COLON
          }
          break
        case STATE_COLON:
          if (buffer[pos] !== COLON) {
            throw new Error('not a colon at: ' + pos.toString(16))
          }
          state = STATE_STRING
          // in case this is a zero length string, there's
          // no bytes to be collected.
          if (0 === strLen) {
            cb(new Buffer(0))
            state = STATE_INITIAL
          }
          break
        case STATE_STRING:
          if (0 === strLen) {
            cb(str)
            state = STATE_INITIAL
          } else {
            //str += String.fromCharCode(buffer[pos]) // not unicode safe..
            str[str.length-strLen] = buffer[pos]
            strLen -= 1
            if (0 === strLen) {
              cb(str)
              state = STATE_INITIAL
            }
          }
          break
        case STATE_STATE_INTEGER:
          state = STATE_INTEGER
          if (buffer[pos] === DASH) {
            neg = true  // handle neg and zero within value.
            break
          } // else fall through
        case STATE_INTEGER:
          if (integer(buffer[pos])) {
            _int *= 10
            _int += buffer[pos] - 0x30
          } else if (buffer[pos] === E) {
            var ret = neg ? 0 - _int : _int
            cb(ret)
            state = STATE_INITIAL
          } else {
            throw new Error('not part of int at:'+pos.toString(16))
          }
          break
      } // switch state
    } // for buffer
  } // function parse

  function integer (value) {
    // check that value is a number and that
    // its value is ascii integer.
    if (typeof value !== 'number') {
      return false
    }
    return between(value, 0x30, 0x39)
  }
  function between (val, min, max) {
    return (min <= val  && val <= max)
  }

} // end BdecodeSMachine

/*
 * The exported decode functionality.
 */
function Bdecode () {
  // markers
  var DICTIONARY_START = {}
  var LIST_START       = {}

  var Context  = function () {
    var self  = this
    var stack = []

    this.cb = function (o) {
      stack.push(o)
    }
    this.cb_list = function () {
      self.cb(LIST_START)
    }
    this.cb_dict = function () {
      self.cb(DICTIONARY_START)
    }

    this.cb_end = function () {

      // unwind the stack until either a DICTIONARY_START or LIST_START is
      // found, create arr or hash, stick unwound stack on, push arr or hash
      // back onto stack

      var obj       = null
      var tmp_stack = []

      while ((obj = stack.pop()) !== undefined) {
        if (LIST_START === obj) {
          var obj2 = null
          var list = []
          while((obj2 = tmp_stack.pop()) !== undefined) {
            list.push(obj2)
          }
          self.cb(list)
          break
        } else if (DICTIONARY_START === obj) {
          var key = null
          var val = null
          var dic = {}
          while ((key = tmp_stack.pop()) !== undefined && (val = tmp_stack.pop()) !== undefined) {
            dic[key.toString()] = val
          }

          if (key !== undefined && dic[key] === undefined) {
            throw new Error('uneven number of keys and values A')
          }
          self.cb(dic)
          break
        } else {
          tmp_stack.push(obj)
        }
      }
      if (tmp_stack.length > 0) {
        // could this case even occur?
        throw new Error('uneven number of keys and values B')
      }
    }
    this.result = function () {
      return stack
    }
  }

  var self     = this
  var ctx      = new Context()
  var smachine = new BdecodeSMachine(ctx.cb, ctx.cb_list, ctx.cb_dict, ctx.cb_end)

  this.result = function () {
    if (!smachine.consistent()) {
      throw new Error('not in consistent state. More bytes coming?')
    }
    return ctx.result()
  }

  this.decode = function (buf, encoding) {
    smachine.parse(buf, encoding)
  }
}

function Bencode (obj) {
  var self = this
  var to_encode = obj
  var buffer = null

  switch (typeof obj) {
    case 'string':
      return encodeString(obj)
    case 'number':
      return encodeNumber(obj)
    case 'object':
      if (obj instanceof Array) {
        return encodeList(obj)
      } else if (Buffer.isBuffer(obj)) {
        return encodeBuffer(obj)
      } else {
        // assume it's a hash
        return encodeDict(obj)
      }
  }

  function encodeString (obj) {
    var blen = Buffer.byteLength(obj)
    var len  = blen.toString(10)
    var buf  = new Buffer(len.length + 1 + blen)

    buf.write(len, 0, 'ascii')
    buf.write(':', len.length, 'ascii')
    buf.write(obj, len.length + 1, 'utf8')

    return buf
  }

  function encodeNumber (num) {
    var n   = num.toString(10)
    var buf = new Buffer(n.length + 2)

    buf.write('i', 0)
    buf.write(n, 1)
    buf.write('e', n.length + 1)

    return buf
  }

  function encodeDict (obj) {
    var func = function (obj, pos) {
      var keys = Object.keys(obj).sort()
      keys.forEach(function (key) {
        var val = new Bencode(obj[key])
        key = new Bencode(key)

        ensure(key.length + val.length, pos)
        key.copy(buffer, pos, 0)
        pos += key.length
        val.copy(buffer, pos, 0)
        pos += val.length
      })
      return pos
    }
    return assemble(obj, 'd', func)
  }

  function encodeList (obj) {
    var func = function(obj, pos) {
      obj.forEach(function (o) {
        var elem = new Bencode(o)

        ensure(elem.length, pos)
        elem.copy(buffer, pos, 0)
        pos += elem.length
      })
      return pos
    }
    return assemble(obj, 'l', func)
  }

  function encodeBuffer (obj) {
    var len = obj.length.toString(10)
    var buf = new Buffer(len.length + 1 + obj.length)

    buf.write(len, 0, 'ascii')
    buf.write(':', len.length, 'ascii')
    obj.copy(buf, len.length + 1, 0)

    return buf
  }

  function assemble (obj, prefix, func) {
    var pos = 0

    ensure(1024, 0)
    buffer.write(prefix, pos++)

    pos = func(obj, pos)
    ensure(1, pos)

    buffer.write('e', pos++)
    return buffer.slice(0, pos)
  }

  function ensure (num, pos) {
    if (!buffer) {
      buffer = new Buffer(num)
    } else {
      if (buffer.length > num + pos + 1) {
        return
      } else {
        var buf2 = new Buffer(buffer.length + num)
        buffer.copy(buf2, 0, 0)
        buffer = buf2
      }
    }
  }
}

function decode (buffer, encoding) {
  var decoder = new Bdecode()

  decoder.decode(buffer, encoding)
  return decoder.result()[0]
}

function Stream (options) {
  options = options || {}
  options.objectMode = true
  Transform.call(this, options)
  this._decoder = new Bdecode()
}

inherits(Stream, Transform)

Stream.prototype._transform = function (chunk, encoding, callback) {
  try {
    this._decoder.decode(chunk, encoding)
    callback(null)
  } catch(err) {
    callback(err)
  }
}

Stream.prototype._flush = function (callback) {
  this.push(this._decoder.result()[0])
  callback(null)
}