okjson.rb

# encoding: UTF-8
#
# Copyright 2011, 2012 Keith Rarick
#
# 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.

# See https://github.com/kr/okjson for updates.

require 'stringio'

# Some parts adapted from
# https://golang.org/src/encoding/json/decode.go and
# https://golang.org/src/unicode/utf8/utf8.go
module OkJson
  Upstream = '45'
  extend self


  # Decodes a json document in string s and
  # returns the corresponding ruby value.
  # String s must be valid UTF-8. If you have
  # a string in some other encoding, convert
  # it first.
  #
  # String values in the resulting structure
  # will be UTF-8.
  def decode(s)
    ts = lex(s)
    v, ts = textparse(ts)
    if ts.length > 0
      raise Error, 'trailing garbage'
    end
    v
  end


  # Encodes x into a json text. It may contain only
  # Array, Hash, String, Numeric, true, false, nil.
  # (Note, this list excludes Symbol.)
  # X itself must be an Array or a Hash.
  # No other value can be encoded, and an error will
  # be raised if x contains any other value, such as
  # Nan, Infinity, Symbol, and Proc, or if a Hash key
  # is not a String.
  # Strings contained in x must be valid UTF-8.
  def encode(x)
    case x
    when Hash    then objenc(x)
    when Array   then arrenc(x)
    else
      raise Error, 'root value must be an Array or a Hash'
    end
  end


  def valenc(x)
    case x
    when Hash    then objenc(x)
    when Array   then arrenc(x)
    when String  then strenc(x)
    when Numeric then numenc(x)
    when true    then "true"
    when false   then "false"
    when nil     then "null"
    else
      raise Error, "cannot encode #{x.class}: #{x.inspect}"
    end
  end


private


  # Parses a "json text" in the sense of RFC 4627.
  # Returns the parsed value and any trailing tokens.
  # Note: this is almost the same as valparse,
  # except that it does not accept atomic values.
  def textparse(ts)
    if ts.length <= 0
      raise Error, 'empty'
    end

    typ, _, val = ts[0]
    case typ
    when '{' then objparse(ts)
    when '[' then arrparse(ts)
    else
      raise Error, "unexpected #{val.inspect}"
    end
  end


  # Parses a "value" in the sense of RFC 4627.
  # Returns the parsed value and any trailing tokens.
  def valparse(ts)
    if ts.length <= 0
      raise Error, 'empty'
    end

    typ, _, val = ts[0]
    case typ
    when '{' then objparse(ts)
    when '[' then arrparse(ts)
    when :val,:str then [val, ts[1..-1]]
    else
      raise Error, "unexpected #{val.inspect}"
    end
  end


  # Parses an "object" in the sense of RFC 4627.
  # Returns the parsed value and any trailing tokens.
  def objparse(ts)
    ts = eat('{', ts)
    obj = {}

    if ts[0][0] == '}'
      return obj, ts[1..-1]
    end

    k, v, ts = pairparse(ts)
    obj[k] = v

    if ts[0][0] == '}'
      return obj, ts[1..-1]
    end

    loop do
      ts = eat(',', ts)

      k, v, ts = pairparse(ts)
      obj[k] = v

      if ts[0][0] == '}'
        return obj, ts[1..-1]
      end
    end
  end


  # Parses a "member" in the sense of RFC 4627.
  # Returns the parsed values and any trailing tokens.
  def pairparse(ts)
    (typ, _, k), ts = ts[0], ts[1..-1]
    if typ != :str
      raise Error, "unexpected #{k.inspect}"
    end
    ts = eat(':', ts)
    v, ts = valparse(ts)
    [k, v, ts]
  end


  # Parses an "array" in the sense of RFC 4627.
  # Returns the parsed value and any trailing tokens.
  def arrparse(ts)
    ts = eat('[', ts)
    arr = []

    if ts[0][0] == ']'
      return arr, ts[1..-1]
    end

    v, ts = valparse(ts)
    arr << v

    if ts[0][0] == ']'
      return arr, ts[1..-1]
    end

    loop do
      ts = eat(',', ts)

      v, ts = valparse(ts)
      arr << v

      if ts[0][0] == ']'
        return arr, ts[1..-1]
      end
    end
  end


  def eat(typ, ts)
    if ts[0][0] != typ
      raise Error, "expected #{typ} (got #{ts[0].inspect})"
    end
    ts[1..-1]
  end


  # Scans s and returns a list of json tokens,
  # excluding white space (as defined in RFC 4627).
  def lex(s)
    ts = []
    while s.length > 0
      typ, lexeme, val = tok(s)
      if typ == nil
        raise Error, "invalid character at #{s[0,10].inspect}"
      end
      if typ != :space
        ts << [typ, lexeme, val]
      end
      s = s[lexeme.length..-1]
    end
    ts
  end


  # Scans the first token in s and
  # returns a 3-element list, or nil
  # if s does not begin with a valid token.
  #
  # The first list element is one of
  # '{', '}', ':', ',', '[', ']',
  # :val, :str, and :space.
  #
  # The second element is the lexeme.
  #
  # The third element is the value of the
  # token for :val and :str, otherwise
  # it is the lexeme.
  def tok(s)
    case s[0]
    when ?{ then ['{', s[0,1], s[0,1]]
    when ?} then ['}', s[0,1], s[0,1]]
    when ?: then [':', s[0,1], s[0,1]]
    when ?, then [',', s[0,1], s[0,1]]
    when ?[ then ['[', s[0,1], s[0,1]]
    when ?] then [']', s[0,1], s[0,1]]
    when ?n then nulltok(s)
    when ?t then truetok(s)
    when ?f then falsetok(s)
    when ?" then strtok(s)
    when Spc, ?\t, ?\n, ?\r then [:space, s[0,1], s[0,1]]
    else
      numtok(s)
    end
  end


  def nulltok(s);  s[0,4] == 'null'  ? [:val, 'null',  nil]   : [] end
  def truetok(s);  s[0,4] == 'true'  ? [:val, 'true',  true]  : [] end
  def falsetok(s); s[0,5] == 'false' ? [:val, 'false', false] : [] end


  def numtok(s)
    m = /(-?(?:[1-9][0-9]+|[0-9]))([.][0-9]+)?([eE][+-]?[0-9]+)?/.match(s)
    if m && m.begin(0) == 0
      if !m[2] && !m[3]
        [:val, m[0], Integer(m[0])]
      elsif m[2]
        [:val, m[0], Float(m[0])]
      else
        [:val, m[0], Integer(m[1])*(10**m[3][1..-1].to_i(10))]
      end
    else
      []
    end
  end


  def strtok(s)
    m = /"([^"\\]|\\["\/\\bfnrt]|\\u[0-9a-fA-F]{4})*"/.match(s)
    if ! m
      raise Error, "invalid string literal at #{abbrev(s)}"
    end
    [:str, m[0], unquote(m[0])]
  end


  def abbrev(s)
    t = s[0,10]
    p = t['`']
    t = t[0,p] if p
    t = t + '...' if t.length < s.length
    '`' + t + '`'
  end


  # Converts a quoted json string literal q into a UTF-8-encoded string.
  # The rules are different than for Ruby, so we cannot use eval.
  # Unquote will raise an error if q contains control characters.
  def unquote(q)
    q = q[1...-1]
    a = q.dup # allocate a big enough string
    # In ruby >= 1.9, a[w] is a codepoint, not a byte.
    if rubydoesenc?
      a.force_encoding('UTF-8')
    end
    r, w = 0, 0
    while r < q.length
      c = q[r]
      if c == ?\\
        r += 1
        if r >= q.length
          raise Error, "string literal ends with a \"\\\": \"#{q}\""
        end

        case q[r]
        when ?",?\\,?/,?'
          a[w] = q[r]
          r += 1
          w += 1
        when ?b,?f,?n,?r,?t
          a[w] = Unesc[q[r]]
          r += 1
          w += 1
        when ?u
          r += 1
          uchar = begin
            hexdec4(q[r,4])
          rescue RuntimeError => e
            raise Error, "invalid escape sequence \\u#{q[r,4]}: #{e}"
          end
          r += 4
          if surrogate? uchar
            if q.length >= r+6
              uchar1 = hexdec4(q[r+2,4])
              uchar = subst(uchar, uchar1)
              if uchar != Ucharerr
                # A valid pair; consume.
                r += 6
              end
            end
          end
          if rubydoesenc?
            a[w] = '' << uchar
            w += 1
          else
            w += ucharenc(a, w, uchar)
          end
        else
          raise Error, "invalid escape char #{q[r]} in \"#{q}\""
        end
      elsif c == ?" || c < Spc
        raise Error, "invalid character in string literal \"#{q}\""
      else
        # Copy anything else byte-for-byte.
        # Valid UTF-8 will remain valid UTF-8.
        # Invalid UTF-8 will remain invalid UTF-8.
        # In ruby >= 1.9, c is a codepoint, not a byte,
        # in which case this is still what we want.
        a[w] = c
        r += 1
        w += 1
      end
    end
    a[0,w]
  end


  # Encodes unicode character u as UTF-8
  # bytes in string a at position i.
  # Returns the number of bytes written.
  def ucharenc(a, i, u)
    if u <= Uchar1max
      a[i] = (u & 0xff).chr
      1
    elsif u <= Uchar2max
      a[i+0] = (Utag2 | ((u>>6)&0xff)).chr
      a[i+1] = (Utagx | (u&Umaskx)).chr
      2
    elsif u <= Uchar3max
      a[i+0] = (Utag3 | ((u>>12)&0xff)).chr
      a[i+1] = (Utagx | ((u>>6)&Umaskx)).chr
      a[i+2] = (Utagx | (u&Umaskx)).chr
      3
    else
      a[i+0] = (Utag4 | ((u>>18)&0xff)).chr
      a[i+1] = (Utagx | ((u>>12)&Umaskx)).chr
      a[i+2] = (Utagx | ((u>>6)&Umaskx)).chr
      a[i+3] = (Utagx | (u&Umaskx)).chr
      4
    end
  end


  def hexdec4(s)
    if s.length != 4
      raise Error, 'short'
    end
    (nibble(s[0])<<12) | (nibble(s[1])<<8) | (nibble(s[2])<<4) | nibble(s[3])
  end


  def subst(u1, u2)
    if Usurr1 <= u1 && u1 < Usurr2 && Usurr2 <= u2 && u2 < Usurr3
      return ((u1-Usurr1)<<10) | (u2-Usurr2) + Usurrself
    end
    return Ucharerr
  end


  def surrogate?(u)
    Usurr1 <= u && u < Usurr3
  end


  def nibble(c)
    if ?0 <= c && c <= ?9 then c.ord - ?0.ord
    elsif ?a <= c && c <= ?z then c.ord - ?a.ord + 10
    elsif ?A <= c && c <= ?Z then c.ord - ?A.ord + 10
    else
      raise Error, "invalid hex code #{c}"
    end
  end


  def objenc(x)
    '{' + x.map{|k,v| keyenc(k) + ':' + valenc(v)}.join(',') + '}'
  end


  def arrenc(a)
    '[' + a.map{|x| valenc(x)}.join(',') + ']'
  end


  def keyenc(k)
    case k
    when String then strenc(k)
    else
      raise Error, "Hash key is not a string: #{k.inspect}"
    end
  end


  def strenc(s)
    t = StringIO.new
    t.putc(?")
    r = 0

    while r < s.length
      case s[r]
      when ?"  then t.print('\\"')
      when ?\\ then t.print('\\\\')
      when ?\b then t.print('\\b')
      when ?\f then t.print('\\f')
      when ?\n then t.print('\\n')
      when ?\r then t.print('\\r')
      when ?\t then t.print('\\t')
      else
        c = s[r]
        # In ruby >= 1.9, s[r] is a codepoint, not a byte.
        if rubydoesenc?
          begin
            # c.ord will raise an error if c is invalid UTF-8
            if c.ord < Spc.ord
              c = "\\u%04x" % [c.ord]
            end
            t.write(c)
          rescue
            t.write(Ustrerr)
          end
        elsif c < Spc
          t.write("\\u%04x" % c)
        elsif Spc <= c && c <= ?~
          t.putc(c)
        else
          n = ucharcopy(t, s, r) # ensure valid UTF-8 output
          r += n - 1 # r is incremented below
        end
      end
      r += 1
    end
    t.putc(?")
    t.string
  end


  def numenc(x)
    if ((x.nan? || x.infinite?) rescue false)
      raise Error, "Numeric cannot be represented: #{x}"
    end
    "#{x}"
  end


  # Copies the valid UTF-8 bytes of a single character
  # from string s at position i to I/O object t, and
  # returns the number of bytes copied.
  # If no valid UTF-8 char exists at position i,
  # ucharcopy writes Ustrerr and returns 1.
  def ucharcopy(t, s, i)
    n = s.length - i
    raise Utf8Error if n < 1

    c0 = s[i].ord

    # 1-byte, 7-bit sequence?
    if c0 < Utagx
      t.putc(c0)
      return 1
    end

    raise Utf8Error if c0 < Utag2 # unexpected continuation byte?

    raise Utf8Error if n < 2 # need continuation byte
    c1 = s[i+1].ord
    raise Utf8Error if c1 < Utagx || Utag2 <= c1

    # 2-byte, 11-bit sequence?
    if c0 < Utag3
      raise Utf8Error if ((c0&Umask2)<<6 | (c1&Umaskx)) <= Uchar1max
      t.putc(c0)
      t.putc(c1)
      return 2
    end

    # need second continuation byte
    raise Utf8Error if n < 3

    c2 = s[i+2].ord
    raise Utf8Error if c2 < Utagx || Utag2 <= c2

    # 3-byte, 16-bit sequence?
    if c0 < Utag4
      u = (c0&Umask3)<<12 | (c1&Umaskx)<<6 | (c2&Umaskx)
      raise Utf8Error if u <= Uchar2max
      t.putc(c0)
      t.putc(c1)
      t.putc(c2)
      return 3
    end

    # need third continuation byte
    raise Utf8Error if n < 4
    c3 = s[i+3].ord
    raise Utf8Error if c3 < Utagx || Utag2 <= c3

    # 4-byte, 21-bit sequence?
    if c0 < Utag5
      u = (c0&Umask4)<<18 | (c1&Umaskx)<<12 | (c2&Umaskx)<<6 | (c3&Umaskx)
      raise Utf8Error if u <= Uchar3max
      t.putc(c0)
      t.putc(c1)
      t.putc(c2)
      t.putc(c3)
      return 4
    end

    raise Utf8Error
  rescue Utf8Error
    t.write(Ustrerr)
    return 1
  end


  def rubydoesenc?
    ::String.method_defined?(:force_encoding)
  end


  class Utf8Error < ::StandardError
  end


  class Error < ::StandardError
  end


  Utagx = 0b1000_0000
  Utag2 = 0b1100_0000
  Utag3 = 0b1110_0000
  Utag4 = 0b1111_0000
  Utag5 = 0b1111_1000
  Umaskx = 0b0011_1111
  Umask2 = 0b0001_1111
  Umask3 = 0b0000_1111
  Umask4 = 0b0000_0111
  Uchar1max = (1<<7) - 1
  Uchar2max = (1<<11) - 1
  Uchar3max = (1<<16) - 1
  Ucharerr = 0xFFFD # unicode "replacement char"
  Ustrerr = "\xef\xbf\xbd" # unicode "replacement char"
  Usurrself = 0x10000
  Usurr1 = 0xd800
  Usurr2 = 0xdc00
  Usurr3 = 0xe000

  Spc = ' '[0]
  Unesc = {?b=>?\b, ?f=>?\f, ?n=>?\n, ?r=>?\r, ?t=>?\t}
end