ruby_parser 3.12.0 → 3.18.1

data/lib/ruby_lexer.rb

@@ -4,36 +4,9 @@
 $DEBUG = true if ENV["DEBUG"]
 
 class RubyLexer
-
   # :stopdoc:
-  HAS_ENC = "".respond_to? :encoding
-
-  IDENT_CHAR = if HAS_ENC then
-                 /[\w\u0080-\u{10ffff}]/u
-               else
-                 /[\w\x80-\xFF]/n
-               end
-
   EOF = :eof_haha!
 
-  # ruby constants for strings (should this be moved somewhere else?)
-
-  STR_FUNC_BORING = 0x00
-  STR_FUNC_ESCAPE = 0x01 # TODO: remove and replace with REGEXP
-  STR_FUNC_EXPAND = 0x02
-  STR_FUNC_REGEXP = 0x04
-  STR_FUNC_QWORDS = 0x08
-  STR_FUNC_SYMBOL = 0x10
-  STR_FUNC_INDENT = 0x20 # <<-HEREDOC
-  STR_FUNC_ICNTNT = 0x40 # <<~HEREDOC
-
-  STR_SQUOTE = STR_FUNC_BORING
-  STR_DQUOTE = STR_FUNC_BORING | STR_FUNC_EXPAND
-  STR_XQUOTE = STR_FUNC_BORING | STR_FUNC_EXPAND
-  STR_REGEXP = STR_FUNC_REGEXP | STR_FUNC_ESCAPE | STR_FUNC_EXPAND
-  STR_SSYM   = STR_FUNC_SYMBOL
-  STR_DSYM   = STR_FUNC_SYMBOL | STR_FUNC_EXPAND
-
   ESCAPES = {
     "a"    => "\007",
     "b"    => "\010",
@@ -50,10 +23,17 @@ class RubyLexer
     "c\?"  => 127.chr,
   }
 
+  HAS_ENC = "".respond_to? :encoding
+
+  BTOKENS = {
+    ".."  => :tBDOT2,
+    "..." => :tBDOT3,
+  }
+
   TOKENS = {
     "!"   => :tBANG,
     "!="  => :tNEQ,
-    # "!@"  => :tUBANG,
+    "!@"  => :tBANG,
     "!~"  => :tNMATCH,
     ","   => :tCOMMA,
     ".."  => :tDOT2,
@@ -66,27 +46,62 @@ class RubyLexer
     "->"  => :tLAMBDA,
   }
 
-  TAB_WIDTH = 8
+  PERCENT_END = {
+    "(" => ")",
+    "[" => "]",
+    "{" => "}",
+    "<" => ">",
+  }
 
-  @@regexp_cache = Hash.new { |h,k| h[k] = Regexp.new(Regexp.escape(k)) }
+  SIMPLE_RE_META = /[\$\*\+\.\?\^\|\)\]\}\>]/
+
+  @@regexp_cache = Hash.new { |h, k| h[k] = Regexp.new(Regexp.escape(k)) }
   @@regexp_cache[nil] = nil
 
+  def regexp_cache
+    @@regexp_cache
+  end
+
+  if $DEBUG then
+    attr_reader :lex_state
+
+    def lex_state= o
+      return if @lex_state == o
+
+      from = ""
+      if ENV["VERBOSE"]
+        path = caller[0]
+        path = caller[1] if path =~ /result/
+        path, line, *_ = path.split(/:/)
+        path.delete_prefix! File.dirname File.dirname __FILE__
+        from = " at .%s:%s" % [path, line]
+      end
+
+      warn "lex_state: %p -> %p%s" % [lex_state, o, from]
+
+      @lex_state = o
+    end
+  end
+
   # :startdoc:
 
-  attr_accessor :lineno # we're bypassing oedipus' lineno handling.
+  attr_accessor :lex_state unless $DEBUG
+
   attr_accessor :brace_nest
   attr_accessor :cmdarg
   attr_accessor :command_start
-  attr_accessor :command_state
+  attr_accessor :cmd_state # temporary--ivar to avoid passing everywhere
   attr_accessor :last_state
   attr_accessor :cond
-  attr_accessor :extra_lineno
+  attr_accessor :old_ss
+  attr_accessor :old_lineno
+
+  # these are generated via ruby_lexer.rex: ss, lineno
 
   ##
   # Additional context surrounding tokens that both the lexer and
   # grammar use.
 
-  attr_accessor :lex_state
   attr_accessor :lex_strterm
   attr_accessor :lpar_beg
   attr_accessor :paren_nest
@@ -95,50 +110,33 @@ class RubyLexer
   attr_accessor :string_buffer
   attr_accessor :string_nest
 
-  if $DEBUG then
-    alias lex_state= lex_state=
-    def lex_state=o
-      return if @lex_state == o
-      c = caller.first
-      c = caller[1] if c =~ /\bresult\b/
-      warn "lex_state: %p -> %p from %s" % [@lex_state, o, c.clean_caller]
-      @lex_state = o
-    end
-  end
-
   # Last token read via next_token.
   attr_accessor :token
 
-  ##
-  # What version of ruby to parse. 18 and 19 are the only valid values
-  # currently supported.
-
-  attr_accessor :version
-
   attr_writer :comments
 
-  def initialize v = 18
-    self.version = v
-    @lex_state = :expr_none
+  def initialize _ = nil
+    @lex_state = nil # remove one warning under $DEBUG
+    self.lex_state = EXPR_NONE
 
-    self.cmdarg = RubyParserStuff::StackState.new(:cmdarg, $DEBUG)
     self.cond   = RubyParserStuff::StackState.new(:cond, $DEBUG)
+    self.cmdarg = RubyParserStuff::StackState.new(:cmdarg, $DEBUG)
+    self.ss     = RPStringScanner.new ""
 
     reset
   end
 
   def arg_ambiguous
-    self.warning("Ambiguous first argument. make sure.")
+    self.warning "Ambiguous first argument. make sure."
   end
 
   def arg_state
-    in_arg_state? ? :expr_arg : :expr_beg
+    is_after_operator? ? EXPR_ARG : EXPR_BEG
   end
 
-  def beginning_of_line?
-    ss.bol?
+  def ignore_body_comments
+    @comments.clear
   end
-  alias :bol? :beginning_of_line? # to make .rex file more readable
 
   def comments # TODO: remove this... maybe comment_string + attr_accessor
     c = @comments.join
@@ -146,184 +144,26 @@ class RubyLexer
     c
   end
 
-  def end_of_stream?
-    ss.eos?
+  def debug n
+    raise "debug #{n}"
   end
 
   def expr_dot?
-    lex_state == :expr_dot
+    lex_state =~ EXPR_DOT
   end
 
-  def expr_fname?
-    lex_state == :expr_fname
+  def expr_fname? # REFACTOR
+    lex_state =~ EXPR_FNAME
   end
 
   def expr_result token, text
     cond.push false
     cmdarg.push false
-    result :expr_beg, token, text
-  end
-
-  def heredoc here # TODO: rewrite / remove
-    _, eos, func, last_line = here
-
-    indent         = (func & STR_FUNC_INDENT) != 0 ? "[ \t]*" : nil
-    content_indent = (func & STR_FUNC_ICNTNT) != 0
-    expand         = (func & STR_FUNC_EXPAND) != 0
-    eos_re         = /#{indent}#{Regexp.escape eos}(\r*\n|\z)/
-    err_msg        = "can't match #{eos_re.inspect} anywhere in "
-
-    rb_compile_error err_msg if end_of_stream?
-
-    if beginning_of_line? && scan(eos_re) then
-      self.lineno += 1
-      ss.unread_many last_line # TODO: figure out how to remove this
-      return :tSTRING_END, eos
-    end
-
-    self.string_buffer = []
-
-    if expand then
-      case
-      when scan(/#[$@]/) then
-        ss.pos -= 1 # FIX omg stupid
-        return :tSTRING_DVAR, matched
-      when scan(/#[{]/) then
-        return :tSTRING_DBEG, matched
-      when scan(/#/) then
-        string_buffer << '#'
-      end
-
-      begin
-        c = tokadd_string func, "\n", nil
-
-        rb_compile_error err_msg if
-          c == RubyLexer::EOF
-
-        if c != "\n" then
-          return :tSTRING_CONTENT, string_buffer.join.delete("\r")
-        else
-          string_buffer << scan(/\n/)
-        end
-
-        rb_compile_error err_msg if end_of_stream?
-      end until check(eos_re)
-    else
-      until check(eos_re) do
-        string_buffer << scan(/.*(\n|\z)/)
-        rb_compile_error err_msg if end_of_stream?
-      end
-    end
-
-    self.lex_strterm = [:heredoc, eos, func, last_line]
-
-    string_content = string_buffer.join.delete("\r")
-
-    string_content = heredoc_dedent(string_content) if content_indent && ruby23plus?
-
-    return :tSTRING_CONTENT, string_content
+    result EXPR_BEG, token, text
   end
 
-  def heredoc_dedent(string_content)
-    width = string_content.scan(/^[ \t]*(?=\S)/).map do |whitespace|
-      heredoc_whitespace_indent_size whitespace
-    end.min || 0
-
-    string_content.split("\n", -1).map do |line|
-      dedent_string line, width
-    end.join "\n"
-  end
-
-  def dedent_string(string, width)
-    characters_skipped = 0
-    indentation_skipped = 0
-
-    string.chars.each do |char|
-      break if indentation_skipped >= width
-      if char == ' '
-        characters_skipped += 1
-        indentation_skipped += 1
-      elsif char == "\t"
-        proposed = TAB_WIDTH * (indentation_skipped / TAB_WIDTH + 1)
-        break if (proposed > width)
-        characters_skipped += 1
-        indentation_skipped = proposed
-      end
-    end
-    string[characters_skipped..-1]
-  end
-
-  def heredoc_whitespace_indent_size(whitespace)
-    whitespace.chars.inject 0 do |size, char|
-      if char == "\t"
-        size + TAB_WIDTH
-      else
-        size + 1
-      end
-    end
-  end
-
-  def heredoc_identifier # TODO: remove / rewrite
-    term, func = nil, STR_FUNC_BORING
-    self.string_buffer = []
-
-    heredoc_indent_mods = '-'
-    heredoc_indent_mods += '\~' if ruby23plus?
-
-    case
-    when scan(/([#{heredoc_indent_mods}]?)([\'\"\`])(.*?)\2/) then
-      term = ss[2]
-      func |= STR_FUNC_INDENT unless ss[1].empty?
-      func |= STR_FUNC_ICNTNT if ss[1] == '~'
-      func |= case term
-              when "\'" then
-                STR_SQUOTE
-              when '"' then
-                STR_DQUOTE
-              else
-                STR_XQUOTE
-              end
-      string_buffer << ss[3]
-    when scan(/[#{heredoc_indent_mods}]?([\'\"\`])(?!\1*\Z)/) then
-      rb_compile_error "unterminated here document identifier"
-    when scan(/([#{heredoc_indent_mods}]?)(#{IDENT_CHAR}+)/) then
-      term = '"'
-      func |= STR_DQUOTE
-      unless ss[1].empty? then
-        func |= STR_FUNC_INDENT
-        func |= STR_FUNC_ICNTNT if ss[1] == '~'
-      end
-      string_buffer << ss[2]
-    else
-      return nil
-    end
-
-    if scan(/.*\n/) then
-      # TODO: think about storing off the char range instead
-      line = matched
-    else
-      line = nil
-    end
-
-    self.lex_strterm = [:heredoc, string_buffer.join, func, line]
-
-    if term == '`' then
-      result nil, :tXSTRING_BEG, "`"
-    else
-      result nil, :tSTRING_BEG, "\""
-    end
-  end
-
-  def in_fname?
-    in_lex_state? :expr_fname
-  end
-
-  def in_arg_state? # TODO: rename is_after_operator?
-    in_lex_state? :expr_fname, :expr_dot
-  end
-
-  def in_lex_state?(*states)
-    states.include? lex_state
+  def in_fname? # REFACTOR
+    lex_state =~ EXPR_FNAME
   end
 
   def int_with_base base
@@ -331,35 +171,35 @@ class RubyLexer
 
     text = matched
     case
-    when text.end_with?('ri')
-      return result(:expr_end, :tIMAGINARY, Complex(0, Rational(text.chop.chop.to_i(base))))
-    when text.end_with?('r')
-      return result(:expr_end, :tRATIONAL, Rational(text.chop.to_i(base)))
-    when text.end_with?('i')
-      return result(:expr_end, :tIMAGINARY, Complex(0, text.chop.to_i(base)))
+    when text.end_with?("ri")
+      result EXPR_NUM, :tIMAGINARY, Complex(0, Rational(text.chop.chop.to_i(base)))
+    when text.end_with?("r")
+      result EXPR_NUM, :tRATIONAL, Rational(text.chop.to_i(base))
+    when text.end_with?("i")
+      result EXPR_NUM, :tIMAGINARY, Complex(0, text.chop.to_i(base))
     else
-      return result(:expr_end, :tINTEGER, text.to_i(base))
+      result EXPR_NUM, :tINTEGER, text.to_i(base)
     end
   end
 
+  def is_after_operator?
+    lex_state =~ EXPR_FNAME|EXPR_DOT
+  end
+
   def is_arg?
-    in_lex_state? :expr_arg, :expr_cmdarg
+    lex_state =~ EXPR_ARG_ANY
   end
 
   def is_beg?
-    in_lex_state? :expr_beg, :expr_value, :expr_mid, :expr_class, :expr_labelarg
+    lex_state =~ EXPR_BEG_ANY || lex_state == EXPR_LAB # yes, == EXPR_LAB
   end
 
   def is_end?
-    in_lex_state? :expr_end, :expr_endarg, :expr_endfn
-  end
-
-  def ruby22_label?
-    ruby22plus? and is_label_possible?
+    lex_state =~ EXPR_END_ANY
   end
 
   def is_label_possible?
-    (in_lex_state?(:expr_beg, :expr_endfn) && !command_state) || is_arg?
+    (lex_state =~ EXPR_LABEL|EXPR_ENDFN && !cmd_state) || is_arg?
   end
 
   def is_label_suffix?
@@ -370,31 +210,51 @@ class RubyLexer
     is_arg? and space_seen and c !~ /\s/
   end
 
-  def matched
-    ss.matched
+  def lambda_beginning?
+    lpar_beg && lpar_beg == paren_nest
+  end
+
+  def is_local_id id
+    # maybe just make this false for now
+    self.parser.env[id.to_sym] == :lvar # HACK: this isn't remotely right
+  end
+
+  def lvar_defined? id
+    # TODO: (dyna_in_block? && dvar_defined?(id)) || local_id?(id)
+    self.parser.env[id.to_sym] == :lvar
   end
 
   def not_end?
     not is_end?
   end
 
+  def possibly_escape_string text, check
+    content = match[1]
+
+    if text =~ check then
+      content.gsub(ESC) { unescape $1 }
+    else
+      content.gsub(/\\\\/, "\\").gsub(/\\\'/, "'")
+    end
+  end
+
   def process_amper text
     token = if is_arg? && space_seen && !check(/\s/) then
                warning("`&' interpreted as argument prefix")
                :tAMPER
-             elsif in_lex_state? :expr_beg, :expr_mid then
+             elsif lex_state =~ EXPR_BEG|EXPR_MID then
                :tAMPER
              else
                :tAMPER2
              end
 
-    return result(:arg_state, token, "&")
+    result :arg_state, token, "&"
   end
 
   def process_backref text
-    token = ss[1].to_sym
+    token = match[1].to_sym
     # TODO: can't do lineno hack w/ symbol
-    result :expr_end, :tBACK_REF, token
+    result EXPR_END, :tBACK_REF, token
   end
 
   def process_begin text
@@ -406,220 +266,256 @@ class RubyLexer
     end
 
     @comments << matched
-    self.lineno += matched.count("\n")
+    self.lineno += matched.count("\n") # HACK?
 
     nil # TODO
   end
 
-  def process_bracing text
-    cond.lexpop
-    cmdarg.lexpop
-
+  def process_brace_close text
     case matched
     when "}" then
       self.brace_nest -= 1
-      self.lex_state   = :expr_endarg
+      return :tSTRING_DEND, matched if brace_nest < 0
+    end
 
-      # TODO
-      # if (c == '}') {
-      #     if (!brace_nest--) c = tSTRING_DEND;
-      # }
+    # matching compare/parse26.y:8099
+    cond.pop
+    cmdarg.pop
 
+    case matched
+    when "}" then
+      self.lex_state   = ruby24minus? ? EXPR_ENDARG : EXPR_END
       return :tRCURLY, matched
     when "]" then
       self.paren_nest -= 1
-      self.lex_state   = :expr_endarg
+      self.lex_state   = ruby24minus? ? EXPR_ENDARG : EXPR_END
       return :tRBRACK, matched
     when ")" then
       self.paren_nest -= 1
-      self.lex_state   = :expr_endfn
+      self.lex_state   = EXPR_ENDFN
       return :tRPAREN, matched
     else
       raise "Unknown bracing: #{matched.inspect}"
     end
   end
 
+  def process_brace_open text
+    # matching compare/parse23.y:8694
+    self.brace_nest += 1
+
+    if lambda_beginning? then
+      self.lpar_beg = nil
+      self.paren_nest -= 1 # close arg list when lambda opens body
+
+      return expr_result(:tLAMBEG, "{")
+    end
+
+    token = case
+            when lex_state =~ EXPR_LABELED then
+              :tLBRACE     # hash
+            when lex_state =~ EXPR_ARG_ANY|EXPR_END|EXPR_ENDFN then
+              :tLCURLY     # block (primary) "{" in parse.y
+            when lex_state =~ EXPR_ENDARG then
+              :tLBRACE_ARG # block (expr)
+            else
+              :tLBRACE     # hash
+            end
+
+    state = token == :tLBRACE_ARG ? EXPR_BEG : EXPR_PAR
+    self.command_start = true if token != :tLBRACE
+
+    cond.push false
+    cmdarg.push false
+    result state, token, text
+  end
+
   def process_colon1 text
     # ?: / then / when
     if is_end? || check(/\s/) then
-      return result :expr_beg, :tCOLON, text
+      return result EXPR_BEG, :tCOLON, text
     end
 
     case
     when scan(/\'/) then
-      string STR_SSYM
+      string STR_SSYM, matched
     when scan(/\"/) then
-      string STR_DSYM
+      string STR_DSYM, matched
     end
 
-    result :expr_fname, :tSYMBEG, text
+    result EXPR_FNAME, :tSYMBEG, text
   end
 
   def process_colon2 text
-    if is_beg? || in_lex_state?(:expr_class) || is_space_arg? then
-      result :expr_beg, :tCOLON3, text
+    if is_beg? || lex_state =~ EXPR_CLASS || is_space_arg? then
+      result EXPR_BEG, :tCOLON3, text
     else
-      result :expr_dot, :tCOLON2, text
+      result EXPR_DOT, :tCOLON2, text
     end
   end
 
-  def process_curly_brace text
-    self.brace_nest += 1
-    if lpar_beg && lpar_beg == paren_nest then
-      self.lpar_beg = nil
-      self.paren_nest -= 1
-
-      return expr_result(:tLAMBEG, "{")
-    end
+  def process_dots text
+    tokens = ruby27plus? && is_beg? ? BTOKENS : TOKENS
 
-    token = if is_arg? || in_lex_state?(:expr_end, :expr_endfn) then
-               :tLCURLY      #  block (primary)
-             elsif in_lex_state?(:expr_endarg) then
-               :tLBRACE_ARG  #  block (expr)
-             else
-               :tLBRACE      #  hash
-             end
-
-    self.command_start = true unless token == :tLBRACE
-
-    return expr_result(token, "{")
+    result EXPR_BEG, tokens[text], text
   end
 
   def process_float text
     rb_compile_error "Invalid numeric format" if text =~ /__/
 
     case
-    when text.end_with?('ri')
-      return result(:expr_end, :tIMAGINARY, Complex(0, Rational(text.chop.chop)))
-    when text.end_with?('r')
-      return result(:expr_end, :tRATIONAL, Rational(text.chop))
-    when text.end_with?('i')
-      return result(:expr_end, :tIMAGINARY, Complex(0, text.chop.to_f))
+    when text.end_with?("ri")
+      result EXPR_NUM, :tIMAGINARY, Complex(0, Rational(text.chop.chop))
+    when text.end_with?("i")
+      result EXPR_NUM, :tIMAGINARY, Complex(0, text.chop.to_f)
+    when text.end_with?("r")
+      result EXPR_NUM, :tRATIONAL,  Rational(text.chop)
     else
-      return result(:expr_end, :tFLOAT, text.to_f)
+      result EXPR_NUM, :tFLOAT, text.to_f
     end
   end
 
   def process_gvar text
-    text.lineno = self.lineno
-    result(:expr_end, :tGVAR, text)
+    if parser.class.version > 20 && text == "$-" then
+      rb_compile_error "unexpected $undefined"
+    end
+
+    result EXPR_END, :tGVAR, text
   end
 
   def process_gvar_oddity text
-    return result :expr_end, "$", "$" if text == "$" # TODO: wtf is this?
     rb_compile_error "#{text.inspect} is not allowed as a global variable name"
   end
 
   def process_ivar text
     tok_id = text =~ /^@@/ ? :tCVAR : :tIVAR
-    text.lineno = self.lineno
-    return result(:expr_end, tok_id, text)
+    result EXPR_END, tok_id, text
+  end
+
+  def process_label text
+    symbol = possibly_escape_string text, /^\"/
+
+    result EXPR_LAB, :tLABEL, symbol
+  end
+
+  def process_label_or_string text
+    if @was_label && text =~ /:\Z/ then
+      @was_label = nil
+      return process_label text
+    elsif text =~ /:\Z/ then
+      self.pos -= 1 # put back ":"
+      text = text[0..-2]
+    end
+
+    orig_line = lineno
+    str = text[1..-2].gsub(/\\\\/, "\\").gsub(/\\\'/, "\'")
+    self.lineno += str.count("\n")
+
+    result EXPR_END, :tSTRING, str, orig_line
   end
 
   def process_lchevron text
-    if (!in_lex_state?(:expr_dot, :expr_class) &&
+    if (lex_state !~ EXPR_DOT|EXPR_CLASS &&
         !is_end? &&
-        (!is_arg? || space_seen)) then
+        (!is_arg? || lex_state =~ EXPR_LABELED || space_seen)) then
       tok = self.heredoc_identifier
       return tok if tok
     end
 
-    return result(:arg_state, :tLSHFT, "\<\<")
+    if is_after_operator? then
+      self.lex_state = EXPR_ARG
+    else
+      self.command_start = true if lex_state =~ EXPR_CLASS
+      self.lex_state = EXPR_BEG
+    end
+
+    result lex_state, :tLSHFT, "\<\<"
   end
 
-  def process_newline_or_comment text
+  def process_newline_or_comment text    # ../compare/parse30.y:9126 ish
     c = matched
-    hit = false
 
-    if c == '#' then
-      ss.pos -= 1
+    if c == "#" then
+      self.pos -= 1
 
       while scan(/\s*\#.*(\n+|\z)/) do
-        hit = true
-        self.lineno += matched.lines.to_a.size
-        @comments << matched.gsub(/^ +#/, '#').gsub(/^ +$/, '')
+        self.lineno += matched.count "\n"
+        @comments << matched.gsub(/^ +#/, "#").gsub(/^ +$/, "")
       end
 
       return nil if end_of_stream?
     end
 
-    self.lineno += 1 unless hit
-
-    # Replace a string of newlines with a single one
-    self.lineno += matched.lines.to_a.size if scan(/\n+/)
-
-    return if in_lex_state?(:expr_beg, :expr_value, :expr_class,
-                            :expr_fname, :expr_dot)
+    c = (lex_state =~ EXPR_BEG|EXPR_CLASS|EXPR_FNAME|EXPR_DOT &&
+         lex_state !~ EXPR_LABELED)
+    if c || self.lex_state == EXPR_LAB then # yes, == EXPR_LAB
+      # ignore if !fallthrough?
+      if !c && parser.in_kwarg then
+        # normal newline
+        self.command_start = true
+        return result EXPR_BEG, :tNL, nil
+      else
+        maybe_pop_stack
+        return # goto retry
+      end
+    end
 
-    if scan(/([\ \t\r\f\v]*)(\.|&)/) then
-      self.space_seen = true unless ss[1].empty?
+    if scan(/[\ \t\r\f\v]+/) then
+      self.space_seen = true
+    end
 
-      ss.pos -= 1
-      return unless check(/\.\./)
+    if check(/#/) then
+      return # goto retry
+    elsif check(/&\.|\.(?!\.)/) then # C version is a hellish obfuscated xnor
+      return # goto retry
     end
 
     self.command_start = true
 
-    return result(:expr_beg, :tNL, nil)
+    result EXPR_BEG, :tNL, nil
   end
 
   def process_nthref text
     # TODO: can't do lineno hack w/ number
-    result :expr_end, :tNTH_REF, ss[1].to_i
+    result EXPR_END, :tNTH_REF, match[1].to_i
   end
 
   def process_paren text
-    token = if ruby18 then
-              process_paren18
+    token = if is_beg? then
+              :tLPAREN
+            elsif !space_seen then
+              # foo( ... ) => method call, no ambiguity
+              :tLPAREN2
+            elsif is_space_arg? then
+              :tLPAREN_ARG
+            elsif lex_state =~ EXPR_ENDFN && !lambda_beginning? then
+              # TODO:
+              # warn("parentheses after method name is interpreted as " \
+              #      "an argument list, not a decomposed argument")
+              :tLPAREN2
             else
-              process_paren19
+              :tLPAREN2 # plain "(" in parse.y
             end
 
     self.paren_nest += 1
 
-    # TODO: add :expr_label to :expr_beg (set in expr_result below)
-    return expr_result(token, "(")
-  end
-
-  def process_paren18
-    self.command_start = true
-    token = :tLPAREN2
-
-    if in_lex_state? :expr_beg, :expr_mid then
-      token = :tLPAREN
-    elsif space_seen then
-      if in_lex_state? :expr_cmdarg then
-        token = :tLPAREN_ARG
-      elsif in_lex_state? :expr_arg then
-        warning "don't put space before argument parentheses"
-      end
-    else
-      # not a ternary -- do nothing?
-    end
-
-    token
+    cond.push false
+    cmdarg.push false
+    result EXPR_PAR, token, text
   end
 
-  def process_paren19
-    if is_beg? then
-      :tLPAREN
-    elsif is_space_arg? then
-      :tLPAREN_ARG
+  def process_percent text
+    case
+    when is_beg? then
+      process_percent_quote
+    when scan(/\=/)
+      result EXPR_BEG, :tOP_ASGN, "%"
+    when is_space_arg?(check(/\s/)) || (lex_state =~ EXPR_FITEM && check(/s/))
+      process_percent_quote
     else
-      :tLPAREN2 # plain '(' in parse.y
+      result :arg_state, :tPERCENT, "%"
     end
   end
 
-  def process_percent text
-    return parse_quote if is_beg?
-
-    return result(:expr_beg, :tOP_ASGN, "%") if scan(/\=/)
-
-    return parse_quote if is_arg? && space_seen && ! check(/\s/)
-
-    return result(:arg_state, :tPERCENT, "%")
-  end
-
   def process_plus_minus text
     sign = matched
     utype, type = if sign == "+" then
@@ -628,34 +524,33 @@ class RubyLexer
                     [:tUMINUS, :tMINUS]
                   end
 
-    if in_arg_state? then
+    if is_after_operator? then
       if scan(/@/) then
-        return result(:expr_arg, utype, "#{sign}@")
+        return result(EXPR_ARG, utype, "#{sign}@")
       else
-        return result(:expr_arg, type, sign)
+        return result(EXPR_ARG, type, sign)
       end
     end
 
-    return result(:expr_beg, :tOP_ASGN, sign) if scan(/\=/)
+    return result(EXPR_BEG, :tOP_ASGN, sign) if scan(/\=/)
 
-    if (is_beg? || (is_arg? && space_seen && !check(/\s/))) then
+    if is_beg? || (is_arg? && space_seen && !check(/\s/)) then
       arg_ambiguous if is_arg?
 
       if check(/\d/) then
         return nil if utype == :tUPLUS
-        return result(:expr_beg, :tUMINUS_NUM, sign)
+        return result EXPR_BEG, :tUMINUS_NUM, sign
       end
 
-      return result(:expr_beg, utype, sign)
+      return result EXPR_BEG, utype, sign
     end
 
-    return result(:expr_beg, type, sign)
+    result EXPR_BEG, type, sign
   end
 
   def process_questionmark text
     if is_end? then
-      state = ruby18 ? :expr_beg : :expr_value # HACK?
-      return result(state, :tEH, "?")
+      return result EXPR_BEG, :tEH, "?"
     end
 
     if end_of_stream? then
@@ -664,12 +559,12 @@ class RubyLexer
 
     if check(/\s|\v/) then
       unless is_arg? then
-        c2 = { " " => 's',
-              "\n" => 'n',
-              "\t" => 't',
-              "\v" => 'v',
-              "\r" => 'r',
-              "\f" => 'f' }[matched]
+        c2 = { " " => "s",
+              "\n" => "n",
+              "\t" => "t",
+              "\v" => "v",
+              "\r" => "r",
+              "\f" => "f" }[matched]
 
         if c2 then
           warning("invalid character syntax; use ?\\" + c2)
@@ -677,34 +572,40 @@ class RubyLexer
       end
 
       # ternary
-      state = ruby18 ? :expr_beg : :expr_value # HACK?
-      return result(state, :tEH, "?")
+      return result EXPR_BEG, :tEH, "?"
     elsif check(/\w(?=\w)/) then # ternary, also
-      return result(:expr_beg, :tEH, "?")
+      return result EXPR_BEG, :tEH, "?"
     end
 
     c = if scan(/\\/) then
           self.read_escape
         else
-          ss.getch
+          getch
         end
 
-    if version == 18 then
-      return result(:expr_end, :tINTEGER, c[0].ord & 0xff)
-    else
-      return result(:expr_end, :tSTRING, c)
-    end
+    result EXPR_END, :tSTRING, c
+  end
+
+  def process_simple_string text
+    orig_line = lineno
+    self.lineno += text.count("\n")
+
+    str = text[1..-2]
+      .gsub(ESC) { unescape($1).b.force_encoding Encoding::UTF_8 }
+    str = str.b unless str.valid_encoding?
+
+    result EXPR_END, :tSTRING, str, orig_line
   end
 
   def process_slash text
     if is_beg? then
-      string STR_REGEXP
+      string STR_REGEXP, matched
 
-      return result(nil, :tREGEXP_BEG, "/")
+      return result nil, :tREGEXP_BEG, "/"
     end
 
     if scan(/\=/) then
-      return result(:expr_beg, :tOP_ASGN, "/")
+      return result(EXPR_BEG, :tOP_ASGN, "/")
     end
 
     if is_arg? && space_seen then
@@ -715,7 +616,7 @@ class RubyLexer
       end
     end
 
-    return result(:arg_state, :tDIVIDE, "/")
+    result :arg_state, :tDIVIDE, "/"
   end
 
   def process_square_bracket text
@@ -723,72 +624,40 @@ class RubyLexer
 
     token = nil
 
-    if in_arg_state? then
+    if is_after_operator? then
       case
       when scan(/\]\=/) then
         self.paren_nest -= 1 # HACK? I dunno, or bug in MRI
-        return result(:expr_arg, :tASET, "[]=")
+        return result EXPR_ARG, :tASET, "[]="
       when scan(/\]/) then
         self.paren_nest -= 1 # HACK? I dunno, or bug in MRI
-        return result(:expr_arg, :tAREF, "[]")
+        return result EXPR_ARG, :tAREF, "[]"
       else
         rb_compile_error "unexpected '['"
       end
     elsif is_beg? then
       token = :tLBRACK
-    elsif is_arg? && space_seen then
+    elsif is_arg? && (space_seen || lex_state =~ EXPR_LABELED) then
       token = :tLBRACK
     else
       token = :tLBRACK2
     end
 
-    return expr_result(token, "[")
-  end
-
-  def possibly_escape_string text, check
-    content = match[1]
-
-    if text =~ check then
-      content.gsub(ESC) { unescape $1 }
-    else
-      content.gsub(/\\\\/, "\\").gsub(/\\'/, "'")
-    end
+    cond.push false
+    cmdarg.push false
+    result EXPR_PAR, token, text
   end
 
   def process_symbol text
-    symbol = possibly_escape_string text, /^:"/
-
-    rb_compile_error "symbol cannot contain '\\0'" if
-      ruby18 && symbol =~ /\0/
-
-    return result(:expr_end, :tSYMBOL, symbol)
-  end
-
-  def was_label?
-    @was_label = ruby22_label?
-    true
-  end
+    symbol = possibly_escape_string text, /^:\"/ # stupid emacs
 
-  def process_label_or_string text
-    if @was_label && text =~ /:\Z/ then
-      @was_label = nil
-      return process_label text
-    elsif text =~ /:\Z/ then
-      ss.pos -= 1 # put back ":"
-      text = text[0..-2]
-    end
-
-    result :expr_end, :tSTRING, text[1..-2].gsub(/\\\\/, "\\").gsub(/\\'/, "'")
-  end
-
-  def process_label text
-    symbol = possibly_escape_string text, /^"/
-
-    result(:expr_labelarg, :tLABEL, [symbol, self.lineno])
+    result EXPR_LIT, :tSYMBOL, symbol
   end
 
   def process_token text
-    # TODO: make this always return [token, lineno]
+    # matching: parse_ident in compare/parse23.y:7989
+    # FIX: remove: self.last_state = lex_state
+
     token = self.token = text
     token << matched if scan(/[\!\?](?!=)/)
 
@@ -796,7 +665,7 @@ class RubyLexer
       case
       when token =~ /[!?]$/ then
         :tFID
-      when in_lex_state?(:expr_fname) && scan(/=(?:(?![~>=])|(?==>))/) then
+      when lex_state =~ EXPR_FNAME && scan(/=(?:(?![~>=])|(?==>))/) then
         # ident=, not =~ => == or followed by =>
         # TODO test lexing of a=>b vs a==>b
         token << matched
@@ -807,216 +676,133 @@ class RubyLexer
         :tIDENTIFIER
       end
 
-    if !ruby18 and is_label_possible? and is_label_suffix? then
+    if is_label_possible? and is_label_suffix? then
       scan(/:/)
-      return result(:expr_labelarg, :tLABEL, [token, self.lineno])
+      return result EXPR_LAB, :tLABEL, token
     end
 
-    unless in_lex_state? :expr_dot then
+    # TODO: mb == ENC_CODERANGE_7BIT && lex_state !~ EXPR_DOT
+    if lex_state !~ EXPR_DOT then
       # See if it is a reserved word.
-      keyword = if ruby18 then # REFACTOR need 18/19 lexer subclasses
-                  RubyParserStuff::Keyword.keyword18 token
-                else
-                  RubyParserStuff::Keyword.keyword19 token
-                end
+      keyword = RubyParserStuff::Keyword.keyword token
 
       return process_token_keyword keyword if keyword
-    end # unless in_lex_state? :expr_dot
-
-    # TODO:
-    # if (mb == ENC_CODERANGE_7BIT && lex_state != EXPR_DOT) {
+    end
 
-    state = if is_beg? or is_arg? or in_lex_state? :expr_dot then
-              command_state ? :expr_cmdarg : :expr_arg
-            elsif not ruby18 and in_lex_state? :expr_fname then
-              :expr_endfn
+    # matching: compare/parse30.y:9039
+    state = if lex_state =~ EXPR_BEG_ANY|EXPR_ARG_ANY|EXPR_DOT then
+              cmd_state ? EXPR_CMDARG : EXPR_ARG
+            elsif lex_state =~ EXPR_FNAME then
+              EXPR_ENDFN
             else
-              :expr_end
+              EXPR_END
             end
+    self.lex_state = state
 
-    if not [:expr_dot, :expr_fname].include? last_state and
-        self.parser.env[token.to_sym] == :lvar then
-      state = :expr_end
-    end
+    tok_id = :tIDENTIFIER if tok_id == :tCONSTANT && is_local_id(token)
 
-    token.lineno = self.lineno # yes, on a string. I know... I know...
+    if last_state !~ EXPR_DOT|EXPR_FNAME and
+        (tok_id == :tIDENTIFIER) and # not EXPR_FNAME, not attrasgn
+        lvar_defined?(token) then
+      state = EXPR_END|EXPR_LABEL
+    end
 
-    return result(state, tok_id, token)
+    result state, tok_id, token
   end
 
   def process_token_keyword keyword
-    state = keyword.state
+    # matching MIDDLE of parse_ident in compare/parse23.y:8046
+    state = lex_state
 
-    value = [token, self.lineno]
+    return result(EXPR_ENDFN, keyword.id0, token) if lex_state =~ EXPR_FNAME
 
-    self.command_start = true if state == :expr_beg and lex_state != :expr_fname
+    self.lex_state = keyword.state
+    self.command_start = true if lex_state =~ EXPR_BEG
 
     case
-    when lex_state == :expr_fname then
-      result(state, keyword.id0, keyword.name)
-    when keyword.id0 == :kDO then
+    when keyword.id0 == :kDO then # parse26.y line 7591
       case
-      when lpar_beg && lpar_beg == paren_nest then
-        self.lpar_beg = nil
-        self.paren_nest -= 1
-        expr_result(:kDO_LAMBDA, value)
+      when lambda_beginning? then
+        self.lpar_beg = nil # lambda_beginning? == FALSE in the body of "-> do ... end"
+        self.paren_nest -= 1 # TODO: question this?
+        result lex_state, :kDO_LAMBDA, token
       when cond.is_in_state then
-        result(state, :kDO_COND, value)
-      when cmdarg.is_in_state && lex_state != :expr_cmdarg then
-        result(state, :kDO_BLOCK, value)
-      when in_lex_state?(:expr_beg, :expr_endarg) then
-        result(state, :kDO_BLOCK, value)
-      when lex_state == :expr_end # eg: a -> do end do end
-        result(state, :kDO_BLOCK, value)
+        result lex_state, :kDO_COND, token
+      when cmdarg.is_in_state && state != EXPR_CMDARG then
+        result lex_state, :kDO_BLOCK, token
       else
-        result(state, :kDO, value)
+        result lex_state, :kDO, token
       end
-    when in_lex_state?(:expr_beg, :expr_value, :expr_labelarg) then
-      result(state, keyword.id0, value)
+    when state =~ EXPR_PAD then
+      result lex_state, keyword.id0, token
     when keyword.id0 != keyword.id1 then
-      result(:expr_beg, keyword.id1, value)
+      result EXPR_PAR, keyword.id1, token
     else
-      result(state, keyword.id1, value)
+      result lex_state, keyword.id1, token
     end
   end
 
   def process_underscore text
-    ss.unscan # put back "_"
+    self.unscan # put back "_"
 
     if beginning_of_line? && scan(/\__END__(\r?\n|\Z)/) then
-      return [RubyLexer::EOF, RubyLexer::EOF]
-    elsif scan(/\_\w*/) then
-      return process_token matched
+      ss.terminate
+      [RubyLexer::EOF, RubyLexer::EOF]
+    elsif scan(/#{IDENT_CHAR}+/) then
+      process_token matched
     end
   end
 
   def rb_compile_error msg
-    msg += ". near line #{self.lineno}: #{ss.rest[/^.*/].inspect}"
+    msg += ". near line #{self.lineno}: #{self.rest[/^.*/].inspect}"
     raise RubyParser::SyntaxError, msg
   end
 
-  def read_escape # TODO: remove / rewrite
-    case
-    when scan(/\\/) then                  # Backslash
-      '\\'
-    when scan(/n/) then                   # newline
-      self.extra_lineno -= 1
-      "\n"
-    when scan(/t/) then                   # horizontal tab
-      "\t"
-    when scan(/r/) then                   # carriage-return
-      "\r"
-    when scan(/f/) then                   # form-feed
-      "\f"
-    when scan(/v/) then                   # vertical tab
-      "\13"
-    when scan(/a/) then                   # alarm(bell)
-      "\007"
-    when scan(/e/) then                   # escape
-      "\033"
-    when scan(/b/) then                   # backspace
-      "\010"
-    when scan(/s/) then                   # space
-      " "
-    when scan(/[0-7]{1,3}/) then          # octal constant
-      (matched.to_i(8) & 0xFF).chr
-    when scan(/x([0-9a-fA-F]{1,2})/) then # hex constant
-      ss[1].to_i(16).chr
-    when check(/M-\\[\\MCc]/) then
-      scan(/M-\\/) # eat it
-      c = self.read_escape
-      c[0] = (c[0].ord | 0x80).chr
-      c
-    when scan(/M-(.)/) then
-      c = ss[1]
-      c[0] = (c[0].ord | 0x80).chr
-      c
-    when check(/(C-|c)\\[\\MCc]/) then
-      scan(/(C-|c)\\/) # eat it
-      c = self.read_escape
-      c[0] = (c[0].ord & 0x9f).chr
-      c
-    when scan(/C-\?|c\?/) then
-      127.chr
-    when scan(/(C-|c)(.)/) then
-      c = ss[2]
-      c[0] = (c[0].ord & 0x9f).chr
-      c
-    when scan(/^[89]/i) then # bad octal or hex... MRI ignores them :(
-      matched
-    when scan(/u([0-9a-fA-F]{2,4}|\{[0-9a-fA-F]{2,6}\})/) then
-      [ss[1].delete("{}").to_i(16)].pack("U")
-    when scan(/[McCx0-9]/) || end_of_stream? then
-      rb_compile_error("Invalid escape character syntax")
-    else
-      ss.getch
-    end.dup
-  end
-
-  def regx_options # TODO: rewrite / remove
-    good, bad = [], []
-
-    if scan(/[a-z]+/) then
-      good, bad = matched.split(//).partition { |s| s =~ /^[ixmonesu]$/ }
-    end
-
-    unless bad.empty? then
-      rb_compile_error("unknown regexp option%s - %s" %
-                       [(bad.size > 1 ? "s" : ""), bad.join.inspect])
-    end
-
-    return good.join
-  end
-
   def reset
+    self.lineno        = 1
     self.brace_nest    = 0
     self.command_start = true
     self.comments      = []
-    self.lex_state     = :expr_none
+    self.lex_state     = EXPR_NONE
     self.lex_strterm   = nil
-    self.lineno        = 1
     self.lpar_beg      = nil
     self.paren_nest    = 0
     self.space_seen    = false
     self.string_nest   = 0
     self.token         = nil
-    self.extra_lineno  = 0
+    self.string_buffer = []
+    self.old_ss        = nil
+    self.old_lineno    = nil
 
-    self.cmdarg.reset
     self.cond.reset
+    self.cmdarg.reset
   end
 
-  def result lex_state, token, text # :nodoc:
-    lex_state = self.arg_state if lex_state == :arg_state
-    self.lex_state = lex_state if lex_state
-    [token, text]
-  end
+  def result new_state, token, text, line = self.lineno # :nodoc:
+    new_state = self.arg_state if new_state == :arg_state
+    self.lex_state = new_state if new_state
 
-  def ruby18
-    RubyParser::V18 === parser
+    [token, [text, line]]
   end
 
-  def scan re
-    ss.scan re
+  def ruby22_label?
+    ruby22plus? and is_label_possible?
   end
 
-  def check re
-    ss.check re
+  def ruby22plus?
+    parser.class.version >= 22
   end
 
-  def eat_whitespace
-    r = scan(/\s+/)
-    self.extra_lineno += r.count("\n") if r
-    r
+  def ruby23plus?
+    parser.class.version >= 23
   end
 
-  def fixup_lineno extra = 0
-    self.lineno += self.extra_lineno + extra
-    self.extra_lineno = 0
+  def ruby24minus?
+    parser.class.version <= 24
   end
 
-  def scanner_class # TODO: design this out of oedipus_lex. or something.
-    RPStringScanner
+  def ruby27plus?
+    parser.class.version >= 27
   end
 
   def space_vs_beginning space_type, beg_type, fallback
@@ -1031,139 +817,9 @@ class RubyLexer
     end
   end
 
-  def string type, beg = matched, nnd = "\0"
-    self.lex_strterm = [:strterm, type, beg, nnd]
-  end
-
-  # TODO: consider
-  # def src= src
-  #   raise "bad src: #{src.inspect}" unless String === src
-  #   @src = RPStringScanner.new(src)
-  # end
-
-  def tokadd_escape term # TODO: rewrite / remove
-    case
-    when scan(/\\\n/) then
-      # just ignore
-    when scan(/\\([0-7]{1,3}|x[0-9a-fA-F]{1,2})/) then
-      self.string_buffer << matched
-    when scan(/\\([MC]-|c)(?=\\)/) then
-      self.string_buffer << matched
-      self.tokadd_escape term
-    when scan(/\\([MC]-|c)(.)/) then
-      self.string_buffer << matched
-    when scan(/\\[McCx]/) then
-      rb_compile_error "Invalid escape character syntax"
-    when scan(/\\(.)/m) then
-      chr = ss[1]
-      prev = self.string_buffer.last
-      if term == chr && prev && prev.end_with?("(?") then
-        self.string_buffer << chr
-      else
-        self.string_buffer << matched
-      end
-    else
-      rb_compile_error "Invalid escape character syntax"
-    end
-  end
-
-  def tokadd_string(func, term, paren) # TODO: rewrite / remove
-    qwords = (func & STR_FUNC_QWORDS) != 0
-    escape = (func & STR_FUNC_ESCAPE) != 0
-    expand = (func & STR_FUNC_EXPAND) != 0
-    regexp = (func & STR_FUNC_REGEXP) != 0
-    symbol = (func & STR_FUNC_SYMBOL) != 0
-
-    paren_re = @@regexp_cache[paren]
-    term_re  = @@regexp_cache[term]
-
-    until end_of_stream? do
-      c = nil
-      handled = true
-
-      case
-      when paren_re && scan(paren_re) then
-        self.string_nest += 1
-      when scan(term_re) then
-        if self.string_nest == 0 then
-          ss.pos -= 1
-          break
-        else
-          self.string_nest -= 1
-        end
-      when expand && scan(/#(?=[\$\@\{])/) then
-        ss.pos -= 1
-        break
-      when qwords && scan(/\s/) then
-        ss.pos -= 1
-        break
-      when expand && scan(/#(?!\n)/) then
-        # do nothing
-      when check(/\\/) then
-        case
-        when qwords && scan(/\\\n/) then
-          string_buffer << "\n"
-          next
-        when qwords && scan(/\\\s/) then
-          c = ' '
-        when expand && scan(/\\\n/) then
-          next
-        when regexp && check(/\\/) then
-          self.tokadd_escape term
-          next
-        when expand && scan(/\\/) then
-          c = self.read_escape
-        when scan(/\\\n/) then
-          # do nothing
-        when scan(/\\\\/) then
-          string_buffer << '\\' if escape
-          c = '\\'
-        when scan(/\\/) then
-          unless scan(term_re) || paren.nil? || scan(paren_re) then
-            string_buffer << "\\"
-          end
-        else
-          handled = false
-        end # inner /\\/ case
-      else
-        handled = false
-      end # top case
-
-      unless handled then
-        t = Regexp.escape term
-        x = Regexp.escape(paren) if paren && paren != "\000"
-        re = if qwords then
-               if HAS_ENC then
-                 /[^#{t}#{x}\#\0\\\s]+|./ # |. to pick up whatever
-               else
-                 /[^#{t}#{x}\#\0\\\s\v]+|./ # argh. 1.8's \s doesn't pick up \v
-               end
-             else
-               /[^#{t}#{x}\#\0\\]+|./
-             end
-
-        scan re
-        c = matched
-
-        rb_compile_error "symbol cannot contain '\\0'" if symbol && c =~ /\0/
-      end # unless handled
-
-      c ||= matched
-      string_buffer << c
-    end # until
-
-    c ||= matched
-    c = RubyLexer::EOF if end_of_stream?
-
-    return c
-  end
-
   def unescape s
     r = ESCAPES[s]
 
-    self.extra_lineno += 1 if s == "\n"     # eg backslash newline strings
-    self.extra_lineno -= 1 if r && s == "n" # literal \n, not newline
-
     return r if r
 
     x = case s
@@ -1179,12 +835,15 @@ class RubyLexer
           s
         when /^[McCx0-9]/ then
           rb_compile_error("Invalid escape character syntax")
-        when /u([0-9a-fA-F]{2,4}|\{[0-9a-fA-F]{2,6}\})/ then
+        when /u(\h{4})/ then
           [$1.delete("{}").to_i(16)].pack("U")
+        when /u(\h{1,3})/ then
+          rb_compile_error("Invalid escape character syntax")
+        when /u\{(\h+(?:\s+\h+)*)\}/ then
+          $1.split.map { |cp| cp.to_i(16) }.pack("U*")
         else
           s
         end
-    x.force_encoding "UTF-8" if HAS_ENC
     x
   end
 
@@ -1192,168 +851,294 @@ class RubyLexer
     # do nothing for now
   end
 
-  def ruby22plus?
-    parser.class.version >= 22
+  def was_label?
+    @was_label = ruby22_label?
+    true
   end
 
-  def ruby23plus?
-    parser.class.version >= 23
-  end
+  class State
+    attr_accessor :n
+    attr_accessor :names
 
-  def process_string # TODO: rewrite / remove
-    token = if lex_strterm[0] == :heredoc then
-              self.heredoc lex_strterm
-            else
-              self.parse_string lex_strterm
-            end
+    # TODO: take a shared hash of strings for inspect/to_s
+    def initialize o, names
+      raise ArgumentError, "bad state: %p" % [o] unless Integer === o # TODO: remove
 
-    token_type, c = token
+      self.n = o
+      self.names = names
+    end
 
-    if ruby22plus? && token_type == :tSTRING_END && ["'", '"'].include?(c) then
-      if (([:expr_beg, :expr_endfn].include?(lex_state) &&
-           !cond.is_in_state) || is_arg?) &&
-          is_label_suffix? then
-        scan(/:/)
-        token_type = token[0] = :tLABEL_END
-      end
+    def == o
+      self.equal?(o) || (o.class == self.class && o.n == self.n)
     end
 
-    if [:tSTRING_END, :tREGEXP_END, :tLABEL_END].include? token_type then
-      self.lex_strterm = nil
-      self.lex_state   = (token_type == :tLABEL_END) ? :expr_labelarg : :expr_end
+    def =~ v
+      (self.n & v.n) != 0
     end
 
-    return token
-  end
+    def | v
+      raise ArgumentError, "Incompatible State: %p vs %p" % [self, v] unless
+        self.names == v.names
+      self.class.new(self.n | v.n, self.names)
+    end
 
-  def parse_quote # TODO: remove / rewrite
-    beg, nnd, short_hand, c = nil, nil, false, nil
+    def inspect
+      return "Value(0)" if n.zero? # HACK?
 
-    if scan(/[a-z0-9]{1,2}/i) then # Long-hand (e.g. %Q{}).
-      rb_compile_error "unknown type of %string" if ss.matched_size == 2
-      c, beg, short_hand = matched, ss.getch, false
-    else                               # Short-hand (e.g. %{, %., %!, etc)
-      c, beg, short_hand = 'Q', ss.getch, true
+      names.map { |v, k| k if self =~ v }.
+        compact.
+        join("|").
+        gsub(/(?:EXPR_|STR_(?:FUNC_)?)/, "")
     end
 
-    if end_of_stream? or c == RubyLexer::EOF or beg == RubyLexer::EOF then
-      rb_compile_error "unterminated quoted string meets end of file"
+    alias to_s inspect
+
+    module Values
+      expr_names = {}
+
+      EXPR_NONE    = State.new    0x0, expr_names
+      EXPR_BEG     = State.new    0x1, expr_names
+      EXPR_END     = State.new    0x2, expr_names
+      EXPR_ENDARG  = State.new    0x4, expr_names
+      EXPR_ENDFN   = State.new    0x8, expr_names
+      EXPR_ARG     = State.new   0x10, expr_names
+      EXPR_CMDARG  = State.new   0x20, expr_names
+      EXPR_MID     = State.new   0x40, expr_names
+      EXPR_FNAME   = State.new   0x80, expr_names
+      EXPR_DOT     = State.new  0x100, expr_names
+      EXPR_CLASS   = State.new  0x200, expr_names
+      EXPR_LABEL   = State.new  0x400, expr_names
+      EXPR_LABELED = State.new  0x800, expr_names
+      EXPR_FITEM   = State.new 0x1000, expr_names
+
+      EXPR_BEG_ANY = EXPR_BEG | EXPR_MID    | EXPR_CLASS
+      EXPR_ARG_ANY = EXPR_ARG | EXPR_CMDARG
+      EXPR_END_ANY = EXPR_END | EXPR_ENDARG | EXPR_ENDFN
+
+      # extra fake lex_state names to make things a bit cleaner
+
+      EXPR_LAB = EXPR_ARG|EXPR_LABELED
+      EXPR_LIT = EXPR_END|EXPR_ENDARG
+      EXPR_PAR = EXPR_BEG|EXPR_LABEL
+      EXPR_PAD = EXPR_BEG|EXPR_LABELED
+
+      EXPR_NUM = EXPR_LIT
+
+      expr_names.merge!(EXPR_NONE    => "EXPR_NONE",
+                        EXPR_BEG     => "EXPR_BEG",
+                        EXPR_END     => "EXPR_END",
+                        EXPR_ENDARG  => "EXPR_ENDARG",
+                        EXPR_ENDFN   => "EXPR_ENDFN",
+                        EXPR_ARG     => "EXPR_ARG",
+                        EXPR_CMDARG  => "EXPR_CMDARG",
+                        EXPR_MID     => "EXPR_MID",
+                        EXPR_FNAME   => "EXPR_FNAME",
+                        EXPR_DOT     => "EXPR_DOT",
+                        EXPR_CLASS   => "EXPR_CLASS",
+                        EXPR_LABEL   => "EXPR_LABEL",
+                        EXPR_LABELED => "EXPR_LABELED",
+                        EXPR_FITEM   => "EXPR_FITEM")
+
+      # ruby constants for strings
+
+      str_func_names = {}
+
+      STR_FUNC_BORING = State.new 0x00,    str_func_names
+      STR_FUNC_ESCAPE = State.new 0x01,    str_func_names
+      STR_FUNC_EXPAND = State.new 0x02,    str_func_names
+      STR_FUNC_REGEXP = State.new 0x04,    str_func_names
+      STR_FUNC_QWORDS = State.new 0x08,    str_func_names
+      STR_FUNC_SYMBOL = State.new 0x10,    str_func_names
+      STR_FUNC_INDENT = State.new 0x20,    str_func_names # <<-HEREDOC
+      STR_FUNC_LABEL  = State.new 0x40,    str_func_names
+      STR_FUNC_LIST   = State.new 0x4000,  str_func_names
+      STR_FUNC_TERM   = State.new 0x8000,  str_func_names
+      STR_FUNC_DEDENT = State.new 0x10000, str_func_names # <<~HEREDOC
+
+      # TODO: check parser25.y on how they do STR_FUNC_INDENT
+
+      STR_SQUOTE = STR_FUNC_BORING
+      STR_DQUOTE = STR_FUNC_EXPAND
+      STR_XQUOTE = STR_FUNC_EXPAND
+      STR_REGEXP = STR_FUNC_REGEXP | STR_FUNC_ESCAPE | STR_FUNC_EXPAND
+      STR_SWORD  = STR_FUNC_QWORDS | STR_FUNC_LIST
+      STR_DWORD  = STR_FUNC_QWORDS | STR_FUNC_EXPAND | STR_FUNC_LIST
+      STR_SSYM   = STR_FUNC_SYMBOL
+      STR_DSYM   = STR_FUNC_SYMBOL | STR_FUNC_EXPAND
+      STR_LABEL  = STR_FUNC_LABEL
+
+      str_func_names.merge!(STR_FUNC_ESCAPE => "STR_FUNC_ESCAPE",
+                            STR_FUNC_EXPAND => "STR_FUNC_EXPAND",
+                            STR_FUNC_REGEXP => "STR_FUNC_REGEXP",
+                            STR_FUNC_QWORDS => "STR_FUNC_QWORDS",
+                            STR_FUNC_SYMBOL => "STR_FUNC_SYMBOL",
+                            STR_FUNC_INDENT => "STR_FUNC_INDENT",
+                            STR_FUNC_LABEL  => "STR_FUNC_LABEL",
+                            STR_FUNC_LIST   => "STR_FUNC_LIST",
+                            STR_FUNC_TERM   => "STR_FUNC_TERM",
+                            STR_FUNC_DEDENT => "STR_FUNC_DEDENT",
+                            STR_SQUOTE      => "STR_SQUOTE")
     end
 
-    # Figure nnd-char.  "\0" is special to indicate beg=nnd and that no nesting?
-    nnd = { "(" => ")", "[" => "]", "{" => "}", "<" => ">" }[beg]
-    nnd, beg = beg, "\0" if nnd.nil?
+    include Values
+  end
 
-    token_type, text = nil, "%#{c}#{beg}"
-    token_type, string_type = case c
-                              when 'Q' then
-                                ch = short_hand ? nnd : c + beg
-                                text = "%#{ch}"
-                                [:tSTRING_BEG,   STR_DQUOTE]
-                              when 'q' then
-                                [:tSTRING_BEG,   STR_SQUOTE]
-                              when 'W' then
-                                eat_whitespace
-                                [:tWORDS_BEG,    STR_DQUOTE | STR_FUNC_QWORDS]
-                              when 'w' then
-                                eat_whitespace
-                                [:tQWORDS_BEG,   STR_SQUOTE | STR_FUNC_QWORDS]
-                              when 'x' then
-                                [:tXSTRING_BEG,  STR_XQUOTE]
-                              when 'r' then
-                                [:tREGEXP_BEG,   STR_REGEXP]
-                              when 's' then
-                                self.lex_state  = :expr_fname
-                                [:tSYMBEG,       STR_SSYM]
-                              when 'I' then
-                                eat_whitespace
-                                [:tSYMBOLS_BEG, STR_DQUOTE | STR_FUNC_QWORDS]
-                              when 'i' then
-                                eat_whitespace
-                                [:tQSYMBOLS_BEG, STR_SQUOTE | STR_FUNC_QWORDS]
-                              end
+  include State::Values
+end
 
-    rb_compile_error "Bad %string type. Expected [QqWwIixrs], found '#{c}'." if
-      token_type.nil?
+class RubyLexer
+  module SSWrapper
+    def string= s
+      ss.string= s
+    end
 
-    raise "huh" unless string_type
+    def beginning_of_line?
+      ss.bol?
+    end
 
-    string string_type, nnd, beg
+    alias bol? beginning_of_line? # to make .rex file more readable
 
-    return token_type, text
-  end
+    def check re
+      maybe_pop_stack
 
-  def parse_string quote # TODO: rewrite / remove
-    _, string_type, term, open = quote
+      ss.check re
+    end
 
-    space = false # FIX: remove these
-    func = string_type
-    paren = open
-    term_re = @@regexp_cache[term]
+    def end_of_stream?
+      ss.eos?
+    end
 
-    qwords = (func & STR_FUNC_QWORDS) != 0
-    regexp = (func & STR_FUNC_REGEXP) != 0
-    expand = (func & STR_FUNC_EXPAND) != 0
+    alias eos? end_of_stream?
 
-    unless func then # nil'ed from qwords below. *sigh*
-      return :tSTRING_END, nil
+    def getch
+      c = ss.getch
+      c = ss.getch if c == "\r" && ss.peek(1) == "\n"
+      c
     end
 
-    space = true if qwords and eat_whitespace
+    def match
+      ss
+    end
 
-    if self.string_nest == 0 && scan(/#{term_re}/) then
-      if qwords then
-        quote[1] = nil
-        return :tSPACE, nil
-      elsif regexp then
-        return :tREGEXP_END, self.regx_options
-      else
-        return :tSTRING_END, term
+    def matched
+      ss.matched
+    end
+
+    def in_heredoc?
+      !!self.old_ss
+    end
+
+    def maybe_pop_stack
+      if ss.eos? && in_heredoc? then
+        self.ss_pop
+        self.lineno_pop
       end
     end
 
-    return :tSPACE, nil if space
+    def pos
+      ss.pos
+    end
 
-    self.string_buffer = []
+    def pos= n
+      ss.pos = n
+    end
 
-    if expand
-      case
-      when scan(/#(?=\$(-.|[a-zA-Z_0-9~\*\$\?!@\/\\;,\.=:<>\"\&\`\'+]))/) then
-        # TODO: !ISASCII
-        # ?! see parser_peek_variable_name
-        return :tSTRING_DVAR, nil
-      when scan(/#(?=\@\@?[a-zA-Z_])/) then
-        # TODO: !ISASCII
-        return :tSTRING_DVAR, nil
-      when scan(/#[{]/) then
-        return :tSTRING_DBEG, nil
-      when scan(/#/) then
-        string_buffer << '#'
-      end
+    def rest
+      ss.rest
+    end
+
+    def scan re
+      maybe_pop_stack
+
+      ss.scan re
     end
 
-    if tokadd_string(func, term, paren) == RubyLexer::EOF then
-      rb_compile_error "unterminated string meets end of file"
+    def scanner_class # TODO: design this out of oedipus_lex. or something.
+      RPStringScanner
     end
 
-    return :tSTRING_CONTENT, string_buffer.join
+    def ss_string
+      ss.string
+    end
+
+    def ss_string= s
+      raise "Probably not"
+      ss.string = s
+    end
+
+    def unscan
+      ss.unscan
+    end
   end
+
+  include SSWrapper
 end
 
-require "ruby_lexer.rex"
+class RubyLexer
+  module SSStackish
+    def lineno_push new_lineno
+      self.old_lineno = self.lineno
+      self.lineno     = new_lineno
+    end
 
-if ENV["RP_LINENO_DEBUG"] then
+    def lineno_pop
+      self.lineno     = self.old_lineno
+      self.old_lineno = nil
+    end
+
+    def ss= o
+      raise "Clearing ss while in heredoc!?!" if in_heredoc?
+      @old_ss = nil
+      super
+    end
+
+    def ss_push new_ss
+      @old_ss = self.ss
+      @ss     = new_ss
+    end
+
+    def ss_pop
+      @ss     = self.old_ss
+      @old_ss = nil
+    end
+  end
+
+  prepend SSStackish
+end
+
+if ENV["RP_STRTERM_DEBUG"] then
   class RubyLexer
-    alias :old_lineno= :lineno=
+    def d o
+      $stderr.puts o.inspect
+    end
+
+    alias old_lex_strterm= lex_strterm=
 
+    def lex_strterm= o
+      self.old_lex_strterm= o
+      where = caller.first.split(/:/).first(2).join(":")
+      $stderr.puts
+      d :lex_strterm => [o, where]
+    end
+  end
+end
+
+require_relative "./ruby_lexer.rex.rb"
+require_relative "./ruby_lexer_strings.rb"
+
+if ENV["RP_LINENO_DEBUG"] then
+  class RubyLexer
     def d o
       $stderr.puts o.inspect
     end
 
+    alias old_lineno= lineno=
+
     def lineno= n
       self.old_lineno= n
       where = caller.first.split(/:/).first(2).join(":")
-      d :lineno => [n, where, ss && ss.rest[0,40]]
+      $stderr.puts
+      d :lineno => [n, where]
     end
   end
 end