ruby_parser 3.13.0 → 3.15.0

data/lib/ruby_lexer.rb

@@ -4,40 +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
-
-  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]
-
   ESCAPES = {
     "a"    => "\007",
     "b"    => "\010",
@@ -54,6 +23,8 @@ class RubyLexer
     "c\?"  => 127.chr,
   }
 
+  HAS_ENC = "".respond_to? :encoding
+
   TOKENS = {
     "!"   => :tBANG,
     "!="  => :tNEQ,
@@ -70,13 +41,26 @@ class RubyLexer
     "->"  => :tLAMBDA,
   }
 
-  TAB_WIDTH = 8
-
-  @@regexp_cache = Hash.new { |h,k| h[k] = Regexp.new(Regexp.escape(k)) }
+  @@regexp_cache = Hash.new { |h, k| h[k] = Regexp.new(Regexp.escape(k)) }
   @@regexp_cache[nil] = nil
 
+  if $DEBUG then
+    attr_reader :lex_state
+
+    def lex_state= o
+      return if @lex_state == o
+      raise ArgumentError, "bad state: %p" % [o] unless State === o
+
+      warn "lex_state: %p -> %p" % [lex_state, o]
+
+      @lex_state = o
+    end
+  end
+
   # :startdoc:
 
+  attr_accessor :lex_state unless $DEBUG
+
   attr_accessor :lineno # we're bypassing oedipus' lineno handling.
   attr_accessor :brace_nest
   attr_accessor :cmdarg
@@ -90,7 +74,6 @@ class RubyLexer
   # 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
@@ -99,24 +82,14 @@ 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
 
   attr_writer :comments
 
   def initialize _ = nil
-    @lex_state = :expr_none
+    @lex_state = nil # remove one warning under $DEBUG
+    self.lex_state = EXPR_NONE
 
     self.cond   = RubyParserStuff::StackState.new(:cond, $DEBUG)
     self.cmdarg = RubyParserStuff::StackState.new(:cmdarg, $DEBUG)
@@ -125,17 +98,22 @@ class RubyLexer
   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?
   end
-  alias :bol? :beginning_of_line? # to make .rex file more readable
+
+  alias bol? beginning_of_line? # to make .rex file more readable
+
+  def check re
+    ss.check re
+  end
 
   def comments # TODO: remove this... maybe comment_string + attr_accessor
     c = @comments.join
@@ -143,30 +121,41 @@ class RubyLexer
     c
   end
 
+  def eat_whitespace
+    r = scan(/\s+/)
+    self.extra_lineno += r.count("\n") if r
+    r
+  end
+
   def end_of_stream?
     ss.eos?
   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
+    result EXPR_BEG, token, text
+  end
+
+  def fixup_lineno extra = 0
+    self.lineno += self.extra_lineno + extra
+    self.extra_lineno = 0
   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
+    indent         = func =~ STR_FUNC_INDENT ? "[ \t]*" : nil
+    expand         = func =~ STR_FUNC_EXPAND
+    eol            = last_line && last_line.end_with?("\r\n") ? "\r\n" : "\n"
     eos_re         = /#{indent}#{Regexp.escape eos}(\r*\n|\z)/
     err_msg        = "can't match #{eos_re.inspect} anywhere in "
 
@@ -175,30 +164,35 @@ class RubyLexer
     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
+      return :tSTRING_END, [eos, func] # TODO: calculate squiggle width at lex?
     end
 
     self.string_buffer = []
 
     if expand then
       case
-      when scan(/#[$@]/) then
-        ss.pos -= 1 # FIX omg stupid
+      when scan(/#(?=\$(-.|[a-zA-Z_0-9~\*\$\?!@\/\\;,\.=:<>\"\&\`\'+]))/) then
+        # TODO: !ISASCII
+        # ?! see parser_peek_variable_name
+        return :tSTRING_DVAR, matched
+      when scan(/#(?=\@\@?[a-zA-Z_])/) then
+        # TODO: !ISASCII
         return :tSTRING_DVAR, matched
       when scan(/#[{]/) then
+        self.command_start = true
         return :tSTRING_DBEG, matched
       when scan(/#/) then
-        string_buffer << '#'
+        string_buffer << "#"
       end
 
       begin
-        c = tokadd_string func, "\n", nil
+        c = tokadd_string func, eol, nil
 
         rb_compile_error err_msg if
           c == RubyLexer::EOF
 
-        if c != "\n" then
-          return :tSTRING_CONTENT, string_buffer.join.delete("\r")
+        if c != eol then
+          return :tSTRING_CONTENT, string_buffer.join
         else
           string_buffer << scan(/\n/)
         end
@@ -214,64 +208,26 @@ class RubyLexer
 
     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?
+    string_content = begin
+                       s = string_buffer.join
+                       s.b.force_encoding Encoding::UTF_8
+                     end
 
     return :tSTRING_CONTENT, string_content
   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 = "-"
     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 |= STR_FUNC_INDENT unless ss[1].empty? # TODO: this seems wrong
+      func |= STR_FUNC_ICNTNT if ss[1] == "~"
       func |= case term
               when "\'" then
                 STR_SQUOTE
@@ -288,7 +244,7 @@ class RubyLexer
       func |= STR_DQUOTE
       unless ss[1].empty? then
         func |= STR_FUNC_INDENT
-        func |= STR_FUNC_ICNTNT if ss[1] == '~'
+        func |= STR_FUNC_ICNTNT if ss[1] == "~"
       end
       string_buffer << ss[2]
     else
@@ -304,23 +260,15 @@ class RubyLexer
 
     self.lex_strterm = [:heredoc, string_buffer.join, func, line]
 
-    if term == '`' then
+    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
@@ -328,42 +276,35 @@ 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")
+      return result(EXPR_NUM, :tIMAGINARY, Complex(0, Rational(text.chop.chop.to_i(base))))
+    when text.end_with?("r")
+      return result(EXPR_NUM, :tRATIONAL, Rational(text.chop.to_i(base)))
+    when text.end_with?("i")
+      return result(EXPR_NUM, :tIMAGINARY, Complex(0, text.chop.to_i(base)))
     else
-      return result(:expr_end, :tINTEGER, text.to_i(base))
+      return 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_ANY)
+    lex_state =~ EXPR_ARG_ANY
   end
 
   def is_beg?
-    # TODO: in_lex_state?(*EXPR_BEG_ANY) || lex_state == [:expr_arg, :expr_labeled]
-    in_lex_state?(*EXPR_BEG_ANY, :expr_value, :expr_labeled)
+    lex_state =~ EXPR_BEG_ANY || lex_state == EXPR_LAB # yes, == EXPR_LAB
   end
 
   def is_end?
-    in_lex_state?(*EXPR_END_ANY)
-  end
-
-  def lvar_defined? id
-    # TODO: (dyna_in_block? && dvar_defined?(id)) || local_id?(id)
-    self.parser.env[id.to_sym] == :lvar
-  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) && !cmd_state) || is_arg?
+    (lex_state =~ EXPR_LABEL|EXPR_ENDFN && !cmd_state) || is_arg?
   end
 
   def is_label_suffix?
@@ -378,6 +319,16 @@ class RubyLexer
     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 matched
     ss.matched
   end
@@ -386,11 +337,139 @@ class RubyLexer
     not is_end?
   end
 
+  def parse_quote # TODO: remove / rewrite
+    beg, nnd, short_hand, c = nil, nil, false, nil
+
+    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, getch, false
+    else                               # Short-hand (e.g. %{, %., %!, etc)
+      c, beg, short_hand = "Q", getch, true
+    end
+
+    if end_of_stream? or c == RubyLexer::EOF or beg == RubyLexer::EOF then
+      rb_compile_error "unterminated quoted string meets end of file"
+    end
+
+    # Figure nnd-char.  "\0" is special to indicate beg=nnd and that no nesting?
+    nnd = { "(" => ")", "[" => "]", "{" => "}", "<" => ">" }[beg]
+    nnd, beg = beg, "\0" if nnd.nil?
+
+    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
+
+    rb_compile_error "Bad %string type. Expected [QqWwIixrs], found '#{c}'." if
+      token_type.nil?
+
+    raise "huh" unless string_type
+
+    string string_type, nnd, beg
+
+    return token_type, text
+  end
+
+  def parse_string quote # TODO: rewrite / remove
+    _, string_type, term, open = quote
+
+    space = false # FIX: remove these
+    func = string_type
+    paren = open
+    term_re = @@regexp_cache[term]
+
+    qwords = func =~ STR_FUNC_QWORDS
+    regexp = func =~ STR_FUNC_REGEXP
+    expand = func =~ STR_FUNC_EXPAND
+
+    unless func then # nil'ed from qwords below. *sigh*
+      return :tSTRING_END, nil
+    end
+
+    space = true if qwords and eat_whitespace
+
+    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
+      end
+    end
+
+    return :tSPACE, nil if space
+
+    self.string_buffer = []
+
+    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
+        self.command_start = true
+        return :tSTRING_DBEG, nil
+      when scan(/#/) then
+        string_buffer << "#"
+      end
+    end
+
+    if tokadd_string(func, term, paren) == RubyLexer::EOF then
+      if func =~ STR_FUNC_REGEXP then
+        rb_compile_error "unterminated regexp meets end of file"
+      else
+        rb_compile_error "unterminated string meets end of file"
+      end
+    end
+
+    return :tSTRING_CONTENT, string_buffer.join
+  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
@@ -402,7 +481,7 @@ class RubyLexer
   def process_backref text
     token = ss[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
@@ -420,54 +499,33 @@ class RubyLexer
   end
 
   def process_brace_close text
-    # matching compare/parse23.y:8561
-    cond.lexpop
-    cmdarg.lexpop
-
     case matched
     when "}" then
       self.brace_nest -= 1
-      self.lex_state   = :expr_endarg # TODO: :expr_end ? Look at 2.6
-
       return :tSTRING_DEND, matched if brace_nest < 0
+    end
+
+    # 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_colon1 text
-    # ?: / then / when
-    if is_end? || check(/\s/) then
-      return result :expr_beg, :tCOLON, text
-    end
-
-    case
-    when scan(/\'/) then
-      string STR_SSYM
-    when scan(/\"/) then
-      string STR_DSYM
-    end
-
-    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
-    else
-      result :expr_dot, :tCOLON2, text
-    end
-  end
-
   def process_brace_open text
     # matching compare/parse23.y:8694
     self.brace_nest += 1
@@ -479,67 +537,111 @@ class RubyLexer
       return expr_result(:tLAMBEG, "{")
     end
 
-    token = case lex_state
-            when :expr_labeled then
+    token = case
+            when lex_state =~ EXPR_LABELED then
               :tLBRACE     # hash
-            when *EXPR_ARG_ANY, :expr_end, :expr_endfn then
-              :tLCURLY     # block (primary)
-            when :expr_endarg
+            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
 
-    # TODO: self.lex_state |= :expr_label if token != :tLBRACE_ARG
+    state = token == :tLBRACE_ARG ? EXPR_BEG : EXPR_PAR
     self.command_start = true if token != :tLBRACE
 
-    return expr_result(token, "{")
+    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
+    end
+
+    case
+    when scan(/\'/) then
+      string STR_SSYM
+    when scan(/\"/) then
+      string STR_DSYM
+    end
+
+    result EXPR_FNAME, :tSYMBEG, text
+  end
+
+  def process_colon2 text
+    if is_beg? || lex_state =~ EXPR_CLASS || is_space_arg? then
+      result EXPR_BEG, :tCOLON3, text
+    else
+      result EXPR_DOT, :tCOLON2, text
+    end
   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")
+      return result EXPR_NUM, :tIMAGINARY, Complex(0, Rational(text.chop.chop))
+    when text.end_with?("i")
+      return result EXPR_NUM, :tIMAGINARY, Complex(0, text.chop.to_f)
+    when text.end_with?("r")
+      return result EXPR_NUM, :tRATIONAL,  Rational(text.chop)
     else
-      return result(:expr_end, :tFLOAT, text.to_f)
+      return result EXPR_NUM, :tFLOAT, text.to_f
     end
   end
 
   def process_gvar text
     text.lineno = self.lineno
-    result(:expr_end, :tGVAR, text)
+    result EXPR_END, :tGVAR, text
   end
 
   def process_gvar_oddity text
-    return result :expr_end, "$", "$" if text == "$" # TODO: wtf is this?
+    return result EXPR_END, "$", "$" if text == "$" # 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, self.lineno]
+  end
+
+  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_lchevron text
-    if (!in_lex_state?(:expr_dot, :expr_class) &&
+    if (lex_state !~ EXPR_DOT|EXPR_CLASS &&
         !is_end? &&
-        (!is_arg? || space_seen)) then # TODO: || in_state(:expr_labeled)
+        (!is_arg? || lex_state =~ EXPR_LABELED || space_seen)) then
       tok = self.heredoc_identifier
       return tok if tok
     end
 
-    if in_arg_state? then
-      self.lex_state = :expr_arg
+    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
+      self.command_start = true if lex_state =~ EXPR_CLASS
+      self.lex_state = EXPR_BEG
     end
 
     return result(lex_state, :tLSHFT, "\<\<")
@@ -549,14 +651,14 @@ class RubyLexer
     c = matched
     hit = false
 
-    if c == '#' then
+    if c == "#" then
       ss.pos -= 1
 
       # TODO: handle magic comments
       while scan(/\s*\#.*(\n+|\z)/) do
         hit = true
         self.lineno += matched.lines.to_a.size
-        @comments << matched.gsub(/^ +#/, '#').gsub(/^ +$/, '')
+        @comments << matched.gsub(/^ +#/, "#").gsub(/^ +$/, "")
       end
 
       return nil if end_of_stream?
@@ -567,17 +669,15 @@ class RubyLexer
     # Replace a string of newlines with a single one
     self.lineno += matched.lines.to_a.size if scan(/\n+/)
 
-    # TODO: remove :expr_value -- audit all uses of it
-    c = in_lex_state?(:expr_beg, :expr_value, :expr_class,
-                      :expr_fname, :expr_dot) && !in_lex_state?(:expr_labeled)
-
+    c = (lex_state =~ EXPR_BEG|EXPR_CLASS|EXPR_FNAME|EXPR_DOT &&
+         lex_state !~ EXPR_LABELED)
     # TODO: figure out what token_seen is for
-    # TODO: if c || self.lex_state == [:expr_beg, :expr_labeled] then
-    if c || self.lex_state == :expr_labeled then
+    if c || self.lex_state == EXPR_LAB then # yes, == EXPR_LAB
       # ignore if !fallthrough?
       if !c && parser.in_kwarg then
         # normal newline
-        return result(:expr_beg, :tNL, nil)
+        self.command_start = true
+        return result EXPR_BEG, :tNL, nil
       else
         return # skip
       end
@@ -592,41 +692,46 @@ class RubyLexer
 
     self.command_start = true
 
-    return result(:expr_beg, :tNL, nil)
+    return 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, ss[1].to_i
   end
 
   def process_paren text
-    token = process_paren19
+    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
+              :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_paren19
-    if is_beg? then
-      :tLPAREN
-    elsif is_space_arg? then
-      :tLPAREN_ARG
-    else
-      :tLPAREN2 # plain '(' in parse.y
-    end
+    cond.push false
+    cmdarg.push false
+    result EXPR_PAR, token, text
   end
 
   def process_percent text
     return parse_quote if is_beg?
 
-    return result(:expr_beg, :tOP_ASGN, "%") if scan(/\=/)
+    return result EXPR_BEG, :tOP_ASGN, "%" if scan(/\=/)
 
-    return parse_quote if is_arg? && space_seen && ! check(/\s/)
+    return parse_quote if is_space_arg?(check(/\s/)) || (lex_state =~ EXPR_FITEM && check(/s/))
 
-    return result(:arg_state, :tPERCENT, "%")
+    return result :arg_state, :tPERCENT, "%"
   end
 
   def process_plus_minus text
@@ -637,33 +742,33 @@ 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
-      return result(:expr_value, :tEH, "?")
+      return result EXPR_BEG, :tEH, "?"
     end
 
     if end_of_stream? then
@@ -672,12 +777,12 @@ 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)
@@ -685,18 +790,28 @@ class RubyLexer
       end
 
       # ternary
-      return result(:expr_value, :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
 
-    return result(:expr_end, :tSTRING, c)
+    result EXPR_END, :tSTRING, c
+  end
+
+  def process_simple_string text
+    replacement = text[1..-2].gsub(ESC) {
+      unescape($1).b.force_encoding Encoding::UTF_8
+    }
+
+    replacement = replacement.b unless replacement.valid_encoding?
+
+    result EXPR_END, :tSTRING, replacement
   end
 
   def process_slash text
@@ -707,7 +822,7 @@ class RubyLexer
     end
 
     if scan(/\=/) then
-      return result(:expr_beg, :tOP_ASGN, "/")
+      return result(EXPR_BEG, :tOP_ASGN, "/")
     end
 
     if is_arg? && space_seen then
@@ -726,73 +841,68 @@ 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
 
-    # TODO: this is done by expr_result except "|EXPR_LABEL")
-    # SET_LEX_STATE(EXPR_BEG|EXPR_LABEL);
-    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, /^:"/
+  def process_string # TODO: rewrite / remove
+    # matches top of parser_yylex in compare/parse23.y:8113
+    token = if lex_strterm[0] == :heredoc then
+              self.heredoc lex_strterm
+            else
+              self.parse_string lex_strterm
+            end
 
-    return result(:expr_end, :tSYMBOL, symbol)
-  end
+    token_type, c = token
 
-  def was_label?
-    @was_label = ruby22_label?
-    true
-  end
+    # matches parser_string_term from 2.3, but way off from 2.5
+    if ruby22plus? && token_type == :tSTRING_END && ["'", '"'].include?(c) then
+      if ((lex_state =~ EXPR_BEG|EXPR_ENDFN &&
+           !cond.is_in_state) || is_arg?) &&
+          is_label_suffix? then
+        scan(/:/)
+        token_type = token[0] = :tLABEL_END
+      end
+    end
 
-  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]
+    if [:tSTRING_END, :tREGEXP_END, :tLABEL_END].include? token_type then
+      self.lex_strterm = nil
+      self.lex_state   = (token_type == :tLABEL_END) ? EXPR_PAR : EXPR_LIT
     end
 
-    result :expr_end, :tSTRING, text[1..-2].gsub(/\\\\/, "\\").gsub(/\\'/, "'")
+    return token
   end
 
-  def process_label text
-    symbol = possibly_escape_string text, /^"/
+  def process_symbol text
+    symbol = possibly_escape_string text, /^:\"/ # stupid emacs
 
-    result(:expr_labeled, :tLABEL, [symbol, self.lineno]) # TODO: expr_arg|expr_labeled
+    result EXPR_LIT, :tSYMBOL, symbol
   end
 
   def process_token text
     # matching: parse_ident in compare/parse23.y:7989
     # TODO: make this always return [token, lineno]
-    self.last_state = lex_state
+    # FIX: remove: self.last_state = lex_state
 
     token = self.token = text
     token << matched if scan(/[\!\?](?!=)/)
@@ -801,7 +911,7 @@ 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
@@ -814,31 +924,33 @@ class RubyLexer
 
     if is_label_possible? and is_label_suffix? then
       scan(/:/)
-      # TODO: :expr_arg|:expr_labeled
-      return result :expr_labeled, :tLABEL, [token, self.lineno]
+      # TODO: propagate the lineno to ALL results
+      return result EXPR_LAB, :tLABEL, [token, self.lineno]
     end
 
-    # TODO: mb == ENC_CODERANGE_7BIT && !in_lex_state?(:expr_dot)
-    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 = RubyParserStuff::Keyword.keyword token
 
       return process_token_keyword keyword if keyword
-    end # unless in_lex_state? :expr_dot
+    end
 
     # matching: compare/parse23.y:8079
-    state = if is_beg? or is_arg? or in_lex_state? :expr_dot then
-              cmd_state ? :expr_cmdarg : :expr_arg
-            elsif in_lex_state? :expr_fname then
-              :expr_endfn
+    state = if is_beg? or is_arg? or lex_state =~ EXPR_DOT then
+              cmd_state ? EXPR_CMDARG : EXPR_ARG
+            elsif lex_state =~ EXPR_FNAME then
+              EXPR_ENDFN
             else
-              :expr_end
+              EXPR_END
             end
 
-    if not [:expr_dot, :expr_fname].include? last_state and
-        (tok_id == :tIDENTIFIER) and # not :expr_fname, not attrasgn
+    tok_id = :tIDENTIFIER if tok_id == :tCONSTANT && is_local_id(token)
+
+    if last_state !~ EXPR_DOT|EXPR_FNAME and
+        (tok_id == :tIDENTIFIER) and # not EXPR_FNAME, not attrasgn
         lvar_defined?(token) then
-      state = :expr_end # TODO: EXPR_END|EXPR_LABEL
+      state = EXPR_END|EXPR_LABEL
     end
 
     token.lineno = self.lineno # yes, on a string. I know... I know...
@@ -853,32 +965,30 @@ class RubyLexer
 
     value = [token, self.lineno]
 
-    return result(lex_state, keyword.id0, value) if state == :expr_fname
+    return result(lex_state, keyword.id0, value) if state =~ EXPR_FNAME
 
-    self.command_start = true if lex_state == :expr_beg
+    self.command_start = true if lex_state =~ EXPR_BEG
 
     case
-    when keyword.id0 == :kDO then
+    when keyword.id0 == :kDO then # parse26.y line 7591
       case
       when lambda_beginning? then
         self.lpar_beg = nil # lambda_beginning? == FALSE in the body of "-> do ... end"
-        self.paren_nest -= 1
-        result(lex_state, :kDO_LAMBDA, value)
+        self.paren_nest -= 1 # TODO: question this?
+        result lex_state, :kDO_LAMBDA, value
       when cond.is_in_state then
-        result(lex_state, :kDO_COND, value)
-      when cmdarg.is_in_state && state != :expr_cmdarg then
-        result(lex_state, :kDO_BLOCK, value)
-      when [:expr_beg, :expr_endarg].include?(state) then
-        result(lex_state, :kDO_BLOCK, value)
+        result lex_state, :kDO_COND, value
+      when cmdarg.is_in_state && state != EXPR_CMDARG then
+        result lex_state, :kDO_BLOCK, value
       else
-        result(lex_state, :kDO, value)
+        result lex_state, :kDO, value
       end
-    when [:expr_beg, :expr_labeled].include?(state) then
-      result(lex_state, keyword.id0, value)
+    when state =~ EXPR_PAD then
+      result lex_state, keyword.id0, value
     when keyword.id0 != keyword.id1 then
-      result(:expr_beg, keyword.id1, value) # TODO: :expr_beg|:expr_label
+      result EXPR_PAR, keyword.id1, value
     else
-      result(lex_state, keyword.id1, value)
+      result lex_state, keyword.id1, value
     end
   end
 
@@ -886,9 +996,9 @@ class RubyLexer
     ss.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
+      [RubyLexer::EOF, RubyLexer::EOF]
+    elsif scan(/#{IDENT_CHAR}+/) then
+      process_token matched
     end
   end
 
@@ -921,10 +1031,11 @@ class RubyLexer
     when scan(/s/) then                   # space
       " "
     when scan(/[0-7]{1,3}/) then          # octal constant
-      (matched.to_i(8) & 0xFF).chr
+      (matched.to_i(8) & 0xFF).chr.force_encoding Encoding::UTF_8
     when scan(/x([0-9a-fA-F]{1,2})/) then # hex constant
-      ss[1].to_i(16).chr
-    when check(/M-\\[\\MCc]/) then
+      # TODO: force encode everything to UTF-8?
+      ss[1].to_i(16).chr.force_encoding Encoding::UTF_8
+    when check(/M-\\./) then
       scan(/M-\\/) # eat it
       c = self.read_escape
       c[0] = (c[0].ord | 0x80).chr
@@ -938,6 +1049,11 @@ class RubyLexer
       c = self.read_escape
       c[0] = (c[0].ord & 0x9f).chr
       c
+    when check(/(C-|c)\\(?!u|\\)/) then
+      scan(/(C-|c)\\/) # eat it
+      c = read_escape
+      c[0] = (c[0].ord & 0x9f).chr
+      c
     when scan(/C-\?|c\?/) then
       127.chr
     when scan(/(C-|c)(.)/) then
@@ -946,15 +1062,25 @@ class RubyLexer
       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(/u(\h{4})/) then
+      [ss[1].to_i(16)].pack("U")
+    when scan(/u(\h{1,3})/) then
+      rb_compile_error "Invalid escape character syntax"
+    when scan(/u\{(\h+(?:\s+\h+)*)\}/) then
+      ss[1].split.map { |s| s.to_i(16) }.pack("U*")
     when scan(/[McCx0-9]/) || end_of_stream? then
       rb_compile_error("Invalid escape character syntax")
     else
-      ss.getch
+      getch
     end.dup
   end
 
+  def getch
+    c = ss.getch
+    c = ss.getch if c == "\r" && ss.peek(1) == "\n"
+    c
+  end
+
   def regx_options # TODO: rewrite / remove
     good, bad = [], []
 
@@ -974,7 +1100,7 @@ class RubyLexer
     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
@@ -988,29 +1114,30 @@ class RubyLexer
     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
+  def result new_state, token, text # :nodoc:
+    new_state = self.arg_state if new_state == :arg_state
+    self.lex_state = new_state if new_state
     [token, text]
   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 scan re
+    ss.scan re
   end
 
   def scanner_class # TODO: design this out of oedipus_lex. or something.
@@ -1033,12 +1160,6 @@ class RubyLexer
     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
@@ -1057,8 +1178,10 @@ class RubyLexer
       prev = self.string_buffer.last
       if term == chr && prev && prev.end_with?("(?") then
         self.string_buffer << chr
+      elsif term == chr || chr.ascii_only? then
+        self.string_buffer << matched # dunno why we keep them for ascii
       else
-        self.string_buffer << matched
+        self.string_buffer << chr # HACK? this is such a rat's nest
       end
     else
       rb_compile_error "Invalid escape character syntax"
@@ -1066,22 +1189,24 @@ class RubyLexer
   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
+    qwords = func =~ STR_FUNC_QWORDS
+    escape = func =~ STR_FUNC_ESCAPE
+    expand = func =~ STR_FUNC_EXPAND
+    regexp = func =~ STR_FUNC_REGEXP
+    symbol = func =~ STR_FUNC_SYMBOL
 
     paren_re = @@regexp_cache[paren]
-    term_re  = @@regexp_cache[term]
+    term_re  = if term == "\n"
+                 /#{Regexp.escape "\r"}?#{Regexp.escape "\n"}/
+               else
+                 @@regexp_cache[term]
+               end
 
     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
@@ -1089,7 +1214,9 @@ class RubyLexer
         else
           self.string_nest -= 1
         end
-      when expand && scan(/#(?=[\$\@\{])/) then
+      when paren_re && scan(paren_re) then
+        self.string_nest += 1
+      when expand && scan(/#(?=[\$\@\{])/) then # TODO: this seems wrong
         ss.pos -= 1
         break
       when qwords && scan(/\s/) then
@@ -1103,7 +1230,7 @@ class RubyLexer
           string_buffer << "\n"
           next
         when qwords && scan(/\\\s/) then
-          c = ' '
+          c = " "
         when expand && scan(/\\\n/) then
           next
         when regexp && check(/\\/) then
@@ -1128,12 +1255,16 @@ class RubyLexer
       end # top case
 
       unless handled then
-        t = Regexp.escape term
-        x = Regexp.escape(paren) if paren && paren != "\000"
+        t = if term == "\n"
+              Regexp.escape "\r\n"
+            else
+              Regexp.escape term
+            end
+        x = Regexp.escape paren if paren && paren != "\000"
         re = if qwords then
-               /[^#{t}#{x}\#\0\\\s]+|./ # |. to pick up whatever
+               /[^#{t}#{x}\#\\\s]+|./ # |. to pick up whatever
              else
-               /[^#{t}#{x}\#\0\\]+|./
+               /[^#{t}#{x}\#\\]+|./
              end
 
         scan re
@@ -1173,12 +1304,15 @@ 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 { |s| s.to_i(16) }.pack("U*")
         else
           s
         end
-    x.force_encoding "UTF-8" if HAS_ENC
     x
   end
 
@@ -1186,172 +1320,154 @@ class RubyLexer
     # do nothing for now
   end
 
-  def ruby22plus?
-    parser.class.version >= 22
-  end
-
-  def ruby23plus?
-    parser.class.version >= 23
+  def was_label?
+    @was_label = ruby22_label?
+    true
   end
 
-  def process_string # TODO: rewrite / remove
-    # matches top of parser_yylex in compare/parse23.y:8113
-    token = if lex_strterm[0] == :heredoc then
-              self.heredoc lex_strterm
-            else
-              self.parse_string lex_strterm
-            end
+  class State
+    attr_accessor :n
+    attr_accessor :names
 
-    token_type, c = token
+    # 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
 
-    # matches parser_string_term
-    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
+      self.n = o
+      self.names = names
     end
 
-    if [:tSTRING_END, :tREGEXP_END, :tLABEL_END].include? token_type then
-      self.lex_strterm = nil
-      # TODO: :expr_beg|:expr_label
-      self.lex_state   = (token_type == :tLABEL_END) ? :expr_label : :expr_end
+    def == o
+      self.equal?(o) || (o.class == self.class && o.n == self.n)
     end
 
-    return token
-  end
-
-  def parse_quote # TODO: remove / rewrite
-    beg, nnd, short_hand, c = nil, nil, false, nil
-
-    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
-    end
-
-    if end_of_stream? or c == RubyLexer::EOF or beg == RubyLexer::EOF then
-      rb_compile_error "unterminated quoted string meets end of file"
-    end
-
-    # Figure nnd-char.  "\0" is special to indicate beg=nnd and that no nesting?
-    nnd = { "(" => ")", "[" => "]", "{" => "}", "<" => ">" }[beg]
-    nnd, beg = beg, "\0" if nnd.nil?
-
-    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
-
-    rb_compile_error "Bad %string type. Expected [QqWwIixrs], found '#{c}'." if
-      token_type.nil?
-
-    raise "huh" unless string_type
-
-    string string_type, nnd, beg
-
-    return token_type, text
-  end
-
-  def parse_string quote # TODO: rewrite / remove
-    _, string_type, term, open = quote
-
-    space = false # FIX: remove these
-    func = string_type
-    paren = open
-    term_re = @@regexp_cache[term]
-
-    qwords = (func & STR_FUNC_QWORDS) != 0
-    regexp = (func & STR_FUNC_REGEXP) != 0
-    expand = (func & STR_FUNC_EXPAND) != 0
-
-    unless func then # nil'ed from qwords below. *sigh*
-      return :tSTRING_END, nil
+    def =~ v
+      (self.n & v.n) != 0
     end
 
-    space = true if qwords and eat_whitespace
-
-    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
-      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
 
-    return :tSPACE, nil if space
+    def inspect
+      return "Value(0)" if n.zero? # HACK?
 
-    self.string_buffer = []
-
-    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
-        self.command_start = true
-        return :tSTRING_DBEG, nil
-      when scan(/#/) then
-        string_buffer << '#'
-      end
+      names.map { |v, k| k if self =~ v }.
+        compact.
+        join("|").
+        gsub(/(?:EXPR_|STR_(?:FUNC_)?)/, "")
     end
 
-    if tokadd_string(func, term, paren) == RubyLexer::EOF then
-      rb_compile_error "unterminated 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_ICNTNT = State.new 0x10000, str_func_names # <<~HEREDOC -- TODO: remove?
+
+      # 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_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_ICNTNT => "STR_FUNC_ICNTNT",
+                            STR_SQUOTE      => "STR_SQUOTE")
     end
 
-    return :tSTRING_CONTENT, string_buffer.join
+    include Values
   end
+
+  include State::Values
 end
 
 require "ruby_lexer.rex"
 
 if ENV["RP_LINENO_DEBUG"] then
   class RubyLexer
-    alias :old_lineno= :lineno=
-
     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]]
+      d :lineno => [n, where, ss && ss.rest[0, 40]]
     end
   end
 end