diff --git a/lib/imbecile.rb b/lib/imbecile.rb
index f25d22e..6ec739f 100644
--- a/lib/imbecile.rb
+++ b/lib/imbecile.rb
@@ -3,6 +3,7 @@ require "set"
 require_relative "imbecile/cli"
 require_relative "imbecile/code_point_range"
 require_relative "imbecile/grammar"
+require_relative "imbecile/grammar/token"
 require_relative "imbecile/regex"
 require_relative "imbecile/regex/fa"
 require_relative "imbecile/regex/fa/state"
diff --git a/lib/imbecile/grammar.rb b/lib/imbecile/grammar.rb
index 3d4ebff..3b7bb36 100644
--- a/lib/imbecile/grammar.rb
+++ b/lib/imbecile/grammar.rb
@@ -1,15 +1,21 @@
 module Imbecile
   class Grammar
 
-    # @return [String, nil] Module name.
+    # @return [String, nil]
+    #   Module name.
     attr_reader :modulename
 
-    # @return [String, nil] Class name.
+    # @return [String, nil]
+    #   Class name.
     attr_reader :classname
 
+    # @return [Array<Token>]
+    #   Tokens.
+    attr_reader :tokens
+
     def initialize(input)
-      @tokens = {}
-      @rules = {}
+      @tokens = []
+      token_names = Set.new
       input.each_line.each_with_index do |line, line_index|
         line = line.chomp
         line_number = line_index + 1
@@ -29,10 +35,11 @@ module Imbecile
           unless name =~ /^[a-zA-Z_][a-zA-Z_0-9]*$/
             raise Error.new("Invalid token name #{name} on line #{line_number}")
           end
-          if @tokens[name]
+          if token_names.include?(name)
             raise Error.new("Duplicate token name #{name} on line #{line_number}")
           end
-          @tokens[name] = {pattern: pattern}
+          @tokens << Token.new(name, pattern, @tokens.size)
+          token_names << name
         else
           raise Error.new("Unexpected input on line #{line_number}: #{line}")
         end
@@ -40,12 +47,14 @@ module Imbecile
 
       # Build NFA from each token expression.
       i = 0
-      @tokens.each do |token_name, token_def|
-        token_def[:regex] = Regex.new(token_def[:pattern])
-        token_def[:regex].nfa.end_state.accepts = "#{i}:#{token_name}"
+      nfas = @tokens.map do |token|
+        regex = Regex.new(token.pattern)
+        regex.nfa.end_state.accepts = "#{i}:#{token.name}"
+        puts regex.nfa
         i += 1
+        regex.nfa
       end
-      dfa = Regex::DFA.new(@tokens.map {|token_name, token_def| token_def[:regex].nfa})
+      dfa = Regex::DFA.new(nfas)
       puts dfa
     end
 
diff --git a/lib/imbecile/grammar/token.rb b/lib/imbecile/grammar/token.rb
new file mode 100644
index 0000000..3b858ee
--- /dev/null
+++ b/lib/imbecile/grammar/token.rb
@@ -0,0 +1,27 @@
+module Imbecile
+  class Grammar
+
+    class Token
+
+      # @return [String]
+      #   Token name.
+      attr_reader :name
+
+      # @return [String]
+      #   Token pattern.
+      attr_reader :pattern
+
+      # @return [Integer]
+      #   Token ID.
+      attr_reader :id
+
+      def initialize(name, pattern, id)
+        @name = name
+        @pattern = pattern
+        @id = id
+      end
+
+    end
+
+  end
+end