Experimental parser: remove hacks from replace_tokens_with_gotos.

tools/lexer_generator/dfa_optimizer.py

 # Copyright 2013 the V8 project authors. All rights reserved.
 # Redistribution and use in source and binary forms, with or without
 # modification, are permitted provided that the following conditions are
 # met:
 #
 #     * Redistributions of source code must retain the above copyright
 #       notice, this list of conditions and the following disclaimer.
 #     * Redistributions in binary form must reproduce the above
 #       copyright notice, this list of conditions and the following
 #       disclaimer in the documentation and/or other materials provided
@@ skipping 48 matching lines @@
     incoming_transitions = {}
     def build_incoming_transitions(state, visitor_state):
       for key, transition_state in state.key_state_iter():
         if not transition_state in incoming_transitions:
           incoming_transitions[transition_state] = []
         incoming_transitions[transition_state].append((key,state))
     dfa.visit_all_states(build_incoming_transitions)
 
     def is_replacement_candidate(state):
       action = state.action()
-      if not action or action.entry_action() or not action.match_action():
+      if not action or not action.match_action():
         return False
       if (action.match_action()[0] == 'token' or
           action.match_action()[0] == 'harmony_token'):
         return True
       return False
 
     replacements = {}
     for state, incoming in incoming_transitions.items():
       if len(incoming) < 10:
         continue
@@ skipping 29 matching lines @@
     name = lambda state : str(state.node_number())
     counters = {
       'removals' : 0,
       'goto_start' : 0,
       'store_token_and_goto' : 0,
       'store_harmony_token_and_goto' : 0,
       }
     store_states = set([])
     # generate a store action to replace an existing action
     def replacement_action(old_action, transition_state):
-      assert not old_action.entry_action()
       assert old_action.match_action()
       state_id = name(transition_state)
       if old_action.match_action()[0] == 'token':
         old_token = old_action.match_action()[1]
         if (transition_state.action().match_action()[0] == 'token' and
             transition_state.action().match_action()[1] == old_token):
           # no need to store token
           match_action = ('goto_start', (state_id,))
           counters['goto_start'] += 1
         else:
           counters['store_token_and_goto'] += 1
           match_action = ('store_token_and_goto', (old_token, state_id))
       elif old_action.match_action()[0] == 'harmony_token':
         match_action = ('store_harmony_token_and_goto',
                         (old_action.match_action()[1][0],
                          old_action.match_action()[1][1],
                          old_action.match_action()[1][2],
                          state_id))
         counters['store_harmony_token_and_goto'] += 1
       else:
         raise Exception(old_action.match_action())
-      return Action(None, match_action, old_action.precedence())
+      return Action(old_action.entry_action(), match_action,
+                    old_action.precedence())
     # map the old state to the new state, with fewer transitions and
     # goto actions
     def replace_state(state, acc):
       new_state = {
         'transitions' : {},
         'terminal' : state in self.__dfa.terminal_set(),
         'action' : state.action(),
       }
       states[name(state)] = new_state
       state_replacements = replacements[state] if state in replacements else {}
       for transition_key, transition_state in state.transitions().items():
         if not transition_key in state_replacements:
           new_state['transitions'][transition_key] = name(transition_state)
           continue
         replacement = state_replacements[transition_key]
         del state_replacements[transition_key]
         # just drop the transition
         if replacement == None:
           counters['removals'] += 1
           continue
         assert replacement == transition_state
         # do goto replacement
         new_state['action'] = replacement_action(state.action(), replacement)
         # will need to patch up transition_state
         store_states.add(name(transition_state))
       assert not state_replacements
     self.__dfa.visit_all_states(replace_state)
     # now patch up all states with stores
+    start_name = name(self.__dfa.start_state())
     for state_id in store_states:
       old_action = states[state_id]['action']
       assert not old_action.entry_action()
       assert old_action.match_action()[0] == 'token', 'unimplemented'
       entry_action = ('store_token', old_action.match_action()[1])
       match_action = ('do_stored_token', state_id)
       precedence = old_action.precedence()
       states[state_id]['action'] = Action(
         entry_action, match_action, precedence)
-    start_name = name(self.__dfa.start_state())
+      # The state might be only reachable via gotos; make sure it's connected in
+      # the state graph by adding a bogus transition from the start state. This
+      # transition doens't match any character.
+      states[start_name]['transitions'][
+          TransitionKey.unique('no_match')] = state_id
+
     if self.__log:
       print 'goto_start inserted %s' % counters['goto_start']
       print 'store_token_and_goto inserted %s' % (
         counters['store_token_and_goto'])
       print 'store_harmony_token_and_goto %s' % (
         counters['store_harmony_token_and_goto'])
       print 'transitions removed %s' % counters['removals']
       print 'do_stored_token actions added %s' % len(store_states)
     self.__dfa = Dfa(self.__dfa.encoding(), start_name, states)
 
   @staticmethod
   def optimize(dfa, log):
     optimizer = DfaOptimizer(dfa, log)
     optimizer.replace_tokens_with_gotos()
     return optimizer.__dfa