def _step()

in gym/gym/envs/board_game/go.py [0:0]

• 35 lines of code
• 9 McCabe index (conditional complexity)

    def _step(self, action):
        assert self.state.color == self.player_color

        # If already terminal, then don't do anything
        if self.done:
            return self.state.board.encode(), 0., True, {'state': self.state}

        # If resigned, then we're done
        if action == _resign_action(self.board_size):
            self.done = True
            return self.state.board.encode(), -1., True, {'state': self.state}

        # Play
        prev_state = self.state
        try:
            self.state = self.state.act(action)
        except pachi_py.IllegalMove:
            if self.illegal_move_mode == 'raise':
                six.reraise(*sys.exc_info())
            elif self.illegal_move_mode == 'lose':
                # Automatic loss on illegal move
                self.done = True
                return self.state.board.encode(), -1., True, {'state': self.state}
            else:
                raise error.Error('Unsupported illegal move action: {}'.format(self.illegal_move_mode))

        # Opponent play
        if not self.state.board.is_terminal:
            self.state, opponent_resigned = self._exec_opponent_play(self.state, prev_state, action)
            # After opponent play, we should be back to the original color
            assert self.state.color == self.player_color

            # If the opponent resigns, then the agent wins
            if opponent_resigned:
                self.done = True
                return self.state.board.encode(), 1., True, {'state': self.state}

        # Reward: if nonterminal, then the reward is 0
        if not self.state.board.is_terminal:
            self.done = False
            return self.state.board.encode(), 0., False, {'state': self.state}

        # We're in a terminal state. Reward is 1 if won, -1 if lost
        assert self.state.board.is_terminal
        self.done = True
        white_wins = self.state.board.official_score > 0
        black_wins = self.state.board.official_score < 0
        player_wins = (white_wins and self.player_color == pachi_py.WHITE) or (black_wins and self.player_color == pachi_py.BLACK)
        reward = 1. if player_wins else -1. if (white_wins or black_wins) else 0.
        return self.state.board.encode(), reward, True, {'state': self.state}