diff options
Diffstat (limited to 'grc/src/gui/StateCache.py')
-rw-r--r-- | grc/src/gui/StateCache.py | 92 |
1 files changed, 0 insertions, 92 deletions
diff --git a/grc/src/gui/StateCache.py b/grc/src/gui/StateCache.py deleted file mode 100644 index 04b18b18a..000000000 --- a/grc/src/gui/StateCache.py +++ /dev/null @@ -1,92 +0,0 @@ -""" -Copyright 2007 Free Software Foundation, Inc. -This file is part of GNU Radio - -GNU Radio Companion is free software; you can redistribute it and/or -modify it under the terms of the GNU General Public License -as published by the Free Software Foundation; either version 2 -of the License, or (at your option) any later version. - -GNU Radio Companion is distributed in the hope that it will be useful, -but WITHOUT ANY WARRANTY; without even the implied warranty of -MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the -GNU General Public License for more details. - -You should have received a copy of the GNU General Public License -along with this program; if not, write to the Free Software -Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA -""" - -from Actions import FLOW_GRAPH_UNDO, FLOW_GRAPH_REDO, get_action_from_name -from Constants import STATE_CACHE_SIZE - -class StateCache(object): - """ - The state cache is an interface to a list to record data/states and to revert to previous states. - States are recorded into the list in a circular fassion by using an index for the current state, - and counters for the range where states are stored. - """ - - def __init__(self, initial_state): - """ - StateCache constructor. - @param initial_state the intial state (nested data) - """ - self.states = [None] * STATE_CACHE_SIZE #fill states - self.current_state_index = 0 - self.num_prev_states = 0 - self.num_next_states = 0 - self.states[0] = initial_state - self.update_actions() - - def save_new_state(self, state): - """ - Save a new state. - Place the new state at the next index and add one to the number of previous states. - @param state the new state - """ - self.current_state_index = (self.current_state_index + 1)%STATE_CACHE_SIZE - self.states[self.current_state_index] = state - self.num_prev_states = self.num_prev_states + 1 - if self.num_prev_states == STATE_CACHE_SIZE: self.num_prev_states = STATE_CACHE_SIZE - 1 - self.num_next_states = 0 - self.update_actions() - - def get_current_state(self): - """ - Get the state at the current index. - @return the current state (nested data) - """ - self.update_actions() - return self.states[self.current_state_index] - - def get_prev_state(self): - """ - Get the previous state and decrement the current index. - @return the previous state or None - """ - if self.num_prev_states > 0: - self.current_state_index = (self.current_state_index + STATE_CACHE_SIZE -1)%STATE_CACHE_SIZE - self.num_next_states = self.num_next_states + 1 - self.num_prev_states = self.num_prev_states - 1 - return self.get_current_state() - return None - - def get_next_state(self): - """ - Get the nest state and increment the current index. - @return the next state or None - """ - if self.num_next_states > 0: - self.current_state_index = (self.current_state_index + 1)%STATE_CACHE_SIZE - self.num_next_states = self.num_next_states - 1 - self.num_prev_states = self.num_prev_states + 1 - return self.get_current_state() - return None - - def update_actions(self): - """ - Update the undo and redo actions based on the number of next and prev states. - """ - get_action_from_name(FLOW_GRAPH_REDO).set_sensitive(self.num_next_states != 0) - get_action_from_name(FLOW_GRAPH_UNDO).set_sensitive(self.num_prev_states != 0) |