Do/Undo using command pattern in Python

Question

I have read that using command pattern is one of the most popular ways to accomplish do/undo functionality. In fact, I have seen that it's possible to stack a bunch of actions and reverse them in order to reach a given state. However, I'm not quite sure how that can be done in Python and most of the tutorials I have read, dabble into concepts but don't show an actual implementation in Python.

Does anyone know how do/undo functionality work in Python?

For reference, this is my (naive and probably ridden with errors) code:

# command
class DrawCommand:
    def __init__(self, draw, point1, point2):
        self.draw = draw
        self.point1 = point1
        self.point2 = point2
    def execute_drawing(self):
        self.draw.execute(self.point1, self.point2)
    def execute_undrawing(self):
        self.draw.unexecute(self.point1, self.point2)
# invoker
class InvokeDrawALine:
    def command(self, command):
        self.command = command
    def click_to_draw(self):
        self.command.execute_drawing()
    def undo(self):
        self.command.execute_undrawing()
# receiver
class DrawALine:
    def execute(self, point1, point2):
        print("Draw a line from {} to {}".format(point1, point2))
    def unexecute(self, point1, point2):
        print("Erase a line from {} to {}".format(point1, point2))

instantiating as follows:

invoke_draw = InvokeDrawALine()
draw_a_line = DrawALine()
draw_command = DrawCommand(draw_a_line, 1, 2)
invoke_draw.command(draw_command)
invoke_draw.click_to_draw()
invoke_draw.undo()

output:

Draw a line from 1 to 2
Erase a line from 1 to 2

Obviously, this test doesn't allow stack several actions to undo. Maybe I'm completely mistaken so I would appreciate some help.

Answer 1

Here is an implementation keeping the commands in a list.

# command
class DrawCommand:
    def __init__(self, draw, point1, point2):
        self.draw = draw
        self.point1 = point1
        self.point2 = point2
    def execute_drawing(self):
        self.draw.execute(self.point1, self.point2)
# invoker
class InvokeDrawLines:
    def __init__(self, data):
        self.commandlist = data
    def addcommand(self, command):
        self.commandlist.append(command)
    def draw(self):
        for cmd in self.commandlist:
            cmd.execute_drawing()
    def undocommand(self, command):
        self.commandlist.remove(command)

# receiver
class DrawALine:
    def execute(self, point1, point2):
        print("Draw a line from" , point1, point2)

Answer 2

How I'd go about this

class Command(object):
    def execute(self, canvas):
         raise NotImplementedError

class DrawLineCommand(Command):
    def __init__(self, point1, point2):
        self._point1 = point1
        self._point2 = point2

    def execute(self, canvas):
        canvas.draw_line(self._point1, self._point2)

 class DrawCircleCommand(Command):
     def __init__(self, point, radius):
        self._point = point
        self._radius = radius

     def execute(self, canvas):
        canvas.draw_circle(self._point, self._radius)

class UndoHistory(object):
    def __init__(self, canvas):
        self._commands = []
        self.canvas = canvas

    def command(self, command):
        self._commands.append(command)
        command.execute(self.canvas)

    def undo(self):
        self._commands.pop() # throw away last command
        self.canvas.clear()
        for command self._commands:
            command.execute(self.canvas)

Some thoughts:

1. Trying to undo an action can be hard. For example, how would you undraw a line? You'd need to recover what used to be under that line. A simpler approach is often to revert to a clean slate and then reapply all the commands.
2. Each command should be contained in a single object. It should store all of the data neccesary for the command.
3. In python you don't need to define the Command class. I do it to provide documentation for what methods I expect Command objects to implement.
4. You may eventually get speed issues reapplying all the command for an undo. Optimization is left as an excersize for the reader.