Difference between revisions of "A priority queue based turn scheduling system"

From RogueBasin
Jump to navigation Jump to search
Line 19: Line 19:
<pre>
<pre>
class PriorityQueue:
class PriorityQueue:
    """
        Taken from http://aspn.activestate.com/ASPN/Cookbook/Python/
          Recipe/68435.
    """
     def __init__(self):
     def __init__(self):
         self.__queue = []
         self.__queue = []

Revision as of 15:55, 13 September 2007

A priority queue based turn scheduling system

This is an implementation in Python of a turn scheduling system that uses a priority queue. Events include monster and player actions, and therefore is able to represent the relative speed of different creatures.

The theory

When creatures take turns in a roguelike, sometimes it is desireable to have faster creatures to receive more turns relative to other creatures. For example, the player may take two turns for every one turn the giant slug makes, while the player is outrun by the bat who takes three turns for every turn the player makes.

To achieve this effect, a creatures speed is thought of giving it a delay between its actions. When a creature take a turn, there is some delay until it may act again. Faster creatures may have turns between this delay. The turn delay is shorter for faster creatures. Referring to the last example, the player would have twice the speed as the giant slug, and therefore receives half the delay between his or her turns.

These delays are handled by using a priority queue and keeping track of the current game time. When the game starts, the game time is set to zero and all creatures on a level (including the player) are inserted into the priority queue with a priority of zero. The priorities in the queue actually represent the time when the creature will act.

To execute a turn, a creature is pulled off the queue. If the priority the creature had in the queue is greater than the current game time, the game time is set to the priority, representing the passage of time. After the creature takes its action, it is placed back into the queue with a priority equal to the sum of the game time and the creature's delay.

The priority queue

Of course, first we need our priority queue. This uses a sorted array, such that things with the same priority are ordered by their insertion order.

class PriorityQueue:
    def __init__(self):
        self.__queue = []

    def enqueue(self, value, priority = 0.0):
        """
            Append a new element to the queue
            according to its priority.
        """

        tuple = (priority, value)

        # Insert the tuple in sorted position in the queue.  If a
        # tuple with equal priority is encountered, the new tuple is
        # inserted after it.
        # This algorithm is based on Python's bisect.insort function,
        # which may be used by calling:
        #       bisect.insort(self.__queue, tuple)

        low = 0
        high = len(self.__queue)

        while low < high:
            mid = (low + high) // 2

            if tuple[0] < self.__queue[mid][0]:
                high = mid
            else:
                low = mid + 1

        self.__queue.insert(low, tuple)

    def dequeue(self):
        """
            Pop the value with the lowest priority.
        """

        return self.__queue.pop(0)[1]

    def dequeueWithKey(self):
        """
            Pop the (priority, value) tuple with the lowest priority.
        """

        return self.__queue.pop(0)

    def erase(self, value):
        """
            Removes an element from the queue by value.
        """

        self.__erase(value, lambda a, b: a == b)

    def erase_ref(self, value):
        """
            Removes an element from the queue by reference.
        """

        self.__erase(value, lambda a, b: a is b)

    def __erase(self, value, test):
        """
            All tupples t are removed from the queue
            if test(t[1], value) evaluates to True.
        """

        i = 0
        while i < len(self.__queue):
            if test(self.__queue[i][1], value):
                del self.__queue[i]
            else:
                i += 1

The time schedule

The time schedule manages the priority queue and keeps track of the current game time.

class TimeSchedule(object):
    """
        Represents a series of events that occur over time.
    """

    def __init__(self):
        self.__currentTime = 0.0
        self.__scheduledEvents = PriorityQueue()

    # Try to enforce read-only access to currentTime (why oh why can't
    # Python have ^private^ like all the other good languages do?
    # >_< ).
    currentTime = property(fget = lambda self: self.__currentTime)

    def scheduleEvent(self, event, delay = 0.0):
        """
            Schedules an event to occur after a certain
            delay relative to the current time.
        """

        self.__scheduledEvents.enqueue( \
          event, delay + self.__currentTime)

    def nextEvent(self):
        """
            Dequeues and returns the next event to occur.  This
            represents the occurence of an event, and therefore
            increases the time.
        """

        time, event = self.__scheduledEvents.dequeueWithKey()

        if time > self.__currentTime:
            self.__currentTime = time

        return event

    def cancelEvent(self, event):
        """
            Cancels a pending event.
        """

        self.__scheduledEvents.erase_ref(event)

Testing

This is some demonstration code showing how the time scheduler may be used to control the turn order of creatures in a game. Here, the delay time for a Thing is a base delay time divided by the thing's speed.

class Thing(object):
    """
        Just something to test.
        Assumes that the maximum speed of a thing is 10.
    """

    BASE_TIME = 10.0

    def __init__(self, id, speed):
        self.id = id
        self.speed = speed

    def __str__(self):
        return self.id

    def actionDelay(self):
        return Thing.BASE_TIME / self.speed

#---------------------------------------------------------------------

TURN_ROUNDS = 2

if __name__ == '__main__':
    q = TimeSchedule()

    things = [Thing('a', 1), Thing('b', 2), Thing('c', 1)]

    turns = 0
    turnsTaken = {}
    for t in things:
        q.scheduleEvent(t)
        turns += t.speed
        turnsTaken[t] = 0

    turns *= TURN_ROUNDS

    while turns > 0:
        thing = q.nextEvent()

        turnsTaken[thing] += 1
        print thing
        turns -= 1

        q.scheduleEvent(thing, thing.actionDelay())

    for thing, numTurns in turnsTaken.iteritems():
        assert numTurns == (thing.speed * TURN_ROUNDS)

    print
    for id, numTurns in turnsTaken.iteritems():
        print id, numTurns

"""
Output:

a
b
c
b
a
b
c
b

a 2
b 4
c 2

"""

Observations

  • The number of turns a creature receives in a repeated sequence of turns is equal to its speed.
  • Doubling the speed of all creatures maintains the same turn pattern (though one may notice the turn pattern will double in length)

Other notes

  • Different creature actions may result in different delays, causing creatures to wait for shorter or longer periods. For example, attacking may take twice as long as walking. To implement this, insert creatures back into the turn scheduler with different delays depending on what action they took. Mind you, don't go overboard with this or the player won't be able to tell what's going on.
  • The turn scheduler does not assume that it is storing creaturs. It may store any kind of events scheduled to occur in the game, and may be modified to handle regeneration, hunger, spell effects, and one-time-only events like time bombs.
  • The regularity of the turn pattern may allow players to employ so-called "pillar dancing" and "hack and back" maneuvers. It is up to the developer whether this is a bug or a feature (if you think it's a feature, consider making an AI that can also pillar dance and hack-and-back).