def runGreedyByOrder(csvFile,orderMethod):
taskList = createTasksFromCsv.getTaskList(csvFile)
helperFunctions.preprocessTimeWindows(taskList)
# create a list of tuples (time of latest time window ending, taskId)
timeWindowsAndPriorities = []
for task in taskList:
latestTime = 0
for day in task.timeWindows:
for window in day:
if window[1] >= latestTime:
latestTime = window[1]
timeWindowsAndPriorities.append([latestTime, task.priority, task.id, float(task.priority)/max(float(task.getNumTimeWindows()), 1), task.dependencyTasks])
# sorting by deadline
timeWindowsAndPriorities = orderMethod(timeWindowsAndPriorities)
# sorting by priority
taskOrdering = []
for timeTaskTuple in timeWindowsAndPriorities:
taskOrdering.append(timeTaskTuple[2])
return helperFunctions.createOptimalSchedule(taskList, taskOrdering)
def runGreedyByOrder(csvFile, orderMethod):
taskList = createTasksFromCsv.getTaskList(csvFile)
helperFunctions.preprocessTimeWindows(taskList)
# create a list of tuples (time of latest time window ending, taskId)
timeWindowsAndPriorities = []
for task in taskList:
latestTime = 0
for day in task.timeWindows:
for window in day:
if window[1] >= latestTime:
latestTime = window[1]
timeWindowsAndPriorities.append([
latestTime, task.priority, task.id,
float(task.priority) / max(float(task.getNumTimeWindows()), 1),
task.dependencyTasks
])
# sorting by deadline
timeWindowsAndPriorities = orderMethod(timeWindowsAndPriorities)
# sorting by priority
taskOrdering = []
for timeTaskTuple in timeWindowsAndPriorities:
taskOrdering.append(timeTaskTuple[2])
return helperFunctions.createOptimalSchedule(taskList, taskOrdering)
def findBestSchedule(taskList, permutations):
helperFunctions.preprocessTimeWindows(taskList)
bestSchedule = Objects.Schedule()
for perm in permutations:
newSchedule = helperFunctions.createOptimalSchedule(taskList, perm)
if isBetterSchedule(newSchedule, bestSchedule):
bestSchedule = newSchedule
return bestSchedule
def findBestSchedule(taskList, permutations):
helperFunctions.preprocessTimeWindows(taskList)
bestSchedule = Objects.Schedule()
for perm in permutations:
newSchedule = helperFunctions.createOptimalSchedule(taskList, perm)
if isBetterSchedule(newSchedule, bestSchedule):
bestSchedule = newSchedule
return bestSchedule
def findBestScheduleWithTimeLimit(taskList, permutations, timeLimit):
'''
A function given a list of task objects and all the potential task permutations
will create a schedule for each ordering and output one of the best schedules in the
time it has that is has tried.
'''
start = time.time()
helperFunctions.preprocessTimeWindows(taskList)
bestSchedule = Objects.Schedule()
for perm in permutations:
newSchedule = helperFunctions.createOptimalSchedule(taskList, perm)
if isBetterSchedule(newSchedule, bestSchedule):
bestSchedule = newSchedule
if time.time() - start >= timeLimit:
return bestSchedule
return bestSchedule
def randomlyPickBestScheduleUnderTime(csvFile, numSeconds):
taskList = createTasksFromCsv.getTaskList(csvFile)
helperFunctions.preprocessTimeWindows(taskList)
bestSchedule, bestProfit = None, 0
taskIds = [task.id for task in taskList]
start = time.time()
while time.time() < (start + numSeconds):
# Get a random permutation
orderToTry = random.sample(taskIds, len(taskIds))
schedule = helperFunctions.createOptimalSchedule(taskList, orderToTry)
if schedule.getProfit() > bestProfit:
bestProfit = schedule.getProfit()
bestSchedule = schedule
return bestSchedule
def randomlyPickBestScheduleUnderTime(csvFile, numSeconds):
taskList = createTasksFromCsv.getTaskList(csvFile)
helperFunctions.preprocessTimeWindows(taskList)
bestSchedule, bestProfit = None, 0
taskIds = [task.id for task in taskList]
start = time.time()
while time.time() < (start + numSeconds):
# Get a random permutation
orderToTry = random.sample(taskIds, len(taskIds))
schedule = helperFunctions.createOptimalSchedule(taskList, orderToTry)
if schedule.getProfit() > bestProfit:
bestProfit = schedule.getProfit()
bestSchedule = schedule
return bestSchedule
def findBestScheduleWithTimeLimit(taskList, permutations, timeLimit):
'''
A function given a list of task objects and all the potential task permutations
will create a schedule for each ordering and output one of the best schedules in the
time it has that is has tried.
'''
start = time.time()
helperFunctions.preprocessTimeWindows(taskList)
bestSchedule = Objects.Schedule()
for perm in permutations:
newSchedule = helperFunctions.createOptimalSchedule(taskList, perm)
if isBetterSchedule(newSchedule, bestSchedule):
bestSchedule = newSchedule
if time.time() - start >= timeLimit:
return bestSchedule
return bestSchedule
def solve(csvFile):
global timeLimit, taskList, delta, tShoe
global startTime
startTime = time.time()
'''
The primal bound holds the priority of best possible schedule we have found so far.
Initialize it to 0. If we find a feasible path with a higher priority (score),
then we update the primal bound.
A multiprocessing manager does all the locking work for you. When you pass it into
a function and the function accesses the managed resource, it is accessed as a proxy,
and the manager makes sure all threads get the updated value. It works with python objects,
which is why it is a list.
'''
manager = multiprocessing.Manager()
#initialize our list of all nodes
taskList = createTasksFromCsv.getTaskList(csvFile)
primalBound = manager.list([0, 0, 0, 0, 0] + [0, 0, 0, 0])
timeLimit = helperFunctions.getLatestDeadline(taskList)
print "timeLimit:", timeLimit
allNodes = [x.id for x in taskList]
allNodes = manager.list(allNodes)
#2-dimensional list with first index value as tao (currentTime), second index value as node
#we can mess with this as we see fit. Make delta larger to speed things up. Make it smaller to have more values.
#delta = timeLimit/10, tShoe = timeLimit/10 make us reach maximum recursion depth
delta = max(10, (timeLimit/dayLength * 3))
tShoe = 9 * timeLimit / 13
print "tShoe", tShoe
boundsInfo = [delta, tShoe]
defineBounds(boundsInfo, primalBound, allNodes)
bestPaths = []
order = reducedBoundsMatrix[min(reducedBoundsMatrix.keys())]
if float("inf") in order:
otherKeys = reducedBoundsMatrix.keys()
otherKeys.remove(min(reducedBoundsMatrix.keys()))
order = reducedBoundsMatrix[min(otherKeys)]
print order
allNodes = sorted(allNodes, key = lambda x: order[x])
print allNodes
print "setting primal bound to ", max(order), " initially"
primalBound[0] = max(order)
'''
Creates a pool of worker threads. Automatically set to the number of
cores in the machine, but you can set that differently with an int argument.
'''
pool = multiprocessing.Pool(len(allNodes)/2 + 1)
'''
This is a partial function, which means it creates an in-between function that
has some set arguments. You can then call it later with the remaining arguments
We use this here so that we can use the map function of the thread pool later
'''
partialPulse = functools.partial(pulse, primalBound, [0], allNodes, allNodes[:], 0)
'''
Calling map on the thread pool with our partial pulse function with added
node args.
'''
#Uncomment the following line for parallel
bestPaths = pool.map(partialPulse, allNodes)
#Uncomment the following line for sequential
# bestPaths = map(partialPulse, allNodes)
print bestPaths
bestSched = max(bestPaths, key = lambda x: getProfit(x))
bestProfit = getProfit(bestSched)
print bestSched
# you have to call this on threads so they stop
pool.close()
# you have to call this on threads after you've closed them
# (read documentation)
pool.join()
bestSched = helperFunctions.createOptimalSchedule(taskList, bestSched[1:])
print "Wow you made it"
print "best schedule\n", bestSched
print "profit", bestProfit
print "primal bound", primalBound
return bestSched
def calendarDrawInit():
fileName = ""
#get the name of the test file we are to use
for b in range(len(fileBooleans)):
if fileBooleans[b]:
fileName = fileListBackEnd[b] + ".csv"
#get an integer value for running time
for b in range(len(timeBooleans)):
if timeBooleans[b]:
timeValue.append(int(timeList[b]))
#do stuff so that we can call different algorithms depending on user's selection
csvFile = pwd(fileName)
#setup for VNS
if buttonList[1] == True:
stoppingCondition = timeValue[0]
schedule[0], useless, useless2 = solve(csvFile, stoppingCondition)
# schedule os.system("python " + vns + " " + file)
#setup for greedy
elif buttonList[2] == True:
sched = greedyByOrder.runGreedyByOrder(csvFile, greedyByOrder.orderByPriority)
greedyByPrioritySol = greedyByOrder.runGreedyByOrder(csvFile, greedyByOrder.orderByPriority)
greedyByDeadlineSol = greedyByOrder.runGreedyByOrder(csvFile, greedyByOrder.orderOptionalByDeadline)
greedyByPresentChoiceSol = greedyByPresentChoice.runGreedyByPresentChoice(csvFile)
solutionList = [greedyByPrioritySol, greedyByDeadlineSol, greedyByPresentChoiceSol]
bestGreedy = max(solutionList, key = lambda sched : sched.getProfit())
schedule[0] = bestGreedy
#setup for Integer
elif buttonList[3] == True:
fileName = pwd(fileName)
output = os.system("python " + pwd("integerProgram.py") + " " + fileName + " " + str(timeValue[0]))
print output
sched = createTasksFromCsv.getTaskList("intProgSched.csv")
order = range(len(sched))
schedule[0] = helperFunctions.createOptimalSchedule(sched, order)
#setup for Pulse
elif buttonList[4]:
fileName = pwd(fileName)
output = os.system("python " + pwd("pulseOPTW.py") + " " + fileName + " " + str(timeValue[0]))
print output
sched = createTasksFromCsv.getTaskList("pulseSched.csv")
order = range(len(sched))
schedule[0] = helperFunctions.createOptimalSchedule(sched, order)
#setup for bruteForce
elif buttonList[5]:
fileName = pwd(fileName)
schedule[0] = runBruteForceAlgWithTimeLimit(csvFile, timeValue[0])
else:
#ERROR: none of our known algorithms were selected when okay when selected
#Default to VNS
pass
#Globals for reference later
dayWidth[0] = (width -sideBarWidth) / len(schedule[0])
#populate the rectColors list with some rainbow colors
setColors()
#determine maximum coordinate values
maxX[0] = 0
maxY[0] = 0
maxNumTimeWindows[0] = 0
for day in range(len(schedule[0])):
for t in range(len(schedule[0][day])):
task = schedule[0][day][t]
taskX = task.x
taskY = task.y
if taskX > maxX[0]:
maxX[0] = taskX
if taskY > maxY[0]:
maxY[0] = taskY
#get maxNumTimeWindows so we can make ghost time windows appropriate width in highlight()
for j in range(len(schedule[0][day][t].timeWindows)):
if len(schedule[0][day][t].timeWindows[j]) > maxNumTimeWindows[0]:
maxNumTimeWindows[0] = len(schedule[0][day][t].timeWindows[j])
background(bgColor)