def testGetMin(self):
pq = PriorityQueue()
randomTuples = self.generateRandomTuples(10000)
for randomT in randomTuples:
pq.insertNode(randomT)
currentMin = pq.getMin()
while len(pq.tree) > 1:
nextMin = pq.getMin()
self.assertLessEqual(currentMin[1], nextMin[1])
currentMin = nextMin
def lowerBoundImprovementReversed(source, target) :
"""
Evolution of the normal label-setting algorithm, with a preprocessing that uses
Dijkstra bicriteria algorithm
@return
number of loops
"""
visitedNodes = __dijkstraBiCrit(target, source, 0) # safest path
bestDanger = source.danger
for v in visitedNodes:
v.resetValue(True)
__dijkstraBiCrit(target, source, 1) # shortest path
bestDistance = source.distance
labelQueue = PriorityQueue()
sourceLabel = (0, 0, source, None, 0, None)
source.labelList.append(sourceLabel)
labelQueue.put(sourceLabel, sourceLabel[0])
ending = False
counter = 0
while not labelQueue.isEmpty() :
actualLabel = labelQueue.getMin()
distSoFar = actualLabel[0]
dangSoFar = actualLabel[1]
actualNode = actualLabel[2]
parentIndex = actualLabel[4]
for nearNode, weight in actualNode.neighbors :
distance, danger = weight
ownIndex = len(nearNode.labelList)
newLabel = (distSoFar + distance, dangSoFar + danger, nearNode, actualNode, ownIndex, parentIndex) # creating of a new label
useLabel = True
if nearNode.bestLabel[0] is not None and nearNode.bestLabel[1] is not None : # if the best label is present, enter
checkDistance = newLabel[0] + nearNode.bestLabel[0]
checkDanger = newLabel[1] + nearNode.bestLabel[1]
if checkDanger >= bestDanger and checkDistance >= bestDistance : # if the values are negative is useless to check
continue #checkLabel = (newLabel[0] + checkDistance, newLabel[1] + checkDanger)
else :
checkLabel = newLabel
else :
checkLabel = newLabel
if ending :
if isDominated(checkLabel, target.labelList) : # if the newlabel + bestLabel is dominated is useless go on
continue # restart 'for' loop with another nearNode
for label in nearNode.labelList :
if isDominated(newLabel, [label]) :
useLabel = False
break # if a value is useless (1st case) the loop - label in labelList - is interrupt, to jump some loops
if useLabel :
nearNode.labelList.append(newLabel)
if nearNode != target :
labelQueue.put(newLabel, newLabel[0])
else :
ending = True
counter += 1
return counter
def a_star(source, target):
listOfCand = PriorityQueue()
listOfCand.put(source, 0)
counter = 0
while not listOfCand.isEmpty():
actualNode = listOfCand.getMin()
counter = counter + 1
actualNode.visited = True
if target.visited:
break
for nextNode, distance in actualNode.neighbors:
if not nextNode.visited:
newDistance = actualNode.minWeight + distance[0]
if newDistance < nextNode.minWeight:
#this path is the best until now, let's record
if nextNode.euclidean is None: # this check is to minimazie the calculation of the "euclidean distance"
nextNode.euclidean = euclideanDistance(
nextNode, target)
nextNode.minWeight = newDistance
nextNode.predecessor = actualNode
priority = newDistance + nextNode.euclidean
listOfCand.put(nextNode, priority)
print("loops:", counter)
def __dijkstraBiCrit(source, target, alpha) :
"""
With the bicriteria algorithm I decided to use the attribute "distance"
of the node, because "minWeight" is used has comparator.
Also in normal dijkstra is calcolated the weight, in this specific case we know
that the "weights" are distance and danger
alpha : to varying of this variable the algorithm can find different results
more α is near to 0, will be found the safest path
more α is near to 1, will be found the shortest path
"""
visitedNodes = []
listOfCandidate = PriorityQueue()
listOfCandidate.put(source, 0)
while not listOfCandidate.isEmpty() :
actualNode = listOfCandidate.getMin()
actualNode.visited = True
visitedNodes.append(actualNode)
if target.visited :
break
for nextNode, weight in actualNode.neighbors :
distance, danger = weight
weightResult = alpha*distance + (1-alpha)*danger
newWeight = actualNode.minWeight + weightResult
if newWeight < nextNode.minWeight :
nextNode.minWeight = newWeight
nextNode.distance = actualNode.distance + distance
nextNode.danger = actualNode.danger + danger
nextNode.predecessor = actualNode
listOfCandidate.put(nextNode, newWeight)
if not isDominated((nextNode.distance, nextNode.danger), nextNode.bests) :
for label in nextNode.bests :
if label[0] > nextNode.distance and label[1] >= nextNode.danger :
nextNode.bests.remove(label)
nextNode.bests.append((nextNode.distance, nextNode.danger))
if alpha == 0 : # Those ifs are useful only with the lowerBound Algorithm ()
# if nextNode.bestLabel[1] is None or nextNode.bestLabel[1] > nextNode.danger :
# nextNode.bestLabel[1] = nextNode.danger
if nextNode.bestDangerLabel is None or nextNode.bestDangerLabel[1] > nextNode.danger :
nextNode.bestDangerLabel = (nextNode.distance, nextNode.danger)
elif alpha == 1 :
# if nextNode.bestLabel[0] is None or nextNode.bestLabel[0] > nextNode.distance :
# nextNode.bestLabel[0] = nextNode.distance
if nextNode.bestDistanceLabel is None or nextNode.bestDangerLabel[0] > nextNode.distance :
nextNode.bestDistanceLabel = (nextNode.distance, nextNode.danger)
return visitedNodes
def dijkstra(self, start_label, end_label):
start_vertex = self.find_vertex_by_label(start_label)
start_vertex.weight = 0
pq = PriorityQueue()
pq.build_heap(self.vertices)
while not pq.isEmpty():
min_vertex = pq.getMin() # return Vertex object
for neighbour_vertex in self.adjacencyList[min_vertex.label]:
tmp_vertex = self.find_vertex_by_label(
neighbour_vertex.label) # vertex object from self.vertices
if min_vertex.weight + neighbour_vertex.weight < tmp_vertex.weight:
tmp_vertex.parent = min_vertex.label
tmp_vertex.weight = min_vertex.weight + neighbour_vertex.weight
pq.deleteMin()
print(self.print_path(end_label))
def dijkstraOneToAll(source):
nodeSet = PriorityQueue()
nodeSet.put(source, 0)
# counter = 0
while not nodeSet.isEmpty():
actualNode = nodeSet.getMin()
# counter = counter + 1
for nextNode, distance in actualNode.neighbors:
newDistance = actualNode.minWeight + distance[0]
if newDistance < nextNode.minWeight:
#this is the better path, until now
nextNode.minWeight = newDistance
nextNode.predecessor = actualNode
nodeSet.put(nextNode, newDistance)
source.addShortestPath(
nextNode, newDistance
) # all the path from the starting node are here
def dijkstraListOfCandidate(source, target):
listOfCand = PriorityQueue()
listOfCand.put(source, 0)
while not listOfCand.isEmpty():
actualNode = listOfCand.getMin()
actualNode.visited = True
if target.visited:
break
for nextNode, distance in actualNode.neighbors:
newDistance = actualNode.minWeight + distance[0]
if newDistance < nextNode.minWeight:
nextNode.minWeight = newDistance
nextNode.predecessor = actualNode
listOfCand.put(nextNode, newDistance)
if not target.visited: # if the target is not been visited, it means that it wasn't in the set
print("ERROR! impossible to reach the target")
def prim(self, start_label):
start_vertex = self.find_vertex_by_label(start_label)
start_vertex.weight = 0
pq = PriorityQueue()
#for vertex in self.vertices:
# pq.insert(vertex)
pq.build_heap(self.vertices)
while not pq.isEmpty():
min_vertex = pq.getMin() # return Vertex object
if min_vertex.parent is not None:
self.path += (min_vertex.parent + " -> " + min_vertex.label +
", ")
for neighbour_vertex in self.adjacencyList[min_vertex.label]:
tmp_vertex = self.find_vertex_by_label(
neighbour_vertex.label) # vertex object from self.vertices
if neighbour_vertex.weight < tmp_vertex.weight:
tmp_vertex.parent = min_vertex.label
tmp_vertex.weight = neighbour_vertex.weight
pq.deleteMin()
print(self.path)
def dijskstra(source):
global dist
global adjList
pq = PriorityQueue()
dist[source] = 0
pq.insertNode((source, 0))
while pq.getSize() > 1:
currPair = pq.getMin()
currNodeIndex = currPair[0]
currWeight = currPair[1]
if( currWeight > dist[currNodeIndex]):
continue
for neighbor in adjList[currNodeIndex]:
nextNodeIndex = neighbor[0]
nextWeight = neighbor[1]
if dist[currNodeIndex] + nextWeight < dist[nextNodeIndex]:
dist[nextNodeIndex] = dist[currNodeIndex] + nextWeight
pq.insertNode(neighbor)
def labelSettingAlgorithm(source, target):
"""
label = (dist, danger, owner, predecessor, ownerListPos, predListPos)
[0] [1] [2] [3] [4] [5]
[ownerListPos] : is the position on the own label list
[predListPos] : is the postion in the labelList of the predecessor label
Step 0 :
create first label (0, 0, source, null, 0, null)
and put it in the labelSet (a priorityQueue)
Step 1 :
if labelSet.isEmpty() -> step 4
else select the label w/ the less distance and calc the label of the neighbors
Step 2 :
foreach neighbor of the current label calc
What to do when i found a new label for a node : - 3 possibility
1st - new label IS DOMINATED by one already existence (in the node.labelList) -> delete new label
2nd - new label DOMINATE one or more already existence ( " " ) -> delete old label, add new label
3rd - new label ISN'T DOMINATED AND CAN'T DOMINATE any other node -> add new label
Step 3 :
return to Step 1
Step 4:
End
@return
number of loops
"""
labelQueue = PriorityQueue()
sourceLabel = (0, 0, source, None, 0, None)
source.labelList.append(sourceLabel)
labelQueue.put(sourceLabel, sourceLabel[0])
ending = False # become True when a "nearNode" is equal to the target
counter = 0
while not labelQueue.isEmpty():
counter += 1
actualLabel = labelQueue.getMin()
distSoFar = actualLabel[0]
dangSoFar = actualLabel[1]
actualNode = actualLabel[2]
parentIndex = actualLabel[4]
for nearNode, weight in actualNode.neighbors:
distance, danger = weight
ownIndex = len(nearNode.labelList)
newLabel = (distSoFar + distance, dangSoFar + danger, nearNode,
actualNode, ownIndex, parentIndex
) # creating of a new label
useLabel = True
if ending: # stop condition
if isDominated(newLabel, target.labelList):
continue # restart for loop if the new label is dominated
for label in nearNode.labelList:
if not __usableLabel(newLabel, label, nearNode.labelList):
useLabel = False
break # if a value is useless (1st case) the loop - label in labelList - is interrupt, to jump some loops
if useLabel:
nearNode.labelList.append(newLabel)
if nearNode != target:
labelQueue.put(newLabel, newLabel[0])
else:
ending = True
return counter
