Python/graphs/edmonds_karp_multiple_source_and_sink.py
2019-10-05 10:14:13 +05:00

194 lines
6.4 KiB
Python

class FlowNetwork:
def __init__(self, graph, sources, sinks):
self.sourceIndex = None
self.sinkIndex = None
self.graph = graph
self._normalizeGraph(sources, sinks)
self.verticesCount = len(graph)
self.maximumFlowAlgorithm = None
# make only one source and one sink
def _normalizeGraph(self, sources, sinks):
if sources is int:
sources = [sources]
if sinks is int:
sinks = [sinks]
if len(sources) == 0 or len(sinks) == 0:
return
self.sourceIndex = sources[0]
self.sinkIndex = sinks[0]
# make fake vertex if there are more
# than one source or sink
if len(sources) > 1 or len(sinks) > 1:
maxInputFlow = 0
for i in sources:
maxInputFlow += sum(self.graph[i])
size = len(self.graph) + 1
for room in self.graph:
room.insert(0, 0)
self.graph.insert(0, [0] * size)
for i in sources:
self.graph[0][i + 1] = maxInputFlow
self.sourceIndex = 0
size = len(self.graph) + 1
for room in self.graph:
room.append(0)
self.graph.append([0] * size)
for i in sinks:
self.graph[i + 1][size - 1] = maxInputFlow
self.sinkIndex = size - 1
def findMaximumFlow(self):
if self.maximumFlowAlgorithm is None:
raise Exception("You need to set maximum flow algorithm before.")
if self.sourceIndex is None or self.sinkIndex is None:
return 0
self.maximumFlowAlgorithm.execute()
return self.maximumFlowAlgorithm.getMaximumFlow()
def setMaximumFlowAlgorithm(self, Algorithm):
self.maximumFlowAlgorithm = Algorithm(self)
class FlowNetworkAlgorithmExecutor(object):
def __init__(self, flowNetwork):
self.flowNetwork = flowNetwork
self.verticesCount = flowNetwork.verticesCount
self.sourceIndex = flowNetwork.sourceIndex
self.sinkIndex = flowNetwork.sinkIndex
# it's just a reference, so you shouldn't change
# it in your algorithms, use deep copy before doing that
self.graph = flowNetwork.graph
self.executed = False
def execute(self):
if not self.executed:
self._algorithm()
self.executed = True
# You should override it
def _algorithm(self):
pass
class MaximumFlowAlgorithmExecutor(FlowNetworkAlgorithmExecutor):
def __init__(self, flowNetwork):
super(MaximumFlowAlgorithmExecutor, self).__init__(flowNetwork)
# use this to save your result
self.maximumFlow = -1
def getMaximumFlow(self):
if not self.executed:
raise Exception("You should execute algorithm before using its result!")
return self.maximumFlow
class PushRelabelExecutor(MaximumFlowAlgorithmExecutor):
def __init__(self, flowNetwork):
super(PushRelabelExecutor, self).__init__(flowNetwork)
self.preflow = [[0] * self.verticesCount for i in range(self.verticesCount)]
self.heights = [0] * self.verticesCount
self.excesses = [0] * self.verticesCount
def _algorithm(self):
self.heights[self.sourceIndex] = self.verticesCount
# push some substance to graph
for nextVertexIndex, bandwidth in enumerate(self.graph[self.sourceIndex]):
self.preflow[self.sourceIndex][nextVertexIndex] += bandwidth
self.preflow[nextVertexIndex][self.sourceIndex] -= bandwidth
self.excesses[nextVertexIndex] += bandwidth
# Relabel-to-front selection rule
verticesList = [
i
for i in range(self.verticesCount)
if i != self.sourceIndex and i != self.sinkIndex
]
# move through list
i = 0
while i < len(verticesList):
vertexIndex = verticesList[i]
previousHeight = self.heights[vertexIndex]
self.processVertex(vertexIndex)
if self.heights[vertexIndex] > previousHeight:
# if it was relabeled, swap elements
# and start from 0 index
verticesList.insert(0, verticesList.pop(i))
i = 0
else:
i += 1
self.maximumFlow = sum(self.preflow[self.sourceIndex])
def processVertex(self, vertexIndex):
while self.excesses[vertexIndex] > 0:
for neighbourIndex in range(self.verticesCount):
# if it's neighbour and current vertex is higher
if (
self.graph[vertexIndex][neighbourIndex]
- self.preflow[vertexIndex][neighbourIndex]
> 0
and self.heights[vertexIndex] > self.heights[neighbourIndex]
):
self.push(vertexIndex, neighbourIndex)
self.relabel(vertexIndex)
def push(self, fromIndex, toIndex):
preflowDelta = min(
self.excesses[fromIndex],
self.graph[fromIndex][toIndex] - self.preflow[fromIndex][toIndex],
)
self.preflow[fromIndex][toIndex] += preflowDelta
self.preflow[toIndex][fromIndex] -= preflowDelta
self.excesses[fromIndex] -= preflowDelta
self.excesses[toIndex] += preflowDelta
def relabel(self, vertexIndex):
minHeight = None
for toIndex in range(self.verticesCount):
if (
self.graph[vertexIndex][toIndex] - self.preflow[vertexIndex][toIndex]
> 0
):
if minHeight is None or self.heights[toIndex] < minHeight:
minHeight = self.heights[toIndex]
if minHeight is not None:
self.heights[vertexIndex] = minHeight + 1
if __name__ == "__main__":
entrances = [0]
exits = [3]
# graph = [
# [0, 0, 4, 6, 0, 0],
# [0, 0, 5, 2, 0, 0],
# [0, 0, 0, 0, 4, 4],
# [0, 0, 0, 0, 6, 6],
# [0, 0, 0, 0, 0, 0],
# [0, 0, 0, 0, 0, 0],
# ]
graph = [[0, 7, 0, 0], [0, 0, 6, 0], [0, 0, 0, 8], [9, 0, 0, 0]]
# prepare our network
flowNetwork = FlowNetwork(graph, entrances, exits)
# set algorithm
flowNetwork.setMaximumFlowAlgorithm(PushRelabelExecutor)
# and calculate
maximumFlow = flowNetwork.findMaximumFlow()
print("maximum flow is {}".format(maximumFlow))