FastPriorityQueue
源码连接:
https://github.com/BlueRaja/High-Speed-Priority-Queue-for-C-Sharp
大致结构:
1节点在内存中的结构还是数组,且首节点为无意义节点,有效节点从索引1开始。(见FastPriorityQueue
2维护的节点必须时固定的继承。(见FastPriorityQueueNode)
3该队列为非线程安全的
4不要对不在优先级队列中的节点调用UpdatePriority(),也不要对在优先级队列的节点调用Enqueue()。
5一个节点只能在一个queue里面,如果要迁移节点到新的队列时应该先将其从原本的queue中移除并重置自身(ResetNode函数)后再加入新queue。
如何实现判断界定是否在容器中。
节点FastPriorityQueueNode中维护了一个数组的索引(见FastPriorityQueueNode的QueueIndex),该索引从1开始。
时间复杂度为O(1)。
是否支持优先级动态变,变更后,如何排序?
支持优先级动态变。(入口见FastPriorityQueueNode的UpdatePriority)
变更后,需要维持大顶堆的大顶,对现有节点(包含新增或者变更节点)根据优先级触发节点上浮CascadeUp或节点下沉CascadeDown。
采用大顶堆的方式,老样子,采用位运算(<<1)方式计算堆中父节点位置,但只维持了大顶,并没有完全排序,即数组中的元素优先级不是从大到小的。
如果是内部节点动态变更优先级,比如随着节点增加,可能构造这样一棵子树:父节点优先级为8,左右子节点的优先级分别为3和5,它本身就不是完全从大到小,当优先级3的节点优先级变成9,只会让它跟父节点交换,交换后依旧不是完全从大到小。
此外,新增一个同优先级的节点是不会把原本节点顶后,而是对新增节点进行下沉操作。(见FastPriorityQueueNode的CascadeDown)
private void CascadeUp(T node)
{
//aka Heapify-up
int parent;
if(node.QueueIndex > 1)
{
parent = node.QueueIndex >> 1;
T parentNode = _nodes[parent];
if(HasHigherOrEqualPriority(parentNode, node))
return;
//Node has lower priority value, so move parent down the heap to make room
_nodes[node.QueueIndex] = parentNode;
parentNode.QueueIndex = node.QueueIndex;
node.QueueIndex = parent;
}
else
{
return;
}
while(parent > 1)
{
parent >>= 1;
T parentNode = _nodes[parent];
if(HasHigherOrEqualPriority(parentNode, node))
break;
//Node has lower priority value, so move parent down the heap to make room
_nodes[node.QueueIndex] = parentNode;
parentNode.QueueIndex = node.QueueIndex;
node.QueueIndex = parent;
}
_nodes[node.QueueIndex] = node;
}
private void CascadeDown(T node)
{
//aka Heapify-down
int finalQueueIndex = node.QueueIndex;
int childLeftIndex = 2 * finalQueueIndex;
// If leaf node, we're done
if(childLeftIndex > _numNodes)
{
return;
}
// Check if the left-child is higher-priority than the current node
int childRightIndex = childLeftIndex + 1;
T childLeft = _nodes[childLeftIndex];
if(HasHigherPriority(childLeft, node))
{
// Check if there is a right child. If not, swap and finish.
if(childRightIndex > _numNodes)
{
node.QueueIndex = childLeftIndex;
childLeft.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = childLeft;
_nodes[childLeftIndex] = node;
return;
}
// Check if the left-child is higher-priority than the right-child
T childRight = _nodes[childRightIndex];
if(HasHigherPriority(childLeft, childRight))
{
// left is highest, move it up and continue
childLeft.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = childLeft;
finalQueueIndex = childLeftIndex;
}
else
{
// right is even higher, move it up and continue
childRight.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = childRight;
finalQueueIndex = childRightIndex;
}
}
// Not swapping with left-child, does right-child exist?
else if(childRightIndex > _numNodes)
{
return;
}
else
{
// Check if the right-child is higher-priority than the current node
T childRight = _nodes[childRightIndex];
if(HasHigherPriority(childRight, node))
{
childRight.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = childRight;
finalQueueIndex = childRightIndex;
}
// Neither child is higher-priority than current, so finish and stop.
else
{
return;
}
}
while(true)
{
childLeftIndex = 2 * finalQueueIndex;
// If leaf node, we're done
if(childLeftIndex > _numNodes)
{
node.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = node;
break;
}
// Check if the left-child is higher-priority than the current node
childRightIndex = childLeftIndex + 1;
childLeft = _nodes[childLeftIndex];
if(HasHigherPriority(childLeft, node))
{
// Check if there is a right child. If not, swap and finish.
if(childRightIndex > _numNodes)
{
node.QueueIndex = childLeftIndex;
childLeft.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = childLeft;
_nodes[childLeftIndex] = node;
break;
}
// Check if the left-child is higher-priority than the right-child
T childRight = _nodes[childRightIndex];
if(HasHigherPriority(childLeft, childRight))
{
// left is highest, move it up and continue
childLeft.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = childLeft;
finalQueueIndex = childLeftIndex;
}
else
{
// right is even higher, move it up and continue
childRight.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = childRight;
finalQueueIndex = childRightIndex;
}
}
// Not swapping with left-child, does right-child exist?
else if(childRightIndex > _numNodes)
{
node.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = node;
break;
}
else
{
// Check if the right-child is higher-priority than the current node
T childRight = _nodes[childRightIndex];
if(HasHigherPriority(childRight, node))
{
childRight.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = childRight;
finalQueueIndex = childRightIndex;
}
// Neither child is higher-priority than current, so finish and stop.
else
{
node.QueueIndex = finalQueueIndex;
_nodes[finalQueueIndex] = node;
break;
}
}
}
}
移除节点如何实现:
先将移除节点A和末尾节点B做个数组位置上的交换。然后对交换后的B进行大顶堆的调整,并置空尾节点。
是否支持元素查询
元素的查询需要自行实现。
标签:node,FastPriorityQueue,childRight,节点,finalQueueIndex,开源,源码,nodes,QueueIndex From: https://www.cnblogs.com/lanyelinxiang/p/17277446.html