2021-05-29：最常使用的K个单词II。在实时数据流中找到最常使用的k个单词，实现TopK类中的三个方法: TopK(k)， 构造方法。add(word)，增加一个新单词。topk()，得到当前最常使用的k个单词。如果两个单词有相同的使用频率，按字典序排名。
福大大 答案2021-05-30：
方法一：
redis的sorted set。hash+跳表实现计数和查找。无代码。
方法二：
节点结构体：有字符串和词频。
词频表：key是字符串，value是节点。
堆：节点数组。刚开始，我以为是大根堆。采用小根堆，如果比堆顶还小，是进不了小根堆的。
反向表：key是节点，value是在堆中的索引。
有代码。
代码用golang编写。代码如下：
package main
import (
"fmt"
"sort"
)
func main() {
a := NewTopK(2)
a.add("fdd")
a.add("moon")
a.add("moonfdd")
a.add("moonfdd")
ret := a.topk()
for i := 0; i < len(ret); i++ {
fmt.Println(ret[i])
}
}
type TopK struct {
//堆
heap []*Node
heapSize int
//字，次数
wordNodeMap map[string]*Node
//反向表
nodeIndexMap map[*Node]int
}
func NewTopK(k int) *TopK {
ret := &TopK{}
ret.heap = make([]*Node, k)
ret.wordNodeMap = make(map[string]*Node)
ret.nodeIndexMap = make(map[*Node]int)
return ret
}
func (this *TopK) add(word string) {
if len(this.heap) == 0 {
return
}
var curNode *Node
preIndex := -1
curNode = this.wordNodeMap[word]
//词频表 反向表
if curNode == nil {
curNode = &Node{word, 1}
this.wordNodeMap[word] = curNode
this.nodeIndexMap[curNode] = -1
} else {
curNode.Times++
preIndex = this.nodeIndexMap[curNode]
}
//小根堆
if preIndex == -1 {
if this.heapSize == len(this.heap) {
if this.compare(curNode, this.heap[0]) {
//不用管了
return
}
curNode, this.heap[0] = this.heap[0], curNode
this.nodeIndexMap[curNode] = -1
this.nodeIndexMap[this.heap[0]] = 0
this.HeapDown(0)
} else {
this.Push(curNode)
}
} else {
this.HeapDown(preIndex)
}
}
func (this *TopK) topk() []string {
heapCopy := make([]*Node, this.heapSize)
copy(heapCopy, this.heap)
sort.Slice(heapCopy, func(i, j int) bool {
return !this.compare(heapCopy[i], heapCopy[j])
})
ans := make([]string, this.heapSize)
for i := 0; i < this.heapSize; i++ {
ans[i] = heapCopy[i].Str
}
return ans
}
type Node struct {
Str string
Times int
}
//索引上移，小根堆
func (this *TopK) HeapUp(index int) {
for (index-1)/2 != index && !this.compare(this.heap[(index-1)/2], this.heap[index]) { //父节点小于当前节点，当前节点必须上移
this.heap[index], this.heap[(index-1)/2] = this.heap[(index-1)/2], this.heap[index]
//加强堆
this.nodeIndexMap[this.heap[index]], this.nodeIndexMap[this.heap[(index-1)/2]] = (index-1)/2, index
index = (index - 1) / 2
}
}
//索引下沉，小根堆
func (this *TopK) HeapDown(index int) {
left := 2*index + 1
for left <= this.heapSize-1 { //左孩子存在
//获取小孩子
largest := left
if left+1 <= this.heapSize-1 && this.compare(this.heap[left+1], this.heap[left]) {
largest++
}
//比较
if !this.compare(this.heap[index], this.heap[largest]) { //当前大于最小孩子，必须下沉
this.heap[index], this.heap[largest] = this.heap[largest], this.heap[index]
//加强堆
this.nodeIndexMap[this.heap[index]], this.nodeIndexMap[this.heap[largest]] = largest, index
} else {
break
}
//下一次遍历
index = largest
left = 2*index + 1
}
}
func (this *TopK) Push(node *Node) {
this.heap[this.heapSize] = node
//加强堆
this.nodeIndexMap[node] = this.heapSize
//索引上移
this.HeapUp(this.heapSize)
this.heapSize++
}
func (this *TopK) Pop() *Node {
ans := this.heap[0]
this.heap[0], this.heap[this.heapSize-1] = this.heap[this.heapSize-1], this.heap[0]
//加强堆
this.nodeIndexMap[this.heap[0]] = 0
this.nodeIndexMap[this.heap[this.heapSize-1]] = -1
this.heapSize--
//索引下沉
this.HeapDown(0)
return ans
}
func (this *TopK) compare(node1 *Node, node2 *Node) bool {
if node1.Times == node2.Times {
return node1.Str > node2.Str
}
return node1.Times < node2.Times
} 执行结果如下：
福大大 答案2021-05-29：
方法一：
redis的sorted set。hash+跳表实现计数和查找。无代码。
方法二：
节点结构体：有字符串和词频。
词频表：key是字符串，value是节点。
堆：节点数组。
反向表：key是节点，value是在堆中的索引。
有代码，但不完整，因为时间紧。
代码用golang编写。代码如下：
package main
import "fmt"
func main() {
a := NewTopK(2)
a.add("lint")
a.add("code")
a.add("code")
fmt.Println(a.topk())
}
type TopK struct {
//堆
heap []*Node
heapSize int
//字，次数
wordNodeMap map[string]*Node
//反向表
nodeIndexMap map[*Node]int
}
func NewTopK(k int) *TopK {
ret := &TopK{}
ret.heap = make([]*Node, k)
return ret
}
func (this *TopK) add(word string) {
if len(this.heap) == 0 {
return
}
var curNode *Node
preIndex := -1
curNode = this.wordNodeMap[word]
if curNode == nil {
curNode = &Node{word, 1}
this.wordNodeMap[word] = curNode
this.nodeIndexMap[curNode] = -1
} else {
//tree set
curNode.Times++
preIndex = this.nodeIndexMap[curNode]
}
if preIndex == -1 {
if this.heapSize == len(this.heap) {
//treeset
} else {
//tree add
this.nodeIndexMap[curNode] = this.heapSize
this.heap[this.heapSize] = curNode
this.HeapUp(preIndex)
}
} else {
//tree add
this.HeapDown(preIndex)
}
}
func (this *TopK) topk() []string {
ans := make([]string, len(this.heap))
return ans
}
type Node struct {
Str string
Times int
}
//索引上移，大根堆
func (this *TopK) HeapUp(index int) {
for this.heap[(index-1)/2].Times < this.heap[index].Times { //父节点小于当前节点，当前节点必须上移
this.heap[index], this.heap[(index-1)/2] = this.heap[(index-1)/2], this.heap[index]
//加强堆
this.nodeIndexMap[this.heap[index]], this.nodeIndexMap[this.heap[(index-1)/2]] = (index-1)/2, index
index = (index - 1) / 2
}
}
//索引下沉，大根堆
func (this *TopK) HeapDown(index int) {
left := 2*index + 1
for left <= this.heapSize-1 { //左孩子存在
//获取大孩子
largest := left
if left+1 <= this.heapSize-1 && this.heap[left+1].Times > this.heap[left].Times {
largest++
}
//比较
if this.heap[index].Times < this.heap[largest].Times { //当前小于最大孩子，必须下沉
this.heap[index], this.heap[largest] = this.heap[largest], this.heap[index]
//加强堆
this.nodeIndexMap[this.heap[index]], this.nodeIndexMap[this.heap[largest]] = largest, index
} else {
break
}
//下一次遍历
index = largest
left = 2*index + 1
}
}
func (this *TopK) Push(node *Node) {
this.heap[this.heapSize] = node
//加强堆
this.nodeIndexMap[node] = this.heapSize
this.heapSize++
//索引上移
this.HeapUp(this.heapSize)
}
func (this *TopK) Pop() *Node {
ans := this.heap[0]
this.heap[0], this.heap[this.heapSize-1] = this.heap[this.heapSize-1], this.heap[0]
//加强堆
this.nodeIndexMap[this.heap[0]] = 0
this.nodeIndexMap[this.heap[this.heapSize-1]] = -1
this.heapSize--
//索引下沉
this.HeapDown(0)
return ans
} 执行结果如下：
京公网安备 11010502036488号