Golang实现词频统计

本例使用golang实现词频统计。步骤：


（1）从文件中读取一篇文章。


（2）统计词频，按单词出现的频率从大到小进行排序。


（3）写入到文件中。


注：任何非英文字母的符号均认为是单词分隔符（即等同于空格）。


效率：使用本程序统计一篇150W单词的文章，大约需要70ms.


1.核心代码：


package wordtest


import (
"bytes"
"fmt"
"io/ioutil"
"os"
"runtime"
"sort"
"strings"
"time"
)


//简单的词频统计任务
func CountTestBase(inputFilePath string, outputFilePath string) {
//时间开始点
start := time.Now().UnixNano() / 1e6


//读取文件
fileData, err := ioutil.ReadFile(inputFilePath)
CheckError(err, "read file")
var fileText string = string(fileData)


//根据CPU核数新开协程
newRountineCount := runtime.NumCPU()*2 - 1
runtime.GOMAXPROCS(newRountineCount + 1)
//切分文件
parts := splitFileText(fileText, newRountineCount)


var ch chan map[string]int = make(chan map[string]int, newRountineCount)
for i := 0; i < newRountineCount; i++ {
go countTest(parts[i], ch)
}


//主线程接收数据
var totalWordsMap map[string]int = make(map[string]int, 0)
completeCount := 0
for {
receiveData := <-ch
for k, v := range receiveData {
totalWordsMap[strings.ToLower(k)] += v
}
completeCount++


if newRountineCount == completeCount {
break
}
}


//添加进slice，并排序
list := make(WordCountBeanList, 0)
for k, v := range totalWordsMap {
list = append(list, NewWordCountBean(k, v))
}
sort.Sort(list)
//时间结束点
end := time.Now().UnixNano() / 1e6
fmt.Printf("time consume:%dms\n", end-start)


//输出
wordsCount := list.totalCount()
var data bytes.Buffer
data.WriteString(fmt.Sprintf("程序执行：%dms\n", end-start))
data.WriteString(fmt.Sprintf("文章总单词数：%d\n\n", wordsCount))
for _, v := range list {
var percent float64 = 100.0 * float64(v.count) / float64(wordsCount)
_, err := data.WriteString(fmt.Sprintf("%s: %d, %3.2f%%\n", v.word, v.count, percent))
CheckError(err, "bytes.Buffer, WriteString")
}


err = ioutil.WriteFile(outputFilePath, []byte(data.String()), os.ModePerm)
CheckError(err, "ioutil.WriteFile")
}


func countTest(text string, ch chan map[string]int) {
var wordMap map[string]int = make(map[string]int, 0)


//按字母读取，除26个字母（大小写）之外的所有字符均认为是分隔符
startIndex := 0
letterStart := false
for i, v := range text {
if (v >= 65 && v <= 90) || (v >= 97 && v <= 122) {
if !letterStart {
letterStart = true
startIndex = i
}
} else {
if letterStart {
wordMap[text[startIndex:i]]++
letterStart = false
}
}
}


//最后一个单词
if letterStart {
wordMap[text[startIndex:]]++
}
ch <- wordMap
}


//将全文分成n段
func splitFileText(fileText string, n int) []string {
length := len(fileText)
parts := make([]string, n)


lastPostion := 0
for i := 0; i < n-1; i++ {
position := length / n * (i + 1)
for string(fileText[position]) != " " {
position++
}


parts[i] = fileText[lastPostion:position]
lastPostion = position
}


//最后一段
parts[n-1] = fileText[lastPostion:]
return parts
}


func CheckError(err error, msg string) {
if err != nil {
panic(msg + "," + err.Error())
}
}
2.一个struct
package wordtest


type WordCountBean struct {
word  string
count int
}


func NewWordCountBean(word string, count int) *WordCountBean {
return &WordCountBean{word, count}
}


type WordCountBeanList []*WordCountBean


func (list WordCountBeanList) Len() int {
return len(list)
}


func (list WordCountBeanList) Less(i, j int) bool {
if list[i].count > list[j].count {
return true
} else if list[i].count < list[j].count {
return false
} else {
return list[i].word < list[j].word
}
}


func (list WordCountBeanList) Swap(i, j int) {
var temp *WordCountBean = list[i]
list[i] = list[j]
list[j] = temp
}


func (list WordCountBeanList) totalCount() int {
totalCount := 0
for _, v := range list {
totalCount += v.count
}


return totalCount
}
3.主函数：
package main


import (
"WordsTest/wordtest"
)


func main() {
inputFilePath := "files/article.txt"
outputFilePath := "files/hanjun-result.txt"


wordtest.CountTestBase(inputFilePath, outputFilePath)
}