简体   繁体   English

缓存在内存中并使用go-routine更新的最佳方法是什么?

[英]What is the best way to cache in memory and update with go-routine?

The Case: Weather API - I will assume that the task is simple and I just want to make an API to return the weather based on another API 案例:天气API-我假设任务很简单,我只想制作一个API以根据另一个API返回天气

The code 编码

package main

import (
    "encoding/json"
    "io/ioutil"
    "net/http"

    "github.com/gorilla/mux"
)

type ResponseBody struct {
    CurrentObservation struct {
        Weather         string `json:"weather"`
        Temperature     string `json:"temperature_string"`
        DisplayLocation struct {
            City string `json:"city"`
        } `json:"display_location"`
    } `json:"current_observation"`
}

var weather ResponseBody

func main() {
    // start the api
    r := mux.NewRouter()
    r.HandleFunc("/", HomeHandler)
    http.ListenAndServe(":8080", r)
}

// handler
func HomeHandler(w http.ResponseWriter, r *http.Request) {
    // load the weather first
    weather = getWeather()
    b, _ := json.Marshal(weather)
    w.Write(b)
}

// get wether from wunderground api
func getWeather() ResponseBody {
    url := "http://api.wunderground.com/api/MY_API_KEY/conditions/q/CA/San_Francisco.json"
    req, err := http.NewRequest("GET", url, nil)
    client := &http.Client{}
    resp, err := client.Do(req)
    if err != nil {
        panic(err)
    }
    defer resp.Body.Close()
    body, _ := ioutil.ReadAll(resp.Body)
    var rb ResponseBody
    json.Unmarshal([]byte(body), &rb)
    return rb
}

Now every time someone hits the API it will send a request to the weather API, but this won't be efficient when I will have concurrent requests, so I will cache it in memory and will update the data in a go-routine every one second 现在,每次有人点击该API时,它将向天气API发送一个请求,但是当我有并发请求时,这将不会很有效,因此我会将其缓存在内存中,并在每个例程中更新数据第二

First: I will move the getWeather call to the main function 首先:我将把getWeather调用移至主要功能

func main() {
    // load the weather first
    weather = getWeather()
    // start the api
    r := mux.NewRouter()
    r.HandleFunc("/", HomeHandler)
    http.ListenAndServe(":8080", r)
}

// handler
func HomeHandler(w http.ResponseWriter, r *http.Request) {
    b, _ := json.Marshal(weather)
    w.Write(b)
}

and will start a go-routine in the main function too 并且也会在主函数中启动go-routine

func main() {
    // load the weather first
    weather = getWeather()
    // update data every 1 second
    go func() {
        for {
            time.Sleep(time.Second)
            weather = getWeather()
        }
    }()
    // start the api
    r := mux.NewRouter()
    r.HandleFunc("/", HomeHandler)
    http.ListenAndServe(":8080", r)
}

so now the application could handle concurrent requests up to minimum 250 concurrent after testing with siege tool 因此,使用攻城工具进行测试后,应用程序现在最多可以处理250个并发请求

Transactions:                250 hits
Availability:             100.00 %
Elapsed time:               0.47 secs
Data transferred:           0.03 MB
Response time:              0.00 secs
Transaction rate:         531.91 trans/sec
Throughput:             0.07 MB/sec
Concurrency:                2.15
Successful transactions:         250
Failed transactions:               0
Longest transaction:            0.04
Shortest transaction:           0.00

So is it right to cache and update data in this way? 那么以这种方式缓存和更新数据是否正确? Or there is something wrong and I should do it in a better way? 还是有什么问题,我应该以更好的方式来做?

The basic approach is OK, but there's a data race on weather . 基本方法是可以的,但是关于weather有一场数据竞赛。 Use a mutex to protect the variable: 使用互斥量来保护变量:

var mu sync.RWMutex
var weather ResponseBody

func main() {
    // load the weather first
    weather = getWeather()
    // update data every 1 second
    go func() {
        for {
            time.Sleep(time.Second)
            mu.Lock()
            weather = getWeather()
            mu.Unlock()
        }
    }()
    // start the api
    r := mux.NewRouter()
    r.HandleFunc("/", HomeHandler)
    http.ListenAndServe(":8080", r)
}

func HomeHandler(w http.ResponseWriter, r *http.Request) {
    mu.RLock()
    b, _ := json.Marshal(weather)
    mu.RUnlock()
    w.Write(b)
}

It is not necessary to guard the first assignment to weather in main because the assignment is guaranteed to happen before the updating goroutine and the request handlers started by ListenAndServer . 不必在main维护对weather的第一个分配,因为可以确保在更新goroutine和ListenAndServer启动的请求处理程序之前进行ListenAndServer

An improvement is to cache the response body bytes: 一种改进是缓存响应主体字节:

var mu sync.RWMutex
var resp []byte

func main() {
    // load the weather first
    weather := getWeather()
    resp, _ = json.Marshal(weather)
    // update data every 1 second
    go func() {
        for {
            time.Sleep(time.Second)
            mu.Lock()
            weather = getWeather()
            resp, _ = json.Marshal(weather)
            mu.Unlock()
        }
    }()
    // start the api
    r := mux.NewRouter()
    r.HandleFunc("/", HomeHandler)
    http.ListenAndServe(":8080", r)
}

func HomeHandler(w http.ResponseWriter, r *http.Request) {
    mu.RLock()
    b := resp
    mu.RUnlock()
    w.Write(b)
}

声明:本站的技术帖子网页,遵循CC BY-SA 4.0协议,如果您需要转载,请注明本站网址或者原文地址。任何问题请咨询:yoyou2525@163.com.

 
粤ICP备18138465号  © 2020-2024 STACKOOM.COM