C++ 手写 HTTP 服务器：从请求解析到响应构建实战

HTTP 服务器本质上是一个 TCP 服务器，区别在于增加了协议解析逻辑：

socket → bind → listen → accept ↓ 创建线程 ↓ read → 解析 → write

关键步骤拆解

1. 读取浏览器请求

使用 read 函数接收客户端数据。注意缓冲区大小要留有余地。

char buffer[1024]; 
ssize_t s = read(sockfd, buffer, sizeof(buffer) - 1);

浏览器发来的原始数据大致如下：

GET /index.html HTTP/1.1
Host: localhost:8080
User-Agent: Mozilla/5.0...

2. 解析请求行

我们需要提取出请求方法和 URL 路径。这里用 stringstream 处理比较方便。

std::stringstream ss(req_header[0]); 
std::string method, url, version; 
ss >> method >> url >> version;

我们主要关注 method 和 url。

3. 拼接文件路径

根据 URL 映射到本地文件系统。根路径 / 通常对应 index.html。

std::string path;
if (url == "/" || url == "/index.html") {
    path = "./wwwroot/index.html";
} else {
    path = "./wwwroot" + url;
}

4. 读取网页内容

使用 ifstream 读取文件内容。注意处理文件不存在的情况。

std::ifstream in(path);
std::string content;
std::string line;
while (std::getline(in, line)) {
    content += line + "\n";
}
in.close();

5. 构造 HTTP 响应

手动拼接符合 HTTP 规范的响应报文。重点是不能漏掉那个关键的空行 \r\n\r\n。

std::string response;
response = "HTTP/1.0 200 OK\r\n";
response += "Content-Length: " + std::to_string(content.size()) + "\r\n";
response += "Content-Type: text/html\r\n";
response += "\r\n"; // 关键的空行
response += content;

write(sockfd, response.c_str(), response.size());

6. 处理 404 错误

当文件找不到时，返回标准的 404 响应。

if (!in.is_open()) {
    std::string body = "<h1>404 Not Found</h1>";
    response = "HTTP/1.0 404 Not Found\r\n";
    response += "Content-Length: " + std::to_string(body.size()) + "\r\n";
    response += "Content-Type: text/html\r\n";
    response += "\r\n";
    response += body;
}

完整代码示例

为了方便大家理解，我将核心逻辑整理为两个文件：Socket.hpp 封装了底层 Socket 操作，HttpServer.cc 实现了业务逻辑。

Socket.hpp

#include <iostream>
#include <string>
#include <unistd.h>
#include <sys/socket.h>
#include <sys/stat.h>
#include <arpa/inet.h>
#include <cstring>

enum Err { SocketErr = 1, BindErr, ListenErr };

class Socket {
public:
    Socket() {}
    void Init() {
        sockfd_ = socket(AF_INET, SOCK_STREAM, 0);
        if (sockfd_ < 0) {
            std::cout << "socket fail!!!" << std::endl;
            exit(SocketErr);
        }
    }
    void Bind(uint16_t port, std::string ip) {
        struct sockaddr_in server;
        memset(&server, 0, sizeof server);
        server.sin_family = AF_INET;
        server.sin_port = htons(port);
        inet_pton(AF_INET, ip.c_str(), &server.sin_addr);
        if (bind(sockfd_, (struct sockaddr *)&server, sizeof server) < 0) {
            std::cout << "bind fail" << std::endl;
            exit(BindErr);
        }
    }
    void Listen() {
        if (listen(sockfd_, 10) < 0) {
            std::cout << "listen fail" << std::endl;
            exit(ListenErr);
        }
    }
    int Accept(uint16_t *client_port, std::string *client_ip) {
        struct sockaddr_in client;
        bzero(&client, sizeof client);
        socklen_t len = sizeof(client);
        int newfd = accept(sockfd_, (struct sockaddr *)&client, &len);
        if (newfd < 0) return -1;
        char ip[64];
        inet_ntop(AF_INET, &client.sin_addr, ip, sizeof ip);
        *client_port = ntohs(client.sin_port);
        *client_ip = ip;
        return newfd;
    }
    bool Connect(uint16_t server_port, std::string server_ip) {
        struct sockaddr_in server;
        bzero(&server, sizeof server);
        server.sin_family = AF_INET;
        server.sin_port = htons(server_port);
        inet_pton(AF_INET, server_ip.c_str(), &server.sin_addr);
        if (connect(sockfd_, (struct sockaddr *)&server, sizeof(server)) < 0) {
            return false;
        }
        return true;
    }
    ~Socket() { close(sockfd_); }
private:
    int sockfd_;
};

HttpServer.cc

#include <iostream>
#include <unistd.h>
#include "Socket.hpp"
#include <fstream>
#include <vector>
#include <sstream>

const std::string wwwroot = "./wwwroot";

class HttpServer;

class ThreadData {
public:
    ThreadData(int sockfd, HttpServer *ts) : sockfd_(sockfd), ts_(ts) {}
public:
    int sockfd_;
    HttpServer *ts_;
};

class HttpServer {
public:
    HttpServer(uint16_t port, std::string ip) : port_(port), ip_(ip) {}
    void Init() {
        listenfd_.Init();
        listenfd_.Bind(port_, ip_);
        listenfd_.Listen();
    }
    void start() {
        while (1) {
            uint16_t client_port;
            std::string client_ip;
            int sockfd = listenfd_.Accept(&client_port, &client_ip);
            if (sockfd < 0) continue;
            ThreadData *td = new ThreadData(sockfd, this);
            pthread_t tid;
            pthread_create(&tid, nullptr, ThreadRun, td);
        }
    }
    static void *ThreadRun(void *arg) {
        pthread_detach(pthread_self());
        ThreadData *td = (ThreadData *)arg;
        td->ts_->HttpHandler(td->sockfd_);
        delete td;
        return nullptr;
    }
    std::string ReadHtmlContent(std::string &htmlpath) {
        std::string content;
        std::ifstream in(htmlpath.c_str());
        if (!in.is_open()) return "404";
        std::string line;
        while (std::getline(in, line)) {
            content += line;
        }
        in.close();
        return content;
    }
    void HttpHandler(int sockfd) {
        char buffer[1024];
        ssize_t s = read(sockfd, buffer, sizeof buffer - 1);
        std::string request;
        if (s > 0) {
            buffer[s] = 0;
            std::cout << buffer << std::endl;
            request = buffer;
            std::vector<std::string> req_header;
            while (!request.empty()) {
                ssize_t pos = request.find("\r\n", 0);
                if (pos == std::string::npos) break;
                std::string line = request.substr(0, pos);
                if (line.empty()) break;
                req_header.push_back(line);
                request.erase(0, pos + 2);
            }
            std::stringstream ss(req_header[0]);
            std::string method, url, http_version;
            ss >> method >> url >> http_version;
            std::string path = url;
            if (path == "/" || path == "/index.html") {
                path = wwwroot + "/index.html";
            } else {
                path = wwwroot + url;
            }
            std::string text = ReadHtmlContent(path);
            std::string response;
            if (text == "404") {
                std::string body = "<h1>404 Not Found</h1>";
                response = "HTTP/1.0 404 Not Found\r\n";
                response += "Content-Length: " + std::to_string(body.size()) + "\r\n";
                response += "Content-Type: text/html\r\n";
                response += "\r\n";
                response += body;
            } else {
                response = "HTTP/1.0 200 OK\r\n";
                response += "Content-Length: " + std::to_string(text.size()) + "\r\n";
                response += "Content-Type: text/html\r\n";
                response += "\r\n";
                response += text;
            }
            write(sockfd, response.c_str(), response.size());
        }
        close(sockfd);
    }
    ~HttpServer() {}
private:
    Socket listenfd_;
    uint16_t port_;
    std::string ip_;
};

int main(int argc, char *argv[]) {
    uint16_t port = std::stoi(argv[2]);
    std::string ip = argv[1];
    HttpServer *svr = new HttpServer(port, ip);
    svr->Init();
    svr->start();
    return 0;
}

前端测试页面

创建一个简单的 index.html 用于测试跳转功能。注意将硬编码的 IP 替换为 localhost 或 127.0.0.1。

<!DOCTYPE html>
<html lang="en">
<head>
<meta charset="UTF-8">
<title>index</title>
</head>
<body>
<h1>测试页面</h1>
<a href="http://127.0.0.1:8080/hello.html">下一页</a>
</body>
</html>

GET 与 POST 的区别

HTTP 中最常用的方法是 GET 和 POST。我们可以通过表单提交来观察它们的差异。

GET 方法

GET 请求会将参数拼接到 URL 后面。这意味着敏感信息（如密码）会直接暴露在地址栏和服务器日志中。

<form action="http://127.0.0.1:8080/login" method="GET">
银行卡账号：<input type="text" name="user"><br>
银行卡密码：<input type="password" name="pass"><br>
<input type="submit" value="登录">
</form>

实际请求中，URL 会变成类似 http://127.0.0.1:8080/login?user=admin&pass=123456 的形式。

POST 方法

POST 请求将参数放在请求正文（Body）中，相对更安全，适合传输敏感数据。

<form action="http://127.0.0.1:8080/login" method="POST">
银行卡账号：<input type="text" name="user"><br>
银行卡密码：<input type="password" name="pass"><br>
<input type="submit" value="登录">
</form>

此时请求体中才会包含 user=admin&pass=123456，URL 保持不变。

总结

从 TCP 通信到 HTTP 服务器的实现，我们实际上已经打通了网络编程的核心链路：

• 理解了请求是如何被服务器解析的。
• 明白了浏览器和服务器之间到底在'说什么'（结构化字符串）。
• 掌握了 GET 和 POST 的本质区别。

HTTP 不再是黑盒，它只是一个遵循特定规则的字符串协议。只要掌握了报文结构和 Socket 操作，就能灵活构建自己的网络服务。