高并发小设计思维

今天看sun的HttpMessages.java文件，虽然文件很小，但是对我的启发很大。
前面定义了大量的HTTP常量：
Java代码

...   
...   
private static final String STATUS_305 = "Use Proxy";   
private static final String STATUS_307 = "Temporary Redirect";   
private static final String STATUS_400 = "Bad Request";   
private static final String STATUS_401 = "Unauthorized";   
private static final String STATUS_402 = "Payment Required";   
private static final String STATUS_403 = "Forbidden";   
private static final String STATUS_404 = "Not Found";   
private static final String STATUS_405 = "Method Not Allowed";   
private static final String STATUS_406 = "Not Acceptable";   
private static final String STATUS_407 = "Proxy Authentication Required";   
private static final String STATUS_408 = "Request Timeout";   
...   
...  

    ...
    ...
    private static final String STATUS_305 = "Use Proxy";
    private static final String STATUS_307 = "Temporary Redirect";
    private static final String STATUS_400 = "Bad Request";
    private static final String STATUS_401 = "Unauthorized";
    private static final String STATUS_402 = "Payment Required";
    private static final String STATUS_403 = "Forbidden";
    private static final String STATUS_404 = "Not Found";
    private static final String STATUS_405 = "Method Not Allowed";
    private static final String STATUS_406 = "Not Acceptable";
    private static final String STATUS_407 = "Proxy Authentication Required";
    private static final String STATUS_408 = "Request Timeout";
    ...
    ...

也可以趁机从这里查看HTTP返回值的意义^_^。然后定义了一个java.util.concurrent.ConcurrentHashMap的变量httpStatusCodeMappings，再往里面添加刚才定义的常量（这个操作放在了一个静态块里），如：
Java代码

static {   
        httpStatusCodeMappings.put("sc.100", STATUS_100);   
        httpStatusCodeMappings.put("sc.101", STATUS_101);   
        httpStatusCodeMappings.put("sc.200", STATUS_200);   
        httpStatusCodeMappings.put("sc.201", STATUS_201);   
        httpStatusCodeMappings.put("sc.202", STATUS_202);   
        httpStatusCodeMappings.put("sc.203", STATUS_203);   
  ...   
  ...  

static {
        httpStatusCodeMappings.put("sc.100", STATUS_100);
        httpStatusCodeMappings.put("sc.101", STATUS_101);
        httpStatusCodeMappings.put("sc.200", STATUS_200);
        httpStatusCodeMappings.put("sc.201", STATUS_201);
        httpStatusCodeMappings.put("sc.202", STATUS_202);
        httpStatusCodeMappings.put("sc.203", STATUS_203);
  ...
  ...

最重要的就是这个方法了，先看代码：
Java代码

public static String getMessage(int status) {   

        // Return the status message for the most frequently used status   
        // codes directly, without any lookup   
        switch (status) {   
            case 200: return STATUS_200;   
            case 302: return STATUS_302;   
            case 400: return STATUS_400;   
            case 404: return STATUS_404;   
    }   

        return httpStatusCodeMappings.get("sc."+ status);   
    }  

public static String getMessage(int status) {

        // Return the status message for the most frequently used status
        // codes directly, without any lookup
        switch (status) {
            case 200: return STATUS_200;
            case 302: return STATUS_302;
            case 400: return STATUS_400;
            case 404: return STATUS_404;
	}

        return httpStatusCodeMappings.get("sc."+ status);
    }

这段代码的用途很简单，就是要返回状态码对应的Message，而这些这些消息都已经存放在httpStatusCodeMappings变量里了，为了提高访问常用的几个返回码的Message，它直接用了一个静态块，而不去lookup那个map了，这就是高手！


sw1982 写道
...lookup 一下hashmap真的那么低效吗？ 建议复习下数据结构哦，你这些总结是没错，可是很表面


典型的没有写过高并发程序的思维方式，明明可以节约的地方，仅仅几行代码就可以优化，偏偏不做。

hashmap再快，也比case 一个 整型满上1w倍。

性能，是一点一点挤牙膏挤出来的，哪能到处浪费啊。