day04

一、            Π值计算案例

1)        设计思路：生成随机数的方法，在单位正方形内随机的产生点，落在单位圆内的点数除以总数=4Π，可以得到Π的值

2)        Mapper实现

生成随机数，判断到圆心的距离，如果距离小于半径，则将一个数存到context内

private static Random rd = new Random();

              public void map(Object key, Text value, Context context) throws IOException, InterruptedException {

                     // TODO Auto-generated method stub

                     int pointNum = Integer.parseInt(value.toString());

                     for (int i = 0; i < pointNum; i++) {

                            //get random number

                            double x = rd.nextDouble();

                            double y = rd.nextDouble();

                            //calculate distance

                            x-=0.5;

                            y-=0.5;

                            double distance = Math.sqrt(x*x + y*y);

                            IntWritable result = new IntWritable(0);

                           

                            if (distance <= 0.5) {

                                   result = new IntWritable(1);

                            }

                            context.write(value, result);

                     }

}

3)        Reducer实现

循环Mapper传入的value，计算出总和，再将值*4得到Π的结果。

private DoubleWritable resule = new DoubleWritable();

              public void reduce(Text key,Iterable<IntWritable> values,Context context) throws IOException, InterruptedException {

                     double pointNum  = Double.parseDouble(key.toString());

                     double sum = 0;

                     for (IntWritable val : values) {

                            sum+=val.get();

                     }

                     resule.set(sum/pointNum*4);

                     context.write(key, resule);

}

二、            日志清洗一案例

1)        设计思路：筛选长度小于11的字段。设计思路：读取文件，切割字符串，判断长度，留下长度大于11的字段

2)        Mapper实现

// 1 获取1行数据

        String line = value.toString();       

        // 2 解析日志

        boolean result = parseLog(line,context);      

        // 3 日志不合法退出

        if (!result) {

            return;

        }    

        // 4 设置key

        k.set(line);  

        // 5 写出数据

        context.write(k, NullWritable.get());

 

3)        Reducer实现

// 1 截取

        String[] fields = line.split(" ");

       

        // 2 日志长度大于11的为合法

        if (fields.length > 11) {

            // 系统计数器

            context.getCounter("map", "true").increment(1);

            return true;

        }else {

            context.getCounter("map", "false").increment(1);

            return false;

        }

 

三、            日志清洗二案例

1)        设计思路：筛选掉日志格式不正确的数据。思路：不正确的格式一般为较长或其他种类的字符较多，所以可以利用字符筛选和长成都筛选。

2)        Mapper实现

// 1 获取1行

        String line = value.toString();       

        // 2 解析日志是否合法

        LogBean bean = pressLog(line);       

        if (!bean.isValid()) {

            return;

        }       

        k.set(bean.toString());       

        // 3 输出

        context.write(k, NullWritable.get());

 

3)        Reducer实现

// 解析日志

    private LogBean pressLog(String line) {

        LogBean logBean = new LogBean();

        // 1 截取

        String[] fields = line.split(" ");  

        if (fields.length > 11) {

            // 2封装数据

            logBean.setRemote_addr(fields[0]);

            logBean.setRemote_user(fields[1]);

            logBean.setTime_local(fields[3].substring(1));

            logBean.setRequest(fields[6]);

            logBean.setStatus(fields[8]);

            logBean.setBody_bytes_sent(fields[9]);

            logBean.setHttp_referer(fields[10]);

           

            if (fields.length > 12) {

                logBean.setHttp_user_agent(fields[11] + " "+ fields[12]);

            }else {

                logBean.setHttp_user_agent(fields[11]);

            }          

            // 大于400，HTTP错误

            if (Integer.parseInt(logBean.getStatus()) >= 400) {

                logBean.setValid(false);

            }

        }else {

            logBean.setValid(false);

        }

        return logBean;

    }

 

四、            共同好友案例

1)        设计思路：我们需要用到逆向思维，要想知道用户之间的共同好友，通过上面的数据我们可以发现B，C，D，E，F，O 有一个共同好友A。A，C，E，K 有一个共同好友B，我们先对数据进行如下处理：A：B，C，D，E，F，O 我们将它输出为<B,A><C,A><D,A><E,A><F,A><O,A>B：A，C，E，K              我们将它输出为<A,B><C,B><E,B><K,B>然后根据对key值相同的value值进行两两组合，得到数据格式为：<A-B,C><A-B,E>接下来我们再进行第二部操作，在map中将key值相同的输出为如下格式:<B-A,C><B-A,E>，表示为B和A之间的公同好友为C、E，然后我们将数据根据key合并得到B-A  C,E。

2)        Mapper实现

protected void map(LongWritable key, Text value, Context context) throws IOException, InterruptedException {

                     String line = value.toString();

                     String[] users = line.split(":");

                     String user = users[0];

                     String[] friends = users[1].split(",");

                     for (String friend : friends) {

                            context.write(new Text(friend), new Text(user));

                     }

              }

String line = value.toString();

                     String[] friend_user = line.split(" ");

                     String friend = friend_user[0];

                     String[] users = friend_user[1].split(",");

                     Arrays.sort(users);

                     System.out.println(users.length);

                     for (int i = 0; i < users.length - 1; i++) {

                            for (int j = i + 1; j < users.length; j++) {

                                   context.write(new Text(users[i] + "-" + users[j]), new Text(friend));

                            }

                     }

Reducer实现

protected void reduce(Text key,Iterable<Text> values,Context context) throws IOException, InterruptedException{

                    

                     StringBuffer sb = new StringBuffer();

                     for(Text user:values){

                            sb.append(user).append(",");

                     }

                           

                     context.write(key,new Text(sb.toString()));                      

              }

protected void reduce(Text key, Iterable<Text> values, Context context)

                            throws IOException, InterruptedException {

                     StringBuffer sb = new StringBuffer();

                     for (Text friend : values) {

                            sb.append(friend).append(" ");

                     }

                     context.write(key, new Text(sb.toString()));

       }

一、            Mapper和Reducer工作原理理解

Mapper自身就相当于循环操作，所以在context进行write操作时，无需在循环，否则会造成输出的数据重复。Recuder原理也一样，在context输入到文件时，不要循环context，每个Reducer输出一次即可。

二、            Mapper和Reducer函数设置

如果Mapper函数体和Reducer、main函数在一个类里，则需要将Mapper和Reducer函数设置为static静态类函数，如果Mapper和Reducer在不同的class内，则可以不设置static静态函数。