MapReduce計數並找到平均值

Question

我想在MapReduce中開發一個程序，該程序從.tbl文件中獲取cust_key和balance值。我已經將2個值連接到字符串中，然后將其發送到Reducer，所以我將計算cust_key並找到每個段的平均余額。這就是為什么我將細分添加為關鍵字。

我想分割字符串並分離兩個值，以便計算保管鍵並求和余額以求平均值。但是splitted array [0]給出了整個字符串，而不是字符串的第一個值。 array [1]引發ArrayoutofBounds異常。我希望這很清楚。

代碼如下

public class MapReduceTest {

        public static class TokenizerMapper extends Mapper<Object, Text, Text, Text>{

         private Text segment = new Text();

         private Text word = new Text();

         private float balance = 0;


         public void map(Object key, Text value, Context context) throws IOException, InterruptedException {
           String[] line = value.toString().split("\\|");

           balance = Float.parseFloat(line[5]);

           String cust_key = line[1];

           int nation = Integer.parseInt(line[3]);

           if((balance > 8000) && ( nation < 15) && (nation > 1)){ 

             segment.set(line[6]);

             //word.set(cust_key+","+balance);

             word.set(cust_key+","+balance);

             context.write(segment,word);
           }
         }

       }

    public static class AvgReducer extends Reducer<Text,Text,Text,Text> {


         Text val = new Text();

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

         String cust_key = "";
         float avg,sum = 0;
         int count = 0;
            for(Text v : values){
                 String[] a = v.toString().trim().split(",");

                 cust_key +=a[0];

            }

            val.set(cust_count);

            context.write(key, val);

     }

   }

輸入數據

8794|Customer#000008794|6dnUgJZGX73Kx1idr6|18|28-434-484-9934|7779.30|HOUSEHOLD|deposits detect furiously even requests. furiously ironic packages are slyly into th
8795|Customer#000008795|oA1cLUtWOAIFz5Douypbq1jHv glSE|9|19-829-732-8102|9794.80|BUILDING|totes. blithely unusual theodolites integrate carefully ironic foxes. unusual excuses cajole carefully carefully fi
8796|Customer#000008796|CzCzpV7SDojXUzi4165j,xYJuRv wZzn grYsyZ|24|34-307-411-6825|4323.03|AUTOMOBILE|s. pending, bold accounts above the sometimes express accounts 
8797|Customer#000008797|TOWDryHNNqp8bvgMW6 FAhRoLyG1ldu2bHcJCM6|2|12-517-522-5820|219.78|FURNITURE|ly bold pinto beans can nod blithely quickly regular requests. fluffily even deposits ru
8798|Customer#000008798|bIegyozQ5kzprN|15|25-472-647-6270|6832.96|AUTOMOBILE|es-- silent instructions nag blithely

堆棧跟蹤

java.lang.Exception: java.lang.ArrayIndexOutOfBoundsException: 1
        at org.apache.hadoop.mapred.LocalJobRunner$Job.runTasks(LocalJobRunner.java:462)
        at org.apache.hadoop.mapred.LocalJobRunner$Job.run(LocalJobRunner.java:529)
Caused by: java.lang.ArrayIndexOutOfBoundsException: 1
        at MapReduceTest$AvgReducer.reduce(MapReduceTest.java:69)
        at MapReduceTest$AvgReducer.reduce(MapReduceTest.java:1)
        at org.apache.hadoop.mapreduce.Reducer.run(Reducer.java:171)
        at org.apache.hadoop.mapred.ReduceTask.runNewReducer(ReduceTask.java:627)
        at org.apache.hadoop.mapred.ReduceTask.run(ReduceTask.java:389)
        at org.apache.hadoop.mapred.LocalJobRunner$Job$ReduceTaskRunnable.run(LocalJobRunner.java:319)
        at java.util.concurrent.Executors$RunnableAdapter.call(Unknown Source)
        at java.util.concurrent.FutureTask.run(Unknown Source)
        at java.util.concurrent.ThreadPoolExecutor.runWorker(Unknown Source)
        at java.util.concurrent.ThreadPoolExecutor$Worker.run(Unknown Source)
        at java.lang.Thread.run(Unknown Source)
17/04/12 18:40:33 INFO mapreduce.Job: Job job_local806960399_0001 running in uber mode : false
17/04/12 18:40:33 INFO mapreduce.Job:  map 100% reduce 0%
17/04/12 18:40:33 INFO mapreduce.Job: Job job_local806960399_0001 failed with state FAILED due to: NA
17/04/12 18:40:33 INFO mapreduce.Job: Counters: 35

更新

減速器

    public static class AvgReducer extends Reducer<Text,Text,Text,Text> {

    Logger log = Logger.getLogger(AvgReducer.class.getName());

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

            float sumBalance=0,avgBalance = 0;

            int cust_count = 1;

            for(Text v : values){
               String[] a = v.toString().trim().split(",");

               //c2 += " i "+i+" "+a[0]+"\n";

               sumBalance +=Float.parseFloat(a[a.length-1]);

               cust_count++;
            }

            avgBalance = sumBalance / cust_count;


            context.write(key,new Text(avgBalance+" "+cust_count));

     }

   }

堆棧跟蹤

java.lang.Exception: java.lang.NumberFormatException: For input string: "8991.715 289"

提前致謝。

Answer 1

Pig運行MapReduce（如果以這種方式配置）。 它比使用MapReduce亂糟糟得多，並且安裝在主要的Hadoop發行版中。

A = LOAD 'test.txt' USING PigStorage('|') AS (f1:int,customer_key:chararray,f3:chararray,nation:int,f5:chararray,balance:float,segment:chararray,f7:chararray);
filtered = FILTER A BY balance > 8000 AND (nation > 1 AND nation < 15);
X = FOREACH filtered generate segment,customer_key,balance;

並輸出

\d X
(BUILDING,Customer#000008795,9794.8)

不確定您是否真的想要平均值，這里只有一個元素，但是您需要在segment和customer_key上使用GROUP BY ，然后可以輕松使用AVG函數 。

---

如果您更熟悉SQL，那么Hive可能也是一種更簡單的方法。

（除非另有配置，否則也通過MapReduce運行）

CREATE EXTERNAL TABLE IF NOT EXISTS records (
    f1 INT,
    customer_key STRING, 
    f3 STRING, 
    nation INT,
    f5 STRING,
    balance FLOAT,
    f8 STRING
) ROW FORMAT DELIMETED 
FIELDS TERMINATED BY '|'
LOCATION 'hdfs://path/test.txt';

那會是這樣

SELECT segment, customer_key, AVG(balance)
FROM records
WHERE balance > 8000 AND nation > 1 AND nation < 15
GROUP BY segment, customer_key;

---

我將介紹Apache Spark示例，但Spark SQL本質上就是該Hive查詢。

Answer 2

如果您真的想在Java MapReduce中嘗試此操作，請嘗試標准化輸入並快速捕獲錯誤。

返回丟棄有問題的記錄

     public void map(Object key, Text value, Context context) throws IOException, InterruptedException {
       float balance = 0.0;
       String custKey = "";
       String segment = "";
       int nation = 0;

       String[] line = value.toString().split("\\|");
       if (line.length < 7) { 
           System.err.println("map: Not enough records");
           return;
       }
       cust_key = line[1];
       try {
           nation = Integer.parseInt(line[3]);
           balance = Float.parseFloat(line[5]);
       } catch (NumberFormatException e) {
           e.printStackTrace();
           return;
       }

       if(balance > 8000 && (nation < 15 && nation > 1)){ 
         segment.set(line[6]);
         word.set(cust_key + "\t" + balance);
         context.write(segment,word);
       }
  }

然后，如果您試圖編寫類似的輸出記錄，則reduce理想情況下應生成相同的格式

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

        float sumBalance=0
        int count = 0;

        for(Text v : values){
           String[] a = v.toString().trim().split("\t");
           if (a.length < 2) {
               System.err.println("reduce: Not enough records");
               continue;
           }

           sumBalance += Float.parseFloat(a[1]);
           count++;
        }

        float avgBalance = count <= 1 ? sumBalance : sumBalance / count;

        context.write(key,new Text(avgBalance + "\t" + count));

 }

（代碼未經測試）