mysql表分为两类：

1. 基本维度表：id（自增），其它值字段。比如dimension_browser、dimension_platform
2. 结果表：联合主键（包含多个单一维度的id值），其它统计结果的字段。比如：stats_user

如果基本维度表中有对应的数据，则获取对应的id；如果没有对应信息则先插入基本维度数据表中，获取id后，再添加到结果表中。

TransformerOutputFormat类中的RecordWriter用于向MySQL输出。

1. cn.itbaizhan.tranformer.mr.NewInstallUserReducer

   context.write(key,mapWritableValue); //Ctrl+点击

2. org.apache.hadoop.mapreduce.TaskInputOutputContext

   void write(KEYOUT var1, VALUEOUT var2) throws IOException, InterruptedException; //Ctrl+Alt+B

3. 选择org.apache.hadoop.mapreduce.lib.chain.ChainReduceContextImpl

   xxxxxxxxxx
   private final RecordWriter<KEYOUT, VALUEOUT> rw;
   public ChainReduceContextImpl(ReduceContext<KEYIN, VALUEIN, KEYOUT, VALUEOUT> base, RecordWriter<KEYOUT, VALUEOUT> output, Configuration conf) {
     this.base = base;
     this.rw = output;
     this.conf = conf;
   }
   public void write(KEYOUT key, VALUEOUT value) throws IOException, InterruptedException {
     this.rw.write(key, value);//Ctrl+单击
   }
   

4. org.apache.hadoop.mapreduce.RecordWriter

   xxxxxxxxxx
   public abstract class RecordWriter<K, V> {
     public RecordWriter() {
     }
     public abstract void write(K var1, V var2) throws IOException, InterruptedException;//Ctrl+Alt+B
     public abstract void close(TaskAttemptContext var1) throws IOException, InterruptedException;
   }
   

5. 选择TransformerRecordWriter,找到了我们自定义类的write方法中。

   xxxxxxxxxx
   @Override
   /**
    * 当reduce任务输出数据是，由计算框架自动调用。把reducer输出的数据写到mysql中
    * key:StatsUserDimension类对象
    * value:MapWritableValue类的对象
    */
   public void write(BaseDimension key, BaseStatsValueWritable value) throws IOException, InterruptedException {
       ......
   }
   

6. TransformerRecordWriter类的write方法实现原理分析

   

7. StatsUserNewInstallUserCollector

8. DimensionConverterImpl