前言

row_nubmer使用說明： row_number接收到的資料是已經分割槽排序的資料， row_number() OVER (PARTITION BYc ORDERBYd)

description = @Description(
    name = "row_number",
    value = "_FUNC_() - The ROW_NUMBER function assigns a unique number (sequentially, starting "
            + "from 1, as defined by ORDER BY) to each row within the partition."
  ),
 

程式碼路徑

hive版本：2.1.0

row_number對應的原始碼類：org.apache.hadoop.hive.ql.udf.generic.GenericUDAFRowNumber

這個是一個UDAF，hive的udf在FunctionRegistry，對應的包在ql目錄下

system.registerGenericUDAF("row_number", new GenericUDAFRowNumber());

原始碼分析

static class RowNumberBuffer implements AggregationBuffer {

    ArrayList<IntWritable> rowNums;
    int nextRow;

    void init() {
      rowNums = new ArrayList<IntWritable>();
    }

    RowNumberBuffer() {
      init();
      nextRow = 1;
    }

    void incr() {
      rowNums.add(new IntWritable(nextRow++));
    }
  }
 

2、iterate方法

           呼叫RowNumberBuffer.incr()，來一條記錄進行加1

public void iterate(AggregationBuffer agg, Object[] parameters) throws HiveException {
      ((RowNumberBuffer) agg).incr();
    }

3、terminate

          返回最終結果，RowNumberBuffer中的list

public Object terminate(AggregationBuffer agg) throws HiveException {
      return ((RowNumberBuffer) agg).rowNums;
    }