hive grouping函数_Merge函数

Grouping函数:
可以接受一列，返回0或1。如果列值为空，则返回1，非空则返回0。它只能在rollup和或cube函数中使用，因为在统计中显示“全部”的那一项统计值时，那一项的标签通常是空的，这时grouping就非常有用，还可以在grouping的基础上进行decode, case等进行美化。

SQL> select grouping(index_type) g_ind, grouping(status) g_st, index_type, status, count(*)
  2  from t group by rollup(index_type, status) order by 1, 2;

     G_IND       G_ST INDEX_TYPE                  STATUS     COUNT(*)
———- ———- ————————— ——– ———-
         0          0 LOB                         N/A               1
         0          0 LOB                         VALID           572
         0          0 FUNCTION-BASED NORMAL       VALID            17
         0          0 FUNCTION-BASED DOMAIN       VALID             1
         0          0 IOT – TOP                   VALID           115
         0          0 CLUSTER                     VALID            10
         0          0 NORMAL                      VALID          4557
         0          0 NORMAL                      N/A              56
         0          0 DOMAIN                      VALID             1
         0          0 BITMAP                      VALID             8
         0          0 BITMAP                      N/A               7
         0          1 IOT – TOP                                   115
         0          1 FUNCTION-BASED DOMAIN                         1
         0          1 DOMAIN                                        1
         0          1 CLUSTER                                      10
         0          1 BITMAP                                       15
         0          1 FUNCTION-BASED NORMAL                        17
         0          1 NORMAL                                     4613
         0          1 LOB                                         573
         1          1                                            5345

20 rows selected

Grouping_id函数：
比grouping还强点，可以接收多个列，这几个列都不为空时，返回0，只要有一个为空，则返回1，如果都为空，则返回3。
下面这条查询把grouping_id牛刀杀鸡，当做grouping用了，可以看到结果跟用grouping是完全一样。

SQL> select grouping_id(index_type) g_ind, grouping_id(status) g_st, index_type, status, count(*)
  2  from t group by rollup(index_type, status) order by 1, 2;

     G_IND       G_ST INDEX_TYPE                  STATUS     COUNT(*)
———- ———- ————————— ——– ———-
         0          0 LOB                         N/A               1
         0          0 LOB                         VALID           572
         0          0 FUNCTION-BASED NORMAL       VALID            17
         0          0 FUNCTION-BASED DOMAIN       VALID             1
         0          0 IOT – TOP                   VALID           115
         0          0 CLUSTER                     VALID            10
         0          0 NORMAL                      VALID          4557
         0          0 NORMAL                      N/A              56
         0          0 DOMAIN                      VALID             1
         0          0 BITMAP                      VALID             8
         0          0 BITMAP                      N/A               7
         0          1 IOT – TOP                                   115
         0          1 FUNCTION-BASED DOMAIN                         1
         0          1 DOMAIN                                        1
         0          1 CLUSTER                                      10
         0          1 BITMAP                                       15
         0          1 FUNCTION-BASED NORMAL                        17
         0          1 NORMAL                                     4613
         0          1 LOB                                         573
         1          1                                            5345

20 rows selected

下面这条查询才体现了grouping_id的作用。

SQL> select grouping_id(index_type, status) g_st, index_type, status, count(*)
  2  from t group by rollup(index_type, status) order by 1, 2;

      G_ST INDEX_TYPE                  STATUS     COUNT(*)
———- ————————— ——– ———-
         0 BITMAP                      VALID             8
         0 BITMAP                      N/A               7
         0 CLUSTER                     VALID            10
         0 DOMAIN                      VALID             1
         0 FUNCTION-BASED DOMAIN       VALID             1
         0 FUNCTION-BASED NORMAL       VALID            17
         0 IOT – TOP                   VALID           115
         0 LOB                         N/A               1
         0 LOB                         VALID           572
         0 NORMAL                      N/A              56
         0 NORMAL                      VALID          4557
         1 BITMAP                                       15
         1 CLUSTER                                      10
         1 DOMAIN                                        1
         1 FUNCTION-BASED DOMAIN                         1
         1 FUNCTION-BASED NORMAL                        17
         1 IOT – TOP                                   115
         1 LOB                                         573
         1 NORMAL                                     4613
         3                                            5345

20 rows selected

根据grouping_id接收不同的列的组合，可以美化出很神奇的结果。