SQL ROLLUP 还是联合？

Question

我正在尝试获取条目总数，但不幸的是我不相信汇总会是最佳选择：

SELECT BUSINESS_STATUS_NAME,
  PENDING_ITEMS,
  DATAGROUP
FROM PAYMENTS
WHERE STATUS LIKE '%PROCESS%';

这会产生：

BUSINESS_STATUS_NAME     PENDING_ITEMS     DATAGROUP
PROCESSING DATA          34                PRODUCT
PROCESSING INS           40                SERVICE

我想得到下面的总计，但 ROLLUP 给了我小计，因为它包括数据组列。我只需要待处理项目的总计，但我需要显示数据组。使用 sum(pending_items) 选择查询 UNION 会更好吗？

BUSINESS_STATUS_NAME     PENDING_ITEMS     DATAGROUP
PROCESSING DATA          34                PRODUCT
PROCESSING INS           40                SERVICE
GRAND TOTAL **           74

谢谢！

Answer 1

为了清晰和性能，我会使用ROLLUP。

假设你有一个这样的示例表：

create table payments (business_status_name, pending_items, datagroup) as (
    select 'PROCESSING DATA', 10, 'PRODUCT' from dual union all
    select 'PROCESSING DATA',  5, 'PRODUCT' from dual union all
    select 'PROCESSING DATA',  2, 'SERVICE' from dual union all
    select 'PROCESSING INS',  10, 'SERVICE' from dual union all
    select 'PROCESSING INS',  10, 'SERVICE' from dual union all
    select 'PROCESSING INS',  10, 'PRODUCT' from dual
)

这是ROLLUP的一种方式（注意括号改变分组逻辑）：

SELECT BUSINESS_STATUS_NAME,
       SUM(PENDING_ITEMS) as PENDING_ITEMS,
       DATAGROUP
FROM PAYMENTS
GROUP BY ROLLUP ((BUSINESS_STATUS_NAME, DATAGROUP))

结果：

BUSINESS_STATUS PENDING_ITEMS DATAGRO
--------------- ------------- -------
PROCESSING INS             10 PRODUCT
PROCESSING INS             20 SERVICE
PROCESSING DATA            15 PRODUCT
PROCESSING DATA             2 SERVICE
                           47

计划：

---------------------------------------------------------------------------------
| Id  | Operation            | Name     | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------
|   0 | SELECT STATEMENT     |          |     6 |   186 |     4  (25)| 00:00:01 |
|   1 |  SORT GROUP BY ROLLUP|          |     6 |   186 |     4  (25)| 00:00:01 |
|   2 |   TABLE ACCESS FULL  | PAYMENTS |     6 |   186 |     3   (0)| 00:00:01 |
---------------------------------------------------------------------------------

这是UNION ALL:

SELECT BUSINESS_STATUS_NAME,
       SUM(PENDING_ITEMS) as PENDING_ITEMS,
       DATAGROUP
FROM PAYMENTS
GROUP BY BUSINESS_STATUS_NAME, DATAGROUP
UNION ALL
SELECT NULL, SUM(PENDING_ITEMS), NULL
FROM PAYMENTS;

结果和ROLLUP一样：

BUSINESS_STATUS PENDING_ITEMS DATAGRO
--------------- ------------- -------
PROCESSING INS             20 SERVICE
PROCESSING INS             10 PRODUCT
PROCESSING DATA            15 PRODUCT
PROCESSING DATA             2 SERVICE
                           47

计划不太好，用TWO FULL SCANS：

--------------------------------------------------------------------------------
| Id  | Operation           | Name     | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |          |     7 |   199 |     7  (58)| 00:00:01 |
|   1 |  UNION-ALL          |          |       |       |            |          |
|   2 |   HASH GROUP BY     |          |     6 |   186 |     4  (25)| 00:00:01 |
|   3 |    TABLE ACCESS FULL| PAYMENTS |     6 |   186 |     3   (0)| 00:00:01 |
|   4 |   SORT AGGREGATE    |          |     1 |    13 |            |          |
|   5 |    TABLE ACCESS FULL| PAYMENTS |     6 |    78 |     3   (0)| 00:00:01 |
--------------------------------------------------------------------------------

这当然只是一个小例子，有几条记录，没有索引，......所以实际表上的情况可能会有所不同，但我仍然相信ROLLUP 应该比UNION ALL 表现更好。

在更简单的情况下，与您的情况完全一样，这将是两种方法的计划：

SELECT BUSINESS_STATUS_NAME,
       SUM(PENDING_ITEMS) as PENDING_ITEMS,
       DATAGROUP
FROM PAYMENTS
GROUP BY ROLLUP ((BUSINESS_STATUS_NAME, DATAGROUP))

---------------------------------------------------------------------------------
| Id  | Operation            | Name     | Rows  | Bytes | Cost (%CPU)| Time     |
---------------------------------------------------------------------------------
|   0 | SELECT STATEMENT     |          |     2 |    62 |     4  (25)| 00:00:01 |
|   1 |  SORT GROUP BY ROLLUP|          |     2 |    62 |     4  (25)| 00:00:01 |
|   2 |   TABLE ACCESS FULL  | PAYMENTS |     2 |    62 |     3   (0)| 00:00:01 |
---------------------------------------------------------------------------------

SELECT BUSINESS_STATUS_NAME,
       PENDING_ITEMS,
       DATAGROUP
FROM PAYMENTS
UNION ALL 
SELECT NULL, 
       SUM(PENDING_ITEMS),
       NULL
FROM PAYMENTS    

--------------------------------------------------------------------------------
| Id  | Operation           | Name     | Rows  | Bytes | Cost (%CPU)| Time     |
--------------------------------------------------------------------------------
|   0 | SELECT STATEMENT    |          |     3 |    75 |     6  (50)| 00:00:01 |
|   1 |  UNION-ALL          |          |       |       |            |          |
|   2 |   TABLE ACCESS FULL | PAYMENTS |     2 |    62 |     3   (0)| 00:00:01 |
|   3 |   SORT AGGREGATE    |          |     1 |    13 |            |          |
|   4 |    TABLE ACCESS FULL| PAYMENTS |     2 |    26 |     3   (0)| 00:00:01 |
--------------------------------------------------------------------------------

ROLLUP 仍然有一个更好的计划，使用单表扫描。

Answer 2

你可以使用rollup，但是你需要一个聚合查询：

SELECT BUSINESS_STATUS_NAME,
       SUM(PENDING_ITEMS) as PENDING_ITEMS,
       DATAGROUP
FROM PAYMENTS
WHERE STATUS LIKE '%PROCESS%'
GROUP BY ROLLUP (BUSINESS_STATUS_NAME, DATAGROUP);

我怀疑这与union all 之间存在性能差异。但是请注意，这可以保证将汇总行作为结果集中的最后一行。