为什么 OLAP 子查询在视图中使用时会出现问题，但在 DB2 for i 中却可以正常工作？

Question

我试图将行转换为有限数量的列，同时跳过空白行，并返回非空白行的计数。我能够解决这个限制，所以我更感兴趣的是为什么以下内容作为独立语句工作，但在前面加上

CREATE VIEW TEST AS

？

时失败

with temp (id,line) as (
  values (1, 'line a'),
         (1, 'line b'),
         (1, null),
         (1, 'line d'),
         (1, 'line e'),
         (1, 'line f'),
         (2, null),
         (2, 'line b'),
         (2, 'line c') 
),
temp2 as (
  select id, line,
   row_number() over (partition by id) rn,
   sum(1) over (partition by id) rows
  from temp
  where line is not null
),
temp3 as (
  select id,
   min(case when rn = 1 then line end) line1,
   min(case when rn = 2 then line end) line2,
   min(case when rn = 3 then line end) line3,
   rows
  from temp2
  group by id,rows
)
select * from temp3

按原样运行上面的代码给出了我的预期结果：

ID  LINE1   LINE2   LINE3   ROWS
1   line a  line b  line d  5
2   line b  line c       -  2

但是当前面加上

create view test as

时，会出现以下错误：

SQL State: 42803 Vendor Code: -122 Message: [SQL0122] Column ID or expression in SELECT list not valid. Cause . . . . . :   One of the following has occurred: -- The statement contains column name ID and an aggregate function in the SELECT clause and no GROUP BY clause is specified. -- Column name ID is specified in the SELECT clause but not in the GROUP BY clause. -- An expression is specified in the SELECT clause but not in the GROUP BY clause. -- A column or expression that is specified in the ORDER BY clause, but not in the SELECT clause, does not conform to the grouping rules listed above. Recovery  . . . :   Do one of the following and try the request again: -- If a GROUP BY clause is required, make certain that all columns or expressions in the SELECT list and ORDER BY clause are also in the GROUP BY clause. -- If a GROUP BY clause is not needed, the SELECT list and ORDER BY clause should not contain aggregate functions with column names.

我知道该错误具有误导性，因为如果我从

rows

中删除对

temp3

的引用，它就可以正常运行。因此，这与 OLAP 语句 (

sum(1) over (partition by id) rows

) 作为子查询的一部分有关，但我不明白为什么它以一种方式运行正常，而另一种则不然。我无法找到文档为什么不应该按原样运行。

Answer 1

我遇到了同样的问题，并通过与相关表的 FK 进行连接和分组来使其工作。别问我为什么。

我之前已经有过表现得很有趣的复杂查询，并且必须让它与这种黑客一起工作。 顺便说一句，我正在使用 DB2 7.4。

这有效：

CREATE OR REPLACE VIEW V_PERSON_KEY_CHANGE AS 
WITH changes AS (
  SELECT 
    personId, 
    dateColumn, 
    date_key, 
    CASE 
      WHEN key = LAG(dateColumn) OVER (PARTITION BY personId ORDER BY dateColumn) 
      THEN 0 
      ELSE 1 
    END AS changed
  FROM person_dates
),
groups AS (
  SELECT 
    personId, 
    dateColumn, 
    date_key, 
    SUM(changed) OVER (PARTITION BY personId ORDER BY dateColumn) AS group_id
  FROM changes
)
SELECT 
    personNames.personId,
    personNames.personName, 
    MIN(dateColumn) as date_from, 
    MAX(dateColumn) as date_to, 
    date_key
FROM groups
INNER JOIN personNames ON groups.personId = personNames.personId 
GROUP BY 
    personNames.personId,
    personNames.personName, 
    date_key, 
    group_id

这不起作用：

CREATE OR REPLACE VIEW V_PERSON_KEY_CHANGE AS 
WITH changes AS (
  SELECT 
    personId, 
    dateColumn, 
    date_key, 
    CASE 
      WHEN key = LAG(dateColumn) OVER (PARTITION BY personId ORDER BY dateColumn) 
      THEN 0 
      ELSE 1 
    END AS changed
  FROM person_dates
),
groups AS (
  SELECT 
    personId, 
    dateColumn, 
    date_key, 
    SUM(changed) OVER (PARTITION BY personId ORDER BY dateColumn) AS group_id
  FROM changes
)
SELECT 
    personId,
    MIN(dateColumn) as date_from, 
    MAX(dateColumn) as date_to, 
    date_key
FROM groups
GROUP BY 
    personId,
    date_key, 
    group_id