如果同一表中存在某些记录，则设置列值

Question

如果表包含某种类型的记录，我如何为表创建一个视图，其中包含表记录的子集、表的所有列以及附加的“标志”列，其值设置为“X”？例如，考虑以下关系表Relations，其中类型的值代表H'-human, 'D'-dog：

 id | type | relation | related
--------------------------------
 H1 |  H   | knows    | D2 
 H1 |  H   | owns     | D2 
 H2 |  H   | knows    | D1 
 H2 |  H   | owns     | D1
 H3 |  H   | knows    | D1
 H3 |  H   | knows    | D2
 H3 |  H   | treats   | D1
 H3 |  H   | treats   | D2
 D1 |  D   | bites    | H3
 D2 |  D   | bites    | H3

此表中可能没有任何特定的记录顺序。

我试图创建一个视图Humans，它将包含来自Relations 的所有人对狗knows 关系以及所有Relations 的列和附加列isOwner，如果存储'X'特定关系中的人拥有某人：

 id | type | relation | related | isOwner
------------------------------------------
 H1 |  H   | knows    |    D2   |    X
 H2 |  H   | knows    |    D1   |    X
 H3 |  H   | knows    |    D1   |    

但为此苦苦挣扎。你知道一种方法吗，最好是在一个CREATE VIEW 电话中，或者真的有什么方法吗？

Answer 1

CREATE VIEW vHumanDogRelations
AS
SELECT
    id,
    type,
    relation,
    related,
    -- Consider using a bit 0/1 instead
    CASE 
        WHEN EXISTS (
            SELECT 1
            FROM Relations
            WHERE 
                id = r.id
                AND related = r.related -- Owns someone or this related only?
                AND relation = 'owns'    
        ) THEN 'X'
        ELSE ''
    END AS isOwner
FROM Relations r
WHERE 
    relation = 'knows'
    AND type = 'H'
    AND related = 'D';

Answer 2

您应该可以将以下select 放入视图定义中

select *, case when exists 
(select * from Relations where id = r.id and relation= 'owns') then 'X'
else '' end as isOwner 
from Relations r

Answer 3

您也可以使用 PIVOT 来实现所需的结果。我将详细解释该方法，因为最终的查询可能看起来令人困惑。

首先，从Relation 派生一个子集，其中type 是H 和relation 或者knows 或owns，将owns 值替换为X：

SELECT
  id,
  type,
  relation = CASE relation WHEN 'owns' THEN 'X' ELSE relation END,
  related
FROM Relations
WHERE type = 'H'
  AND relation IN ('knows', 'owns')

根据你的例子，你会得到这个：

id  type  relation  related
--  ----  --------  -------
H1   H    knows     D2
H1   H    owns      D2
H2   H    knows     D1
H2   H    owns      D1
H3   H    knows     D1
H3   H    knows     D2

接下来，将此 PIVOT 子句应用于第一个查询的结果：

PIVOT (
  MAX(relation) FOR relation IN (knows, X)
) AS p

它将具有相同id, type, related 值的行分组为一行并将relation 拆分为两列，knows 和X：

id  type  related  knows  X
--  ----  -------  -----  ----
H1   H    D2       knows  X
H2   H    D1       knows  X
H3   H    D1       knows  NULL
H3   H    D2       knows  NULL

此时您只需要重新排列主SELECT子句中为输出设置的列，将knows重命名为relation和X重命名为isOwner：

SELECT
  id,
  type,
  relation = knows,
  related,
  isOwner = X
...

输出：

id  type  relation  related  isOwner
--  ----  --------  -------  -------
H1   H    knows     D2       X
H2   H    knows     D1       X
H3   H    knows     D1       NULL
H3   H    knows     D2       NULL

当然，如果有必要，NULL 可以很容易地用空字符串替换。

最后一点可能是将这个额外的过滤器添加到主查询中：

WHERE knows IS NOT NULL

以防万一有人在不认识狗的情况下拥有狗（并且您不希望在输出中包含这些狗）。

所以，完整的查询应该是这样的：

SELECT
  id,
  type,
  relation = knows,
  related,
  isOwner = X
FROM (
  SELECT
    id,
    type,
    relation = CASE relation WHEN 'owns' THEN 'X' ELSE relation END,
    related
  FROM Relations
  WHERE type = 'H'
    AND relation IN ('knows', 'owns')
) AS s
PIVOT (
  MAX(relation) FOR relation IN (knows, X)
) AS p
WHERE knows IS NOT NULL
;

此解决方案的 SQL Fiddle 演示可用 here。