几个表上的 FULL JOIN

Question

我有一个项目清单，按存款、库存、采购订单、销售订单、消费或生产预测等。我想为每个项目设置一个查询返回给我所有这些信息，存款由订金。显然，并不是说每个项目都有这个信息。

例如，考虑以下表格：

T1
|    REF |
|--------|
|      1 |

T2
|    REF |  DEPOT |
|--------|--------|
|      1 |      A |
|      1 |      B |

T3
|    REF |  DEPOT |
|--------|--------|
|      1 |      A |
|      1 |      C |

T4
|    REF |  DEPOT |
|--------|--------|
|      1 |      B |
|      1 |      C |
|      1 |      D |

如果我只取前三个表（刚开始），我还没有找到更好的方法：

SELECT T1.ref AS T1ref
  , T2.ref AS T2ref, T2.depot AS T2depot
  , T3.ref AS T3ref, T3.depot AS T3depot
FROM T1
  LEFT JOIN T2
    ON T2.ref = T1.ref
  FULL JOIN T3
    ON T3.ref = T1.ref
    AND T3.depot = T2.depot

输出：

|  T1REF |  T2REF | T2DEPOT |  T3REF | T3DEPOT |
|--------|--------|---------|--------|---------|
|      1 |      1 |       A |      1 |       A |
|      1 |      1 |       B | (null) |  (null) |
| (null) | (null) |  (null) |      1 |       C |

我想要什么：

|  T1REF |  T2REF | T2DEPOT |  T3REF | T3DEPOT |
|--------|--------|---------|--------|---------|
|      1 |      1 |       A |      1 |       A |
|      1 |      1 |       B | (null) |  (null) |
|      1 | (null) |  (null) |      1 |       C |

必须有一个干净的方法来做到这一点，但我没有找到任何东西。而且很难找到这方面的资料。

有人有提示吗？

sqlfiddle：http://sqlfiddle.com/#!3/19014/2

谢谢你， 大卫。

编辑：
T4：

|  T1REF |  T2REF | T2DEPOT |  T3REF | T3DEPOT |  T4REF | T4DEPOT |
|--------|--------|---------|--------|---------|--------|---------|
|      1 |      1 |       A |      1 |       A | (null) |  (null) |
|      1 |      1 |       B | (null) |  (null) |      1 |       B |
|      1 | (null) |  (null) |      1 |       C |      1 |       C |
|      1 | (null) |  (null) | (null) |  (null) |      1 |       D |

我应该用更好的名称来命名我的表：T1 = 物品，T2 = 库存，T3 = 购买，T4 = 销售。所以，T1 将始终拥有所有的 refs，以及我需要的许多其他信息。

Answer 1

您可以使用coalesce() 修复您的特定查询：

SELECT coalesce(T1.ref, t2.ref, t3.ref) AS T1ref

但是，我发现从我关心的列表和组合开始，而不是使用full outer join。在这种情况下，您似乎关心所有表格中的t1.ref 和depots。也许这更接近您真正想做的事情：

SELECT t1ref.ref,
       T2.ref AS T2ref, T2.depot AS T2depot,
       T3.ref AS T3ref, T3.depot AS T3depot
FROM (select ref from T1 union
      select ref from T2 union
      select ref from T3
     ) t1ref cross join
     (select depot from T2 union
      select depot from t3
     ) d LEFT JOIN T2
     ON T2.ref = T1ref.ref and
        t2.depot = d.depot LEFT JOIN
     T3
     ON T3.ref = T1ref.ref AND
        T3.depot = d.depot --OR T2.depot IS NULL)

Answer 2

Gordon Linoff 所述的所有 4 个表和 Coalesce 的示例

;with[T1]([REF])as(
       select * from (values(1),(2))[A]([REF])
),
[T2]([REF],[DEPOT])as(
       select*from(values
             (1,'A'),
             (1,'B'),
             (1,'E')
       )[a]([REF],[DEPOT])
),
[T3]([REF],[DEPOT])as(
       select*from(values
             (1,'A'),
             (1,'C')
       )[a]([REF],[DEPOT])
),
[T4]([REF],[DEPOT])as(
       select*from(values
             (1,'B'),
             (1,'C'),
             (1,'D'),
             (1,'E')
       )[a]([REF],[DEPOT])
)

select
*
from [T1]
outer apply (
      select
            T2.REF T2REF,
            T2.DEPOT T2DEPOT,
            T3.REF T3REF,
            T3.DEPOT T3DEPOT,
            T4.REF T4REF,
            T4.DEPOT T4DEPOT
      from T2
      full outer join T3 on
            T2.REF = T3.REF and
            T2.DEPOT = T3.DEPOT
      full outer join T4 on   
            COALESCE(T3.REF,T2.REF) = T4.REF and
            COALESCE(T3.DEPOT,T2.DEPOT) = T4.DEPOT
      where
        COALESCE(T2.REF,T3.REF,T4.REF) = T1.REF
) TR

Answer 3

在你的帮助下，我发现如果我暂时忘记了物品信息（表T1），我几乎可以得到我想要的：

SELECT *
FROM T2
  FULL JOIN T3
    ON T3.ref = T2.ref
    AND T3.depot = T2.depot
  FULL JOIN T4
    ON (
      T4.ref = T2.ref
      AND T4.depot = T2.depot
    )
    OR (
      T4.ref = T3.ref
      AND T4.depot = T3.depot
    )

它给出：

|  T2REF | T2DEPOT |  T3REF | T3DEPOT |  T4REF | T4DEPOT |
|--------|---------|--------|---------|--------|---------|
|      1 |       A |      1 |       A | (null) |  (null) |
|      1 |       B | (null) |  (null) |      1 |       B |
| (null) |  (null) |      1 |       C |      1 |       C |
| (null) |  (null) | (null) |  (null) |      1 |       D |

然后我只需要获取 T1 信息：

SELECT *
FROM T2
  FULL JOIN T3
    ON T3.ref = T2.ref
    AND T3.depot = T2.depot
  FULL JOIN T4
    ON (
      T4.ref = T2.ref
      AND T4.depot = T2.depot
    )
    OR (
      T4.ref = T3.ref
      AND T4.depot = T3.depot
    )
  INNER/RIGHT JOIN T1
    ON T1.ref = COALESCE(T2.ref, T3.ref, T4.ref)

获取：

|  T2REF | T2DEPOT |  T3REF | T3DEPOT |  T4REF | T4DEPOT |  T1REF |
|--------|---------|--------|---------|--------|---------|--------|
|      1 |       A |      1 |       A | (null) |  (null) |      1 |
|      1 |       B | (null) |  (null) |      1 |       B |      1 |
| (null) |  (null) |      1 |       C |      1 |       C |      1 |
| (null) |  (null) | (null) |  (null) |      1 |       D |      1 |

为了好玩（以及更好地理解这个请求的目的），如果我将 T1 作为第一个表，我仍在尝试完成这项工作......

（太糟糕了 sqlfiddle 似乎没有管理多个具有相同名称的字段...）