使用外连接时，where 子句中“OR”的 SQL 替代方法

Question

我有以下疑问：

select * from
from assignments dah, employees emp 
where 
    dah.person_id=emp.person_id(+)
and 
(dah.effective_end_date between emp.date_from(+) and emp.date_to(+)
and dah.effective_end_date between emp.valid_from(+) and emp.valid_to(+))
or   
(dah.effective_start_date between emp.date_from(+) and emp.date_to(+)
and dah.effective_start_date between emp.valid_from(+) and emp.valid_to(+))

我收到以下消息：“OR 或 IN 的操作数中不允许外连接运算符 (+)”。我知道使用带有内部连接的 2 个联合是一种解决方案，但我不能使用它，因为我实际上有很多代码（我提供的代码只是一个示例）。

编辑：我需要通过 oracle 语法完成此操作，因为我使用数据仓库，而我们的 ETL 不完全支持显式语法。也许有些东西我没有看到，这可以用不同的方式写吗？

编辑 nr.2：也许日期重叠逻辑可以在不使用 OR 和 oracle 语法的情况下以某种方式实现？

Answer 1

如果您将已弃用的外连接运算符 ((+)) 转换为显式外连接，它应该可以工作：

SELECT          *
FROM            assignments dah
LEFT OUTER JOIN employees emp ON
                dah.person_id = emp.person_id AND
                ((dah.effective_end_date BETWEEN emp.date_from AND 
                                                 emp.date_to AND 
                  dah.effective_end_date BETWEEN emp.valid_from AND 
                                                 emp.valid_to) OR
                (dah.effective_start_date BETWEEN emp.date_from AND 
                                                  emp.date_to AND
                 dah.effective_start_date BETWEEN emp.valid_from AND 
                                                  emp.valid_to)
                )

Answer 2

使用明确的left join 语法：

select *
from employees emp left join
     assignments dah 
     on dah.person_id = emp.person_id and
        ((dah.effective_end_date between emp.date_from and emp.date_to and
          dah.effective_end_date between emp.valid_from and emp.valid_to
         ) or
         (dah.effective_start_date between emp.date_from and emp.date_to and
          dah.effective_start_date between emp.valid_from and emp.valid_to
         )
        );

一个简单的规则是永远不要在from 子句中使用逗号。始终使用显式 join 语法。

注意：从技术上讲，您的外连接语法将具有 other 顺序的表：

from assignments dah left join
     employees emp 
     on . . .

我故意交换了它们。 left join 保留第一个表中的所有行，即使是那些没有匹配项的行。 + 语法更难遵循。 + 位于将获得 NULL 值的一侧。但是，对我来说，不匹配的行似乎不太可能出现在 assignments 表中。

如果你有正确的外键关系，那么所有的分配都应该有一个正确的人。但是，我可能不理解您的数据，您可能想要反转您的表格以实现您真正想要做的事情。

编辑：

至于重叠，我倾向于使用更简单的：

     on dah.person_id = emp.person_id and
        (dah.effective_end_date >= emp.date_from and
         dah.effective_start_date <= emp.date_to 
        )

如果您愿意，您甚至可以使用古老的 + 符号来编写此代码。另请注意：这些不做完全相同的事情。这将检测一个周期完全嵌入另一个周期的重叠。

Answer 3

您需要像这样使用显式 LEFT JOIN：

SELECT * FROM
    assignments dah LEFT JOIN 
    employees emp ON dah.person_id=emp.person_id
AND 
(dah.effective_end_date between emp.date_from and emp.date_to
and dah.effective_end_date between emp.valid_from and emp.valid_to)
or   
(dah.effective_start_date between emp.date_from and emp.date_to
and dah.effective_start_date between emp.valid_from and emp.valid_to)

Answer 4

如果不能使用 ANSI LEFT OUTER JOIN 语法，请只使用这个：

首先：您在查询中缺少括号 - 除了最初的 JOIN 之外，您可以将 x between min(+) AND max(+) 重写为 (min is NULL OR x >= min) AND (max is NULL OR x <= max)

SELECT *
FROM assignments dah, employees emp 
WHERE
    dah.person_id = emp.person_id(+)
AND 
(
      (emp.date_from IS NULL OR dah.effective_start_date >= emp.date_from)
  AND (emp.date_to IS NULL OR dah.effective_start_date <= emp.date_to)
  AND (emp.valid_from IS NULL OR dah.effective_start_date >= emp.valid_from)
  AND (emp.valid_to IS NULL OR dah.effective_start_date <= emp.valid_to)

  OR

      (emp.date_from IS NULL OR dah.effective_end_date >= emp.date_from)
  AND (emp.date_to IS NULL OR dah.effective_end_date <= emp.date_to)
  AND (emp.valid_from IS NULL OR dah.effective_end_date >= emp.valid_from)
  AND (emp.valid_to IS NULL OR dah.effective_end_date <= emp.valid_to)
)

我认为这会选择您想要的 - 与所有行的左连接，其中 start_date 或 end_date 在两个日期之间。

您希望所有行都来自具有 id 和右开始日期的 LEFT JOIN 或仅具有结束日期的行，没有任何 id 可以加入...您的查询基本上是这样的：WHERE ( id1=id2(+) AND ...) OR ( ... ) 因为 AND比 OR 的结合力更强。

如果您知道emp.date_from 和emp.date_to 都有效或为NULL

因此，如果从不存在只有 date_from 为 NULL，但 date_to 有效的情况，则可以大大缩短语句：

SELECT *
FROM assignments dah, employees emp 
WHERE
    dah.person_id = emp.person_id(+)
AND 
(
     emp.date_from IS NULL
  OR dah.effective_start_date BETWEEN emp.date_from AND emp.date_to
     AND dah.effective_start_date BETWEEN emp.valid_from AND emp.valid_to
  OR dah.effective_end_date BETWEEN emp.date_from AND emp.date_to
     AND dah.effective_end_date BETWEEN emp.valid_from AND emp.valid_to
)

Answer 5

由于您必须使用旧式外连接语法，所以这里有一种方法（简化，因为您没有向我们提供示例数据和/或表创建脚本）：

with assignments as (select 1 assignment_id, 1 person_id, to_date('01/08/2015', 'dd/mm/yyyy') start_date, to_date('03/08/2015', 'dd/mm/yyyy') end_date from dual union all
                     select 2 assignment_id, 1 person_id, to_date('02/08/2015', 'dd/mm/yyyy') start_date, to_date('04/08/2015', 'dd/mm/yyyy') end_date from dual union all
                     select 3 assignment_id, 1 person_id, to_date('06/08/2015', 'dd/mm/yyyy') start_date, to_date('10/08/2015', 'dd/mm/yyyy') end_date from dual union all
                     select 4 assignment_id, 2 person_id, to_date('02/08/2015', 'dd/mm/yyyy') start_date, to_date('03/08/2015', 'dd/mm/yyyy') end_date from dual),
       employees as (select 1 person_id, to_date('01/08/2015', 'dd/mm/yyyy') start_date, to_date('03/08/2015', 'dd/mm/yyyy') end_date from dual union all
                     select 3 person_id, to_date('01/08/2015', 'dd/mm/yyyy') start_date, to_date('03/08/2015', 'dd/mm/yyyy') end_date from dual)
select *
from   assignments dah,
       employees emp
where  dah.person_id = emp.person_id (+)
and    dah.start_date <= emp.end_date (+)
and    dah.end_date >= emp.start_date (+);

ASSIGNMENT_ID  PERSON_ID START_DATE END_DATE   PERSON_ID_1 START_DATE_1 END_DATE_1
------------- ---------- ---------- ---------- ----------- ------------ ----------
            2          1 02/08/2015 04/08/2015           1 01/08/2015   03/08/2015
            1          1 01/08/2015 03/08/2015           1 01/08/2015   03/08/2015
            3          1 06/08/2015 10/08/2015                                    
            4          2 02/08/2015 03/08/2015          

你确定你的外连接是正确的吗？您确定您实际上不是在追求以下内容吗？：

with assignments as (select 1 assignment_id, 1 person_id, to_date('01/08/2015', 'dd/mm/yyyy') start_date, to_date('03/08/2015', 'dd/mm/yyyy') end_date from dual union all
                     select 2 assignment_id, 1 person_id, to_date('02/08/2015', 'dd/mm/yyyy') start_date, to_date('04/08/2015', 'dd/mm/yyyy') end_date from dual union all
                     select 3 assignment_id, 1 person_id, to_date('06/08/2015', 'dd/mm/yyyy') start_date, to_date('10/08/2015', 'dd/mm/yyyy') end_date from dual union all
                     select 4 assignment_id, 2 person_id, to_date('02/08/2015', 'dd/mm/yyyy') start_date, to_date('03/08/2015', 'dd/mm/yyyy') end_date from dual),
       employees as (select 1 person_id, to_date('01/08/2015', 'dd/mm/yyyy') start_date, to_date('03/08/2015', 'dd/mm/yyyy') end_date from dual union all
                     select 3 person_id, to_date('01/08/2015', 'dd/mm/yyyy') start_date, to_date('03/08/2015', 'dd/mm/yyyy') end_date from dual)
select *
from   assignments dah,
       employees emp
where  dah.person_id (+) = emp.person_id
and    dah.start_date (+) <= emp.end_date
and    dah.end_date (+) >= emp.start_date;

ASSIGNMENT_ID  PERSON_ID START_DATE END_DATE   PERSON_ID_1 START_DATE_1 END_DATE_1
------------- ---------- ---------- ---------- ----------- ------------ ----------
            1          1 01/08/2015 03/08/2015           1 01/08/2015   03/08/2015
            2          1 02/08/2015 04/08/2015           1 01/08/2015   03/08/2015
                                                         3 01/08/2015   03/08/2015