UNION ALL 正在返回多个结果 [重复]

Question

此查询返回多个结果。在底部，您可以看到每个 ID 号有两个。如何区分仍然获得 25 条唯一值记录的记录？

        SELECT *
            FROM(
                SELECT
                    DISTINCT cards.id, cards.name, cards.created_at, cards.updated_at
                    FROM cards
                        INNER JOIN card_process
                            ON card_process.card_id = cards.id
                            INNER JOIN processes
                                ON processes.id = card_process.process_id
                                INNER JOIN category_process
                                    ON category_process.process_id = processes.id
                                    INNER JOIN categories
                                        ON categories.id = category_process.category_id
                                        INNER JOIN series
                                            ON series.id = categories.serie_id
                                            INNER JOIN serie_user
                                                ON serie_user.serie_id = series.id
                        AND `cards`.`type` NOT IN ('', 'libraries')
                        AND NOT `cards`.`deleted`
                        AND NOT `categories`.`deleted`
                        AND NOT `series`.`deleted`
                        AND `cards`.`type` IN ('forms')
                        AND `series`.`id` IN (124,235,126,126,201,236,207,207,207,207,247,234,131,131,221,225,225,222)
            UNION ALL
                SELECT
                    DISTINCT cards.id, cards.name, cards.created_at, cards.updated_at
                    FROM cards
                        WHERE `cards`.`account_user_id`='9'
                            AND NOT `cards`.`deleted`
                            AND `cards`.`type` IN ('forms')
            ) AS qry
                ORDER BY `updated_at` ASC
                LIMIT 0, 25

Answer 1

是的，UNION ALL 就是这样工作的，它只是为您提供两个查询的所有行。

如果要删除重复项，请使用 UNION ，不带 ALL 部分。

Answer 2

对架构或数据形状一无所知，您可以通过简化查询来获得所需的信息。假设一个大致符合标准的 SQL 方言

• 您不需要封闭的select。 全选（联合选择表达式）采用order by，并且该顺序适用于整个全选。

• 您不需要在union 中的单个select 语句中使用distinct 关键字：union 本身会消除重复行。

并且，当您编写一个长而复杂的查询时，请花一些时间对其进行格式化，以便下一个必须理解它的人（可能是几年后的您）可以轻松地理解它。

这会让你得到等价物：

SELECT cards.id   ,
       cards.name ,
       cards.created_at ,
       cards.updated_at
FROM       cards
INNER JOIN card_process     ON card_process.card_id = cards.id
INNER JOIN processes        ON processes.id = card_process.process_id
INNER JOIN category_process ON category_process.process_id = processes.id
INNER JOIN categories       ON categories.id =  category_process.category_id
INNER JOIN series           ON series.id = categories.serie_id
INNER JOIN serie_user       ON serie_user.serie_id = series.id
                           AND `cards`.`type` NOT IN ('', 'libraries')
                           AND NOT `cards`.`deleted`
                           AND NOT `categories`.`deleted`
                           AND NOT `series`.`deleted`
                           AND `cards`.`type` IN ('forms')
                           AND `series`.`id` IN (124,235,126,126,201,236,207,207,207,207,247,234,131,131,221,225,225,222)
UNION
SELECT cards.id   ,
       cards.name ,
       cards.created_at ,
       cards.updated_at
FROM cards
WHERE `cards`.`account_user_id`='9'
  AND NOT `cards`.`deleted`
  AND `cards`.`type` IN ('forms')
ORDER BY `updated_at` ASC
LIMIT 0, 25

如果 MySql 不允许限制 全选，那么您可能需要封闭的 select 语句。

另外，您可能会注意到您的联接条件几乎肯定不正确。在第一次选择中，serie_user 表的联接条件有一堆测试：

AND `cards`.`type` NOT IN ('', 'libraries')
AND NOT `cards`.`deleted`
AND NOT `categories`.`deleted`
AND NOT `series`.`deleted`
AND `cards`.`type` IN ('forms')
AND `series`.`id` IN (124,235,126,126,201,236,207,207,207,207,247,234,131,131,221,225,225,222)

它们仅适用于加入来自serie_user 的候选行。它们不适用于整个结果集。它们应该被重构为 where 子句和相关表的连接条件，因此：

SELECT cards.id   ,
       cards.name ,
       cards.created_at ,
       cards.updated_at
FROM       cards
INNER JOIN card_process     ON card_process.card_id        = cards.id
INNER JOIN processes        ON processes.id                = card_process.process_id
INNER JOIN category_process ON category_process.process_id = processes.id
INNER JOIN categories       ON categories.id               =  category_process.category_id
                           AND NOT `categories`.`deleted`
INNER JOIN series           ON series.id                   = categories.serie_id
                           AND NOT `series`.`deleted`
                           AND     `series`.`id`           IN (124,235,126,126,201,236,207,207,207,207,247,234,131,131,221,225,225,222)
INNER JOIN serie_user       ON serie_user.serie_id         = series.id
WHERE `cards`.`type` NOT IN ('', 'libraries')
  AND NOT `cards`.`deleted`
  AND     `cards`.`type` IN ('forms')
UNION
SELECT cards.id   ,
       cards.name ,
       cards.created_at ,
       cards.updated_at
FROM cards
WHERE     `cards`.`account_user_id` = '9'
  AND NOT `cards`.`deleted`
  AND     `cards`.`type` IN ('forms')
ORDER BY `updated_at` ASC
LIMIT 0, 25

我要注意的最后一件事是，您正在加入大量未在结果集中使用的表。您所做的只是从cards 表中获取行的子集。这向我表明，如果您简单地摆脱 union 和所有无关的连接并简单地提出正确的问题，您可能会过得更好。

这完全消除了重复的可能性。

稍微重构一下就可以解决这个问题（不能保证我 100% 正确，但你应该能够理解它的要点）：

select c.id         ,
       c.name       ,
       c.created_at ,
       c.updated_at
from cards c
where   not c.deleted
  and       c.type            = 'forms'
  and (     c.account_user_id = '9'
        OR exists ( select *
                    from card_process     cp
                    join processes        p   on     p.id           = cp.process_id
                    join category_process cpx on     cpx.process_id = p.id
                    join categories       c   on     c.id           = cpx.category_id
                                             and not c.deleted
                    join series           s   on     s.id           = categories.serid_id
                                             and not s.deleted
                                             and     series.id IN ( 124 , 235 , 126 , ... )
                    join serie_user       su  on     su.serid_id     = s.id
                    where cp.card_id = c.card_id
                  )
      )
ORDER BY updated_at
LIMIT 0, 25

一个简单的查询通常隐藏在一个复杂的查询中。