MySQL - 列的条件连接

Question

我有以下疑问：

select ad_st_id_state, count(distinct id_visit) as Visits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address

group by ad_st_id_state
order by ad_st_id_state

我也有这个：

select ad_st_id_state, count(distinct id_visit) as DoneVisits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address
join sf_visit_file_time on id_visit = vft_vi_id_visit /* Another join has been added */

group by ad_st_id_state
order by ad_st_id_state

正如您所见，除了额外的join 语句之外，查询几乎相同。 这两个查询都返回了我需要的正确值，但我需要将它们加入一个表中，所以我这样做：

select fffuuu.ad_st_id_state, count(distinct id_visit) as Visitas, fffuuu.doneVisits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address

join (
select ad_st_id_state, count(distinct id_visit) as doneVisits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address
join sf_visit_file_time on id_visit = vft_vi_id_visit

group by ad_st_id_state
order by ad_st_id_state
) as fffuuu on sf_address.ad_st_id_state = fffuuu.ad_st_id_state

group by ad_st_id_state
order by ad_st_id_state

或者换句话说，我将第一个查询与第二个查询作为子查询连接起来。结果集很好且正确，但花费的时间太长，所以我在另一个系统中超时，这个查询正在运行。每个查询独立运行很快，但加入它们对我的需求来说太慢了......

我想知道是否有办法对此进行优化，我在想是否有一些加入条件语句之类的。我搜索了信息，但我没有任何运气。我在想这样的事情：

select ad_st_id_state, count(distinct id_visit) as Visits, if(@someVariable := true) as DoneVisits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address
if (@someVariable == true) then join sf_visit_file_time on id_visit = vft_vi_id_visit

group by ad_st_id_state
order by ad_st_id_state

或者类似的东西。我该如何优化它？

Answer 1

您能否在计数中使用 sf_visit_file_time 的外连接和 case 语句？显然，我在本地没有您的架构，但类似于：

select ad_st_id_state, 
   count(distinct id_visit) as Visits,
   count(distinct case when vft_vi_id_visit is not null then id_visit end) as DoneVisits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address
left join sf_visit_file_time on id_visit = vft_vi_id_visit
group by ad_st_id_state
order by ad_st_id_state

Answer 2

如果两个查询都快速且运行良好，并且您只需要一个表中的结果，则可以使用联合选择 http://dev.mysql.com/doc/refman/5.0/en/union.html

select ad_st_id_state, count(distinct id_visit) as Visits, '' as DoneVisits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address
group by ad_st_id_state
UNION
select ad_st_id_state, '' as Visits, count(distinct id_visit) as DoneVisits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address
join sf_visit_file_time on id_visit = vft_vi_id_visit /* Another join has been added */
group by ad_st_id_state
order by ad_st_id_state

Answer 3

您的两个查询都干净高效。因此，只需将它们连接在一起进行演示，将它们中的每一个都视为子查询。

SELECT a.ad_st_id_state, a.Visits, b.DoneVisits
  FROM (
        /* put your first query here */
       ) AS a
  LEFT JOIN (
        /* put your second query here */
       ) AS b ON a.ad_st_id_state = b.ad_st_id_state
 ORDER BY a.ad_st_id_state

这会产生更大的查询，但它应该在您拥有的两个查询的时间总和内运行。您可以将 ORDER BY 子句从您的子查询中删除。

所以它看起来像这样......一个名副其实的俱乐部三明治查询。

SELECT a.ad_st_id_state, a.Visits, b.DoneVisits
  FROM (
          select ad_st_id_state, count(distinct id_visit) as Visits
            from sf_visit
            join vr_users on vi_us_id_user = sus_us_id_user
            join sf_pdv on vi_pdv_id_pdv = id_pdv
            join sf_address on pdv_ad_id_address = id_address
           group by ad_st_id_state
       ) AS a
  LEFT JOIN (
           select ad_st_id_state, count(distinct id_visit) as DoneVisits
            from sf_visit
            join vr_users on vi_us_id_user = sus_us_id_user
            join sf_pdv on vi_pdv_id_pdv = id_pdv
            join sf_address on pdv_ad_id_address = id_address
            join sf_visit_file_time on id_visit = vft_vi_id_visit /* Another join */
           group by ad_st_id_state
       ) AS b ON a.ad_st_id_state = b.ad_st_id_state
 ORDER BY a.ad_st_id_state

Answer 4

在这种情况下，您的“条件连接”是在告诉您“如果我在另一个表中找到一条记录，则表示访问完成。”

为了创建该“条件”，您可以使用 LEFT OUTER JOIN 而不是 INNER JOIN

OUTER JOIN 与 INNER JOIN 相对，如果在一个表中而不是另一个表中找到记录，则不会中断。它仍然返回记录。有LEFT 和RIGHT 外连接。你应该自己研究它们。但基本上，要实现你想要的，你可以这样做：

select ad_st_id_state, count(distinct vft_vi_id_visit) as DoneVisits
from sf_visit
join vr_users on vi_us_id_user = sus_us_id_user
join sf_pdv on vi_pdv_id_pdv = id_pdv
join sf_address on pdv_ad_id_address = id_address
left outer join sf_visit_file_time on id_visit = vft_vi_id_visit /* Another join has been added */
group by ad_st_id_state
order by ad_st_id_state

所以现在，DoneVisits 将一直存在——当没有 sf_visit_file_time 记录时，它会为零。