mysql多个左连接并按主表分组

Question

我有以下情况。一个地区有多个地区，一个地区有多个地址，一个地址在一个月内被访问多次。现在我想生成一份关于某个地区的月度报告。 （一个区域被访问了多少次）。我已经编写了查询，但结果集产生的区域较少，因为有些地址没有被访问。我有以下结构

表格

areas: id|name (180 rows) //名称是唯一的

区域：id|name|area_id（1k 行）

地址：id|name|territory_id（80k 行）

visiting_addresses：id|address_id|date|status（1M+ 行）//状态 => 1 = 已访问，2 = 待处理

我的查询如下。

select ar.id as area_id, ar.name as area,
sum(case when va.status = 1 then 1 else 0 end) as visited,
sum(case when va.status = 2 then 1 else 0 end) as pending,
count(va.id) as total

from      areas ar
left join territories t on t.area_id=ar.id
left join addresses a on a.territory_id=t.id
left join visiting_addresses va on va.address_id=a.id
where     month(va.date) = '01'
and       year(va.date)='2020'
group by  ar.id

区域表包含 180 个区域，但结果集仅显示 144 个区域。我的错误在哪里，对此有何解释？那些区域丢失了，因为他们没有来访。

Answer 1

您的 WHERE 子句正在将带有 visiting_addresses 的 LEFT JOIN 转换为 INNER JOIN。由于它是 LEFT-JOIN 链中最右边的表，所有连接都将转换为 INNER JOINS。为了防止这种情况，您应该将相应的条件从 WHERE 子句移到 ON 子句：

select ar.id as area_id, ar.name as area,
sum(case when va.status = 1 then 1 else 0 end) as visited,
sum(case when va.status = 2 then 1 else 0 end) as pending,
count(va.id) as total

from      areas ar
left join territories t on t.area_id=ar.id
left join addresses a on a.territory_id=t.id
left join visiting_addresses va
  on  va.address_id=a.id
  and month(va.date) = '01'
  and year(va.date)='2020'

group by  ar.id

但由于您有很多行，我宁愿运行两个查询。首先使用内部连接仅获取上个月具有地址的区域。您应该更改va.date 上的条件以使用索引：

select ar.id as area_id, ar.name as area,
sum(case when va.status = 1 then 1 else 0 end) as visited,
sum(case when va.status = 2 then 1 else 0 end) as pending,
count(va.id) as total

from areas ar
join territories t on t.area_id=ar.id
join addresses a on a.territory_id=t.id
join visiting_addresses va on  va.address_id=a.id
where va.date >= '2020-01-01'
  and va.date <  '2020-02-01'

group by  ar.id

确保您在visiting_addresses(date) 上有索引，或者在visiting_addresses(date, address_id, status) 上有更好的索引。

然后用一个简单的方法得到所有区域

select ar.id as area_id, ar.name as area 
from areas ar

并在将visited、pending 和total 设置为零（在应用程序代码中）时将缺失区域添加到第一个结果中。

INNER JOIN 应该快得多，因为现在引擎可以使用 WHERE 条件的索引从 visiting_addresses 开始仅读取必要的行。

您还可以使用更复杂但单一的查询。想法是使用带有预聚合子查询的 LEFT JOIN：

select ar.id as area_id, ar.name as area,
    coalesce(visited, 0) as visited,
    coalesce(pending, 0) as pending,
    coalesce(total, 0) as total
from areas ar
left join (
    select t.area_id
    sum(case when va.status = 1 then 1 else 0 end) as visited,
    sum(case when va.status = 2 then 1 else 0 end) as pending,
    count(va.id) as total
    from territories t
    join addresses a on a.territory_id=t.id
    join visiting_addresses va on  va.address_id=a.id
    where va.date >= '2020-01-01'
      and va.date <  '2020-02-01'
    group by t.area_id
) x on x.area_id = ar.id

Answer 2

尝试将WHERE 子句中的逻辑移动到相应连接的ON 子句中：

SELECT
    ar.id AS area_id,
    ar.name AS area,
    COUNT(CASE WHEN va.status = 1 THEN 1 END) AS visited,
    COUNT(CASE WHEN va.status = 2 THEN 1 END) AS pending,
    COUNT(va.id) AS total
FROM areas ar
LEFT JOIN territories t ON t.area_id = ar.id
LEFT JOIN addresses a ON a.territory_id = t.id
LEFT JOIN visiting_addresses va ON va.address_id = a.id AND
    va.date >= '2020-01-01' AND va.date <  '2020-02-01'
GROUP BY
    ar.id;

请注意，选择 name 字段时仅按 id 进行聚合在 MySQL 中是有效的，假设 id 是 areas 表中的唯一字段。

您也可以尝试将以下索引添加到visiting_addresses 表中：

CREATE INDEX date_idx ON visiting_addresses (address_id, date, status);

这可能有助于加快加入此表的速度。