从累积和表中获取数据

Question

这是我的主表，它根据日期存储关键字展示次数。

+---------+-----------+------------+-------------+
| ID      | KeywordId | ReportDate | Impressions |
+---------+-----------+------------+-------------+
| 3324620 | 969274443 | 2018-12-21 | 1           |
+---------+-----------+------------+-------------+
| 3334253 | 969274443 | 2018-12-22 | 0           |
+---------+-----------+------------+-------------+
| 3354119 | 969274443 | 2018-12-23 | 20          |
+---------+-----------+------------+-------------+
| 3373097 | 969274443 | 2018-12-24 | 11          |
+---------+-----------+------------+-------------+
| 3392168 | 969274443 | 2018-12-25 | 1           |
+---------+-----------+------------+-------------+

这是我的累积总和表，它每天都会累加展示次数

+----+-----------+------------+-------------+
| ID | keywordId | ReportDate | Impressions |
+----+-----------+------------+-------------+
| 1  | 969274443 | 2018-12-21 | 1           |
+----+-----------+------------+-------------+
| 2  | 969274443 | 2018-12-22 | 1           |
+----+-----------+------------+-------------+
| 3  | 969274443 | 2018-12-23 | 21          |
+----+-----------+------------+-------------+
| 4  | 969274443 | 2018-12-24 | 32          |
+----+-----------+------------+-------------+
| 5  | 969274443 | 2018-12-25 | 33          |
+----+-----------+------------+-------------+

要获得日期 2018-12-21 和 2018-12-25 之间的展示次数，只需减去

33 - 1 得到结果 22.

我使用下面的查询来得到这个结果。

select  keyword.keywordId,(second.impressions - first.impressions) as imp from keyword
inner join diffreportkeyword as first on first.keywordId = keyword.keywordId
                                     and first.reportDate = '2018-12-21'
inner join diffreportkeyword as second on second.keywordId = keyword.keywordId
                                      and second.reportDate = '2018-12-25'

但有时，特定日期没有数据，查询返回 0 行。但这意味着对于特定日期，展示次数为 0。

如果我用左连接改变这个，

select  keyword.keywordId,(second.impressions - first.impressions) as imp from keyword
left join diffreportkeyword as first on first.keywordId = keyword.keywordId
                                     and first.reportDate = '2018-12-18'
left join diffreportkeyword as second on second.keywordId = keyword.keywordId
                                      and second.reportDate = '2018-12-25'

这可行，但 diffreportkeyword 表有 500 万行，关键字表有 300k 个不同的“keywordId”，左连接会使查询非常慢。

我的问题是，

我怎样才能让它更快？

谢谢

Answer 1

这个

  on second.keywordId = keyword.keywordId
 and second.reportDate = '2018-12-25'

需要

INDEX(keywordId, reportDate)  -- in either order

旁注：说起来可能更简单

    '2018-12-21' - INTERVAL 3 DAY

而不是

    '2018-12-18'