0

我有 MySql 数据库，其中包含许多带有日期（时间戳）和更多属性的记录。例如，“testTable”看起来像


a varchar(255)
b int(11)
时间戳 bigint(20)

我需要在 1 月 1 日至 1 月 15 日的一段时间内每天找到 sum(b) 的前 10 名，这些日期可以由用户指定。

迭代查询会是什么样子？粗略的方法可能是单独的选择语句，其间有 UNION ALL。

select a, sum(b) from testTable where FROM_UNIXTIME( timestamp ) between '2012-01-01 05:10:00' and '2012-01-02 05:10:00' group by a order by sum(b) desc LIMIT 10
UNION ALL
select a, sum(b) from testTable where FROM_UNIXTIME( timestamp ) between '2012-01-02 05:10:00' and '2012-01-03 05:10:00' group by a  order by sum(b) desc LIMIT 10
UNION ALL
select a, sum(b) from testTable where FROM_UNIXTIME( timestamp ) between '2012-01-03 05:10:00' and '2012-01-04 05:10:00' group by a order by sum(b) desc LIMIT 10
..
..
..
UNION ALL
select a, sum(b) from testTable where FROM_UNIXTIME( timestamp ) between '2012-01-14 05:10:00' and '2012-01-15 05:10:00' group by a order by sum(b) desc LIMIT 10 ;</br>

但我希望它更通用，用户可以运行具有 2 个给定日期的脚本。

输出就像
一个 | FROM_UNIXTIME（时间戳）| 总和 (b)
------------+------------+------
测试 | 2012-01-01 03:24:41-04 | 500
测试 | 2012-01-01 03:19:40-04 | 420
测试| 2012-01-01 03:14:39-04 | 261
测试 | 2012-01-01 03:09:38-04 | 244
测试| 2012-01-01 03:04:37-04 | 231
测试| 2012-01-01 02:59:36-04 | 223
测试| 2012-01-01 02:54:35-04 | 211
测试1 | 2012-01-01 02:49:34-04 | 199
测试1 | 2012-01-01 03:24:41-04 | 195
测试1 | 2012-01-01 03:19:40-04 | 191
新 | 2012-01-02 06:11:06-04 | 1000
新| 2012-01-02 06:06:06-04 | 978
新| 2012-01-02 06:01:06-04 | 867
新 | 2012-01-02 05:56:05-04 | 786
新| 2012-01-02 05:51:05-04 | 698
新| 2012-01-02 05:46:05-04 | 第598
章 2012-01-02 06:11:06-04 | 第476
章 2012-01-02 05:41:04-04 | 第345
章 2012-01-02 06:06:06-04 | 250 新2
| 2012-01-02 06:01:06-04 | 125

4

1 回答 1

1

试试这个...更改介于范围内的日期以仅通过该范围一次。

更正了错字、省略的行和缺少逗号：

select day, a, tot
from 
   (
   select 
      *,
      @num := if(@day =  tt4.day, @num + 1, 1) as row_number,
      @day := tt4.day as dummy
   from
      (
      select
         ts as day, 
         tt1.a, 
         sum(tt1.b) as tot
      from 
         testTable tt1, 
         ( select distinct date(FROM_UNIXTIME(tt2.timestamp)) as ts
           from   testTable tt2
           where  date(FROM_UNIXTIME(tt2.timestamp)) between cast('2012/01/01' as date) and cast('2012/01/15' as date) ) as tt3
      where 
         date(FROM_UNIXTIME(tt1.timestamp)) = tt3.ts
      group by 
         date(FROM_UNIXTIME(tt1.timestamp)), 
         tt1.a
      order by 
         date(FROM_UNIXTIME(tt1.timestamp)),
         sum(tt1.b) desc,
         tt1.a
      ) as tt4
   ) as tt5
where 
   tt5.row_number <=10

已修改- 为 Vertica 更改了 SQL 的风格...语法可能已关闭（我没有要测试的 Vertica 安装），但要点就在那里。

select day, a, tot
from 
   (
   select 
      *,
      ROW_NUMBER() OVER (PARTITION BY tt4.day) as row_number
   from
      (
      select
         ts as day, 
         tt1.a, 
         sum(tt1.b) as tot
      from 
         testTable tt1, 
         ( select distinct date(TO_TIMESTAMP(tt2.timestamp)) as ts
           from   testTable tt2
           where  date(TO_TIMESTAMP(tt2.timestamp)) between cast('2012/01/01' as date) and cast('2012/01/15' as date) ) as tt3
      where 
         date(TO_TIMESTAMP(tt1.timestamp)) = tt3.ts
      group by 
         date(TO_TIMESTAMP(tt1.timestamp)), 
         tt1.a
      order by 
         date(TO_TIMESTAMP(tt1.timestamp)),
         sum(tt1.b) desc,
         tt1.a
      ) as tt4
   ) as tt5
where 
   tt5.row_number <=10
于 2012-07-16T19:15:54.310 回答