按公司规模和月份计算答案

【问题标题】：count by company size and month按公司规模和月份计算
【发布时间】：2018-04-17 12:40:08
【问题描述】：

我有一个带有列的表员工：

company_id
employee_id
opted_out_on（日期）
opt_out_window_starts_on（日期）

两者都与员工相关，因此同一公司的员工之间存在差异。我想要选择退出 opted_out_on 的月份等于 opt_out_window_starts_on 的月份的员工数量，按公司规模分组。我写了下面的查询，但数字是错误的。

源数据如下：

company_id | employee_id | opt_out_on | opt_out_window_starts_on
-----------+-------------+------------+-------------------------
     23101 |   920190500 | 20/01/2017 | 09/01/2017
     12264 |   920190508 | 02/03/2017 | 04/02/2017
     12039 |   920190513 | 07/11/2017 | 31/10/2017
     12039 |   920190516 | 13/02/2017 | 11/02/2017
     23109 |   920190520 | 06/01/2017 | 06/01/2017
     21875 |   920190529 | 22/05/2017 | 15/05/2017
     21875 |   920190532 | 19/05/2017 | 15/05/2017
     21875 |   920190533 | 22/05/2017 | 15/05/2017
     21875 |   920190534 | 15/05/2017 | 15/05/2017
     21875 |   920190535 | 22/05/2017 | 15/05/2017
     21875 |   920190536 | 19/05/2017 | 15/05/2017
     18112 |   920190545 | 11/01/2017 | 05/01/2017
      9764 |   920190546 | 02/05/2017 | 02/05/2017
     18365 |   920190665 | 30/03/2017 | 28/03/2017
     18365 |   920190668 | 28/03/2017 | 28/03/2017
     18365 |   920190678 | 28/03/2017 | 28/03/2017
     20928 |   920190699 | 14/06/2017 | 13/06/2017
     20928 |   920190705 | 20/06/2017 | 13/06/2017
     16860 |   920190717 | 17/01/2017 | 16/01/2017
     16860 |   920190718 | 16/01/2017 | 16/01/2017

此数据可在SQL Fiddle 上获得。

我的代码是：

with size as 
(
    select distinct company_id
    , date_trunc('month', opt_out_window_starts_on) :: date months
    , case 
        when count(id) over (partition by company_id ) = 1 then 
            count(id)
            filter 
            (
                where opt_state = 'opted_out' 
                and 
                (
                    date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
                )
            )
        else 0 
    end as "one"
    , case 
        when count(id) over (partition by company_id ) = 2 then 
            count(id)
            filter 
            (
                where opt_state = 'opted_out' 
                and 
                (
                    date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
                )
            )
        else 0 
    end as "two"
    , case 
        when 
            count(id) over (partition by company_id ) >= 3 
            and count(id) over (partition by company_id ) <= 4
        then 
            count(id) filter 
            (
                where opt_state = 'opted_out' 
                and 
                (
                    date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
                )
            )
        else 0 
    end as "3-4"
    , case 
        when 
            count(id) over (partition by company_id ) >= 5 
            and count(id) over (partition by company_id ) <= 9
        then 
            count(id)
            filter 
            (
                where opt_state = 'opted_out' 
                and 
                (
                    date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
                )
            )
        else 0 
    end as "5-9"
    , case 
        when 
            count(id) over (partition by company_id ) >= 10 
            and count(id) over (partition by company_id ) <= 29
        then 
            count(id)
            filter 
            (
                where opt_state = 'opted_out' 
                and 
                (
                    date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
                )
            )
        else 0 
    end as "10-29"
    , case 
        when 
            count(id) over (partition by company_id ) >= 30 
            and count(id) over (partition by company_id ) <= 49
        then 
            count(id)
            filter (
                where opt_state = 'opted_out' 
                and 
                (
                    date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
                )
            )
        else 0 
    end as "30-49"
    , case 
        when 
            count(id) over (partition by company_id) >= 50 
            and count(id) over (partition by company_id) <= 249
        then 
            count(id)
            filter 
            (
                where opt_state = 'opted_out' 
                and 
                (
                    date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
                )
            )
        else 0 
    end as "50-249"
    , case 
        when 
            count(id) over (partition by company_id ) >= 250
        then 
            count(id)
            filter 
            (
                where opt_state = 'opted_out' 
                and 
                (
                    date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
                )
            )
        else 0 
    end as "250+"
    from employees
    group by company_id
    , id
) 
select to_char(date_trunc('month', months),'YYYY MON')
, sum(one)
, sum(two)
, sum("3-4")
, sum("5-9")
, sum("10-29")
, sum("30-49")
, sum("50-249")
    , sum("250+")
from size
where to_char(date_trunc('month', months),'YYYY MON') is not null
group by date_trunc('month', months)
order by date_trunc('month', months) asc
;

错误结果示例：

  to_char  | sum | sum | sum | sum | sum | sum | sum | sum
  ---------+-----+-----+-----+-----+-----+-----+-----+----
  2017 JAN |  35 |  37 |  85 | 119 | 131 |  39 |  19 |   3
  2017 FEB |  49 |  53 | 112 | 165 | 170 |  41 |  23 |   2
  2017 MAR |  29 |  33 |  62 |  85 | 104 |  24 |  23 |   1
  2017 APR |  37 |  50 | 117 | 150 | 159 |  33 |  28 |   3
  2017 MAY |  71 |  84 | 150 | 182 | 208 |  51 |  44 |   4
  2017 JUN |  35 |  51 |  80 |  90 | 128 |  26 |  43 |   7
  2017 JUL |  69 |  84 | 146 | 177 | 173 |  51 |  45 |   5
  2017 AUG |  93 | 137 | 168 | 194 | 234 |  63 |  51 |   5
  2017 SEP |  75 |  60 |  97 | 127 | 136 |  44 |  44 |   6
  2017 OCT |  91 | 113 | 132 | 157 | 126 |  38 |  49 |   6
  2017 NOV | 125 | 158 | 168 | 198 | 162 |  36 |  58 |  11
  2017 DEC |  49 |  81 |  83 | 103 |  91 |  27 |  40 |  10
  2018 JAN | 134 | 132 | 119 | 158 | 132 |  37 |  41 |  12
  2018 FEB | 127 | 111 | 155 | 142 | 121 |  37 |  45 |  15
  2018 MAR | 112 |  92 | 119 | 115 | 116 |  35 |  37 |  15
  2018 APR |  37 |  51 |  64 |  57 |  56 |  14 |  26 |  10
  2018 MAY |   0 |   0 |   0 |   0 |   0 |   0 |   0 |   0

如果您以 2017 年 1 月为例，选择退出的员工总数为 1404，而我的查询显示为 468 (35 + 37 + 85 + 119 + 131 + 39 + 19 + 3)。

我核对了号码：

select count(id)
, date_trunc('month', opt_out_window_starts_on)
from employees
where opt_state = 'opted_out'
and date_trunc('month', opt_out_window_starts_on) = date_trunc('month', opt_out_on)
group by date_trunc('month', opt_out_window_starts_on)

【问题讨论】：

向我们展示样本数据和样本结果。然后指出错误的地方并给出正确的结果。
嗨，抱歉，如何添加数据样本？谢谢
恕我直言，好像放在 SQL 中的东西应该被移植到“视图”层，f.o.报告引擎
您可以在表格中选择样本数据（或虚拟数据），然后将其粘贴到您的 SO 帖子/问题中。然后显示哪些数据/结果不正确以及您对结果的期望。查看其他一些帖子/问题以获得想法。
或者，您可以设置一个 SQL Fiddle 并将示例数据放在那里。

标签： sql postgresql

【解决方案1】：

如果我正确地理解了你想要的问题：

表示给定公司规模范围（即员工总数）的列
给出年份和月份的行
每个单元格的值代表那些公司中在提供退出选项的同一个月内选择退出某项服务的员工人数。

这段代码解决了这个问题的第一部分；获取公司列表、规模和离职人数：

select company_id
, date_trunc('month', opt_out_window_starts_on) optOutMonth
, count(employee_id) as companySize
, count(case 
    when opt_state = 'opted_out'
    and date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
    then employee_id
    else null
end)
 as leaversInOptOutWindow
from employee
group by company_id, date_trunc('month', opt_out_window_starts_on)

SQL Fiddle

然后，您可以将其与另一个表连接起来，在该表中定义了您的范围以将公司分组为大小，并对结果执行透视。

遗憾的是 Postgres 不支持数据透视表，但似乎可以通过 crosstab 函数实现相同的功能：PostgreSQL Crosstab Query

我没有让这个在 SQL Fiddle 中工作；不确定这是否归结为我的代码中的一个问题，或者缺乏对这个功能的支持（因为我对 Postgres 不够熟悉）......但是我相信这应该是你所追求的领域：

SELECT * FROM crosstab(
'
  select e.optOutMonth
    , s.name
    , sum(e.leaversInOptOutWindow) LeaverCount
    from
    (
        select company_id
        , date_trunc(''month'', opt_out_window_starts_on) optOutMonth
        , count(employee_id) as companySize
        , count(case 
            when opt_state = ''opted_out''
            and date_trunc(''month'', opt_out_on) = date_trunc(''month'', opt_out_window_starts_on)
            then employee_id
            else null
        end) as leaversInOptOutWindow
        from employee 
        group by company_id
        , date_trunc(''month'', opt_out_window_starts_on)
    ) e
    left outer join companySizeRange s
    on s.minSize <= e.companySize
    and (s.maxSize is null or s.maxSize >= e.companySize)
    group by e.optOutMonth, s.name
    order by e.optOutMonth, s.name
'
  , 'select name from companySizeRange order by minSize'
 ) as x(
   optOutMonth date, "1" int, "2" int, "3-4" int, "5-9" int, "10-29" int, "30-49" int, "50-249" int, "250+" int
 );

SQL Fiddle

更新

这是一个不需要crosstab/pivot的版本：

select e.optOutMonth
, sum(case when s.companySize = 1 then e.leaversInOptOutWindow else 0 end) "1"
, sum(case when s.companySize = 2 then e.leaversInOptOutWindow else 0 end) "2"
, sum(case when s.companySize between 3 and 4 then e.leaversInOptOutWindow else 0 end) "3-4"
, sum(case when s.companySize between 5 and 9 then e.leaversInOptOutWindow else 0 end) "5-9"
, sum(case when s.companySize between 10 and 29 then e.leaversInOptOutWindow else 0 end) "10-29"
, sum(case when s.companySize between 30 and 49 then e.leaversInOptOutWindow else 0 end) "30-49"
, sum(case when s.companySize between 50 and 249 then e.leaversInOptOutWindow else 0 end) "50-249"
, sum(case when s.companySize >= 250 then e.leaversInOptOutWindow else 0 end) "250+"
from
(
    select company_id
    , count(employee_id) as companySize
    from employee 
    group by company_id
) s
inner join
(
    select company_id
    , date_trunc('month', opt_out_window_starts_on) optOutMonth
    , count(employee_id) as leaversInOptOutWindow
    from employee 
    where opt_state = 'opted_out'
    and date_trunc('month', opt_out_on) = date_trunc('month', opt_out_window_starts_on)
    group by company_id
    , date_trunc('month', opt_out_window_starts_on)
) e
on e.company_id = s.company_id
group by e.optOutMonth
order by e.optOutMonth

SQL Fiddle

【讨论】：

是的，这正是我想要的，等你更新非常感谢
已更新，但尚未在 SQL Fiddle 中运行......所以也许我做错了什么；或者可能不支持此功能...遗憾的是我不完全确定...
非常感谢您的尝试，它出错 [42P01] 错误：关系“companysizerange”不存在
不用担心。看一下 SQL Fiddle 的左窗格；这是我创建的用于定义您的范围的新表（不是严格需要；但我觉得这可以稍微简化代码和维护）。
这是完美的工作，结果是 100% 正确，非常感谢您的帮助 @JohnLBevan