如何计算 SQL 服务器中变量的平均值

Question

我正在尝试计算行之间的日期范围，然后计算平均值。

declare @date0 date = (
  select top 1 my_date 
  from someTable 
  order by my_date desc)
declare @date1 date = (
  select my_date 
  from someTable 
  order by my_date desc
  offset 1 rows 
  fetch next 1 row only)
declare @date2 date = (
  select my_date 
  from someTable 
  order by my_date desc
  offset 2 rows 
  fetch next 1 row only)
declare @date3 date = (
  select my_date 
  from someTable 
  order by my_date desc
  offset 3 rows 
  fetch next 1 row only)

select 
[Range 1]   = dateDiff(day, @date1, @date0),
[Range 2]   = dateDiff(day, @date2, @date1),
[Range 3]   = dateDiff(day, @date3, @date2),
[Avg Range] = avg(
                nullIf(@date0, 0), 
                nullIf(@date1, 0), 
                nullIf(@date2, 0), 
                nullIf(@date3, 0)
              )

范围计算工作正常，但有点笨拙。

但是，我不确定如何处理平均值。看起来该函数应该运行针对 table 而不是数组，但是我在将变量插入临时 table 列时遇到了问题。

如何获得这些范围的平均值（不包括 range = 0）？

Answer 1

AVG 是一个聚合函数，设计用于 GROUP BY 或 windows。您可以简单地在查询中进行数学运算：

select 
[Range 1]   = dateDiff(day, @date1, @date0),
[Range 2]   = dateDiff(day, @date2, @date1),
[Range 3]   = dateDiff(day, @date3, @date2),
[Avg Range] = (
               nullIf(@date0, 0) +
               nullIf(@date1, 0) + 
               nullIf(@date2, 0) + 
               nullIf(@date3, 0)
              ) /
              (
               CASE WHEN @date0 IS NULL THEN 0 ELSE 1 END +
               CASE WHEN @date1 IS NULL THEN 0 ELSE 1 END +
               CASE WHEN @date2 IS NULL THEN 0 ELSE 1 END +
               CASE WHEN @date3 IS NULL THEN 0 ELSE 1 END
              )

Answer 2

使用 UNION ALL 为每种情况 return 一行：

select avg(t.[Range]) [Avg Range]
from (
  select dateDiff(day, @date1, @date0) [Range]
  union all
  select dateDiff(day, @date2, @date1)
  union all
  select dateDiff(day, @date3, @date2)
) t
where t.[Range] <> 0

Answer 3

没有理由使用四个不同的查询：

with dates as (
    select
        row_number() over (order by my_date desc) rn,
        datediff(days, lag(my_date) over (order by my_date desc), my_date) diff
    from T
)
select avg(nullif(diff, 0)) from dates where rn <= 3;

或

with dates as (
    select
        datediff(days, lag(my_date) over (order by my_date desc), my_date) diff
    from T
    order by my_date desc
    fetch next three rows only
)
select avg(nullif(diff, 0)) from dates;

使用 distinct 还可以让您轻松获得前三个日期，而不必乱用 nullif()。

with dates as (
    select distinct
        datediff(days, lag(my_date) over (order by my_date desc), my_date) diff
    from T
    order by my_date desc
    fetch next three rows only
)
select avg(diff) from dates;

如何计算 SQL 服务器中变量的平均值

How to calculate average of variables in SQL Server

sql

average

range