postgresql中的移动平均线

Question

我的Postgresql 9.1数据库中有以下表格:

select * from ro;
date       |  shop_id | amount 
-----------+----------+--------
2013-02-07 |     1001 |      3
2013-01-31 |     1001 |      2
2013-01-24 |     1001 |      1
2013-01-17 |     1001 |      5
2013-02-10 |     1001 |     10
2013-02-03 |     1001 |      4
2012-12-27 |     1001 |      6
2012-12-20 |     1001 |      8
2012-12-13 |     1001 |      4
2012-12-06 |     1001 |      3
2012-10-29 |     1001 |      3

我试图得到一个移动平均线,比较过去3个星期四的数据而不包括当前的星期四.这是我的查询:

select date, shop_id, amount, extract(dow from date),
avg(amount) OVER (PARTITION BY extract(dow from date) ORDER BY date DESC
                      ROWS BETWEEN 0 PRECEDING AND 2 FOLLOWING)                          
from ro
where extract(dow from date) = 4

这是给出的结果

date       |  shop_id | amount | date_part |        avg         
-----------+----------+--------+-----------+--------------------
2013-02-07 |     1001 |      3 |         4 | 2.0000000000000000
2013-01-31 |     1001 |      2 |         4 | 2.6666666666666667
2013-01-24 |     1001 |      1 |         4 | 4.0000000000000000
2013-01-17 |     1001 |      5 |         4 | 6.3333333333333333
2012-12-27 |     1001 |      6 |         4 | 6.0000000000000000
2012-12-20 |     1001 |      8 |         4 | 5.0000000000000000
2012-12-13 |     1001 |      4 |         4 | 3.5000000000000000
2012-12-06 |     1001 |      3 |         4 | 3.0000000000000000

我预计

date       |  shop_id | amount | date_part |        avg         
-----------+----------+--------+-----------+--------------------
2013-02-07 |     1001 |      3 |         4 | 2.6666666666666667
2013-01-31 |     1001 |      2 |         4 | 4.0000000000000000
2013-01-24 |     1001 |      1 |         4 | 6.3333333333333333
2013-01-17 |     1001 |      5 |         4 | 6.0000000000000000
2012-12-27 |     1001 |      6 |         4 | 5.0000000000000000
2012-12-20 |     1001 |      8 |         4 |
2012-12-13 |     1001 |      4 |         4 |
2012-12-06 |     1001 |      3 |         4 |

Answer 1

SQL小提琴

select
    "date",
    shop_id,
    amount,
    extract(dow from date),
    case when
        row_number() over (order by date) > 3
        then
            avg(amount) OVER (
                ORDER BY date DESC
                ROWS BETWEEN 1 following AND 3 FOLLOWING
            )
        else null end
from (
    select *
    from ro
    where extract(dow from date) = 4
) s

OP的查询有什么问题是框架规范:

ROWS BETWEEN 0 PRECEDING AND 2 FOLLOWING

除此之外,我的查询通过在应用昂贵的窗口函数之前过滤星期四来避免不需要的计算.

如果有必要按shop_id进行分区,那么显然partition by shop_id要将两个函数添加到,avg并且row_number.

Answer 2

我认为更好的答案可能是:

SELECT date, shop_id, amount, 
    extract(dow from date) AS dow,
    CASE WHEN count(amount) OVER w = 3 
        THEN avg(amount) OVER w END AS average_amt             
FROM ro
WHERE extract(dow from date) = 4 
WINDOW w AS (ORDER BY date DESC ROWS BETWEEN 1 FOLLOWING AND 3 FOLLOWING)

我认为使用相同的窗口来检查窗口中的行数并取平均值是更清晰的.(这也可以保存两个窗口聚合,如原始答案中所示.)

关于早期答案中的声明"我的查询通过在应用昂贵的窗口函数之前过滤星期四来避免不需要的计算",这也适用于OP和我的查询建议的查询,如附加EXPLAIN到任一节目.