如何在带有 Postgres 的动态框架中使用窗口函数中的列值？

Question

我不确定 Postgres 是否可以做到这一点，但我希望每一行都计算其窗口函数“框架”。

那么如何在 Postgres 的窗口函数中使用当前行中的列作为 ROWS 的参数？

例如..

create table items (
  id serial primary key, name varchar, price integer, precede_by integer
);

insert into items (name, price, precede_by) 
  values 
  ('A', 1, 2), ('A', 1, 3), 
  ('B', 1, 1),('B',1, 2), ('B', 1, 1), ('B', 2, 2), 
  ('C', 1, 2), ('C', 1,2), ('C', 1, 2), ('C', 1, 5);

我想使用在precede_by 列中设置的整数值。例如，像这样..

select name, sum(price) over 
  (partition by name rows between precede_by preceding and current row) 
from items;

但这不起作用。我收到ERROR: argument of ROWS must not contain variables

Answer 1

不，不可能随心所欲地编写窗口函数。我相信CROSS JOIN LATERAL 可以解决问题

with oitems as
(
  select *, row_number() over (partition by name order by id) rn from items
)
select * 
from oitems i1
cross join lateral 
(
  select sum(price) 
  from  oitems i2
  where i1.name = i2.name and i2.rn >= i1.rn - i1.precede_by and i2.rn <= i1.rn
) t

另一个（可能更优雅）的解决方案是使用 GROUP BY 的自联接

with oitems as
(
  select *, row_number() over (partition by name order by id) rn from items
)
select i1.id, i1.name, i1.price, sum(i2.price) 
from oitems i1
join  oitems i2 on i1.name = i2.name and i2.rn >= i1.rn - i1.precede_by and i2.rn <= i1.rn
group by i1.id, i1.name, i1.price

dbfiddle demo