在 sql 中，多次使用子查询的最佳语法

Question

从桌子上

create table seeplai.item(
  itm_id  serial  primary key,
  itp_id  int     not null references seeplai.item_type,
  itm_sq  int2    not null default 10,
  ... )
create index item_itp_sq on seeplai.item( itp_id, itm_sq );

使用item 记录作为基础，我希望找到具有相同itp_id 并且具有更少itm_sq 的最大itp_sq。现在使用两个子查询，在语法和逻辑上都是正确的，但是感觉不对。

select  max(itm_sq)
into    v_prev_sq
from    seeplai.item itm
where   itm.itp_id = (select itp_id from seeplai.item where itm_id=p_itm_id)
and     itm.itm_sq < (select itm_sq from seeplai.item where itm_id=p_itm_id);

我知道我使用了更好的方法，但是它的语法已经融入了我的脑海中

Answer 1

要获得一个最大的itm_sq，它小于你所在行中的那个：

SELECT  i2.itm_sq
FROM    seeplai.item i
JOIN    seeplai.item i2 USING (itp_id)
WHERE   i.itm_id = p_itm_id
AND     i2.itm_sq < i.itm_sq 
ORDER   BY i2.itm_sq DESC
LIMIT   1;

要打破平局（具有相同itm_sq 的多行），您可以ORDER BY 附加列。

这样您就不必重复使用任何东西。如果必须，可以使用子查询或CTE。

---

如果您需要之前的itm_sq 来为每个 共享相同itp_id 的项目，则窗口函数lag()（或倒序的lead()）将很有帮助：

SELECT  i2.*, lag(i2.itm_sq) OVER (ORDER BY i2.itm_sq) AS previous_itm_sq
FROM    seeplai.item i
JOIN    seeplai.item i2 USING (itp_id)
WHERE   i.itm_id = p_itm_id
ORDER   BY i2.itm_sq DESC;

Answer 2

您需要将表连接到自身：

select max(itm_sq)
into v_prev_sq
from seeplai.item itm
join seeplai.item itm2 on itm.itm_tp = itm2.itm_tp and itm2.itm_id=p_itm_id
join seeplai.item itm3 on itm.itm_sq < itm3.itm_sq and itm3.itm_id=p_itm_id

这是您查询的“直接翻译”，但我认为这种简化也可以：

select max(itm_sq)
into v_prev_sq
from seeplai.item itm
join seeplai.item itm2
    on itm.itm_tp = itm2.itm_tp 
    and itm2.itm_id = p_itm_id
    and itm.itm_sq < itm2.itm_sq

Answer 3

我不确定，但也许你可以使用 PostgreSQL 的 WITH Queries (Common Table Expressions)。

Answer 4

也许您正在寻找公用表表达式？ http://www.postgresql.org/docs/9.2/static/queries-with.html

虽然我认为实际上使用连接更容易做到这一点......

select  max(itm_0.itm_sq)
into    v_prev_sq
from    seeplai.item itm_0,
join itm_1
on itm_0.itm_tp = itm_1.itm_tp
and itm_0.itm_sq < itm_1.itm_sq
and itm_0.itm_id = itm_1.p_itm_id
group by itm_1.p_itm_id

如果需要，可能使用DISTINCT ON

Answer 5

只需将子查询的值存储在一个变量中：

declare @retVal varchar(20);
set @retVal = select itm_tp from seeplai.item where itm_id=p_itm_id

select  max(itm_sq)
into    v_prev_sq
from    seeplai.item itm
where   itm.itm_tp = @retVal
and     itm.itm_sq < @retVal;