Question

我有三张大表如下......

property
--------
property_id
other_prop_data

transfer_property
-----------------
property_id
transfer_id

transfer
--------
transfer_id
contract_date
transfer_price

我想返回“2012-01-01”和“2012-06-30”之间发生的所有转移的唯一属性ID列表。这是我到目前为止的代码......

SELECT *
FROM property p
JOIN
(
  SELECT t.transfer_id, t.contract_date, t.transfer_price::integer, tp.property_id
  FROM transfer t
  LEFT JOIN transfer_property tp ON tp.transfer_id = t.transfer_id
  WHERE t.contract_date BETWEEN '2012-01-01' AND '2012-06-30'
) transfer1 ON transfer1.property_id = p.property_id

AND NOT EXISTS
(
  SELECT transfer2.transfer_id
  FROM
  (
    SELECT t.transfer_id, t.contract_date, t.transfer_price::integer, tp.property_id
    FROM transfer t
    LEFT JOIN transfer_property tp ON tp.transfer_id = t.transfer_id
    WHERE t.contract_date BETWEEN '2012-01-01' AND '2012-06-30'
  ) AS transfer2
  WHERE transfer2.property_id = transfer1.property_id
  AND transfer2.contract_date > transfer1.contract_date
)

这可行（我认为），但速度很慢。

我在...中发现了几个类似的查询 https://stackoverflow.com/questions/tagged/greatest-n-per-group ...但是我找到的大多数都是同一个表的自连接，而不是如上所述加入关系表。

我知道在MySQL中您可以使用用户变量，但我不知道如何在PostgreSQL中执行此操作，或者在这种情况下它是否是理想的解决方案。

是否有人对如何改进此查询有任何建议（甚至是如何使用与上面完全不同的方法来完成此操作）？

非常感谢任何帮助。谢谢！

此致

克里斯

PS：还尝试了DISTINCT和MAX的变体，但不相信他们会按照我使用它的方式选择最近日期的记录。

编辑：对不起，我还应该补充一点，我在PGADMIN 1.12.3中运行我的查询

Answer 1

尝试在PostgreSQL中使用ROW_NUMBER() OVER。这是一个SQLFiddle example：

SELECT *
FROM property p
JOIN
(
  SELECT t.transfer_id, t.contract_date, 
         t.transfer_price::integer, tp.property_id,
         row_number() over 
           (PARTITION BY tp.property_id 
            ORDER BY t.contract_date desc) as rn
  FROM transfer t
  LEFT JOIN transfer_property tp 
        ON tp.transfer_id = t.transfer_id
  WHERE t.contract_date BETWEEN '2012-01-01' 
                            AND '2012-06-30'
) transfer1 
       ON transfer1.property_id = p.property_id
where transfer1.rn = 1

Answer 2

给出骨架表：

create table property( property_id serial primary key );

create table transfer(
    transfer_id serial primary key,
    contract_date date not null
);

create table transfer_property (
    property_id integer references property(property_id),
    transfer_id integer references transfer(transfer_id)
);

和数据：

insert into property
select nextval('property_property_id_seq') 
from generate_series(1,10);

insert into transfer 
select nextval('transfer_transfer_id_seq'), 
       DATE '2012-01-01' + x * INTERVAL '1 month'
from generate_series(1,10) x;

-- Repeat this 4 or 5 times to produce a pile of duplicate entries
insert into transfer_property (transfer_id,property_id)
select transfer_id, property_id
from property cross join transfer
order by random()
limit 40;

使用：

select distinct property_id 
from transfer_property tp inner join transfer t on (tp.transfer_id = t.transfer_id)
where t.contract_date between  '2012-01-01' and '2012-06-30';

不足/曲解？ 请发布示例数据和显示有意义的关系和预期结果的真实模式。

Answer 3

“我想返回”2012-01-01“和”2012-06-30“之间发生的所有转移的唯一属性ID列表。”

对我而言，显示为：

SELECT DISTINCT tp.property_id
  FROM transfer t
  JOIN transfer_property tp ON tp.transfer_id = t.transfer_id
  WHERE t.contract_date BETWEEN '2012-01-01' AND '2012-06-30'
     ;

现在把它放在CTE或子查询中，你就完成了：

WITH x1 AS (
      SELECT DISTINCT tp.property_id AS property_id
      FROM transfer t
      JOIN transfer_property tp ON tp.transfer_id = t.transfer_id
      WHERE t.contract_date BETWEEN '2012-01-01' AND '2012-06-30'
      )
SELECT ...
FROM property p
JOIN x1 ON x1.property_id = p.property_id
    ;

我不明白NOT EXISTS子查询的目的。您只对MAX感兴趣吗？

更新：出现（从标题中）您只需要maxdate。可以通过你不存在的构造，或者子查询中的这个MAX（...）来完成;像...：

WITH m1 AS (
      SELECT DISTINCT tp.property_id AS property_id
        , MAX(t.contract_date) AS contract_date
      FROM transfer t
      JOIN transfer_property tp ON tp.transfer_id = t.transfer_id
      WHERE t.contract_date BETWEEN '2012-01-01' AND '2012-06-30'
        GROUP BY tp.property_id
      )
SELECT ...
FROM property p
JOIN m1 ON m1.property_id = p.property_id
    ;

如何改进查询以选择使用PostgreSQL中使用连接表的MAX日期的记录？

3 个答案: