如何在 PostgreSQL 查询中对不同的元组进行排序

Question

我正在尝试在 Postgres 中提交一个只返回不同元组的查询。在我的示例查询中，我不希望一个cluster_id/feed_id 组合存在多次的重复条目。如果我做一个简单的：

select distinct on (cluster_info.cluster_id, feed_id) 
   cluster_info.cluster_id, num_docs, feed_id, url_time 
   from url_info 
   join cluster_info on (cluster_info.cluster_id = url_info.cluster_id) 
   where feed_id in (select pot_seeder from potentials) 
   and num_docs > 5 and url_time > '2012-04-16';

我明白了，但我也想根据num_docs 进行分组。因此，当我执行以下操作时：

select distinct on (cluster_info.cluster_id, feed_id) 
   cluster_info.cluster_id, num_docs, feed_id, url_time 
   from url_info join cluster_info 
   on (cluster_info.cluster_id = url_info.cluster_id) 
   where feed_id in (select pot_seeder from potentials) 
   and num_docs > 5 and url_time > '2012-04-16' 
   order by num_docs desc;

我收到以下错误：

ERROR:  SELECT DISTINCT ON expressions must match initial ORDER BY expressions
LINE 1: select distinct on (cluster_info.cluster_id, feed_id) cluste...

我想我明白为什么会出现错误（除非我以某种方式明确描述该组，否则无法按元组分组）但我该怎么做呢？或者如果我对错误的解释不正确，有没有办法实现我的初始目标？

Answer 1

最左边的ORDER BY 项不能与DISTINCT 子句的项不一致。我引用the manual about DISTINCT：

DISTINCT ON 表达式必须匹配最左边的ORDER BY 表达式。 ORDER BY 子句通常包含额外的 确定所需行优先级的表达式 每个DISTINCT ON 组。

试试：

SELECT *
FROM  (
    SELECT DISTINCT ON (c.cluster_id, feed_id) 
           c.cluster_id, num_docs, feed_id, url_time 
    FROM   url_info u
    JOIN   cluster_info c ON (c.cluster_id = u.cluster_id) 
    WHERE  feed_id IN (SELECT pot_seeder FROM potentials) 
    AND    num_docs > 5
    AND    url_time > '2012-04-16'
    ORDER  BY c.cluster_id, feed_id, num_docs, url_time
           -- first columns match DISTINCT
           -- the rest to pick certain values for dupes
           -- or did you want to pick random values for dupes?
    ) x
ORDER  BY num_docs DESC;

或者使用GROUP BY:

SELECT c.cluster_id
     , num_docs
     , feed_id
     , url_time 
FROM   url_info u
JOIN   cluster_info c ON (c.cluster_id = u.cluster_id) 
WHERE  feed_id IN (SELECT pot_seeder FROM potentials) 
AND    num_docs > 5
AND    url_time > '2012-04-16'
GROUP  BY c.cluster_id, feed_id 
ORDER  BY num_docs DESC;

如果c.cluster_id, feed_id 是您在SELECT 列表中包含列的所有（在这种情况下都是）表的主键列，那么这仅适用于PostgreSQL 9.1 或更高版本。

否则，您需要GROUP BY 其余列或汇总或提供更多信息。