Question

将记录合并到一个记录中时删除重复项的最佳方法是什么？

我有一种情况，表会跟踪玩家姓名及其记录，如下所示：

stats
-------------------------------
nick     totalgames     wins   ...
John     100            40
john     200            97
Whistle  50             47
wHiStLe  75             72
...

我需要合并nick重复的行（当忽略大小写时）并将记录合并为一个，如下所示：

    stats
    -------------------------------
    nick     totalgames     wins   ...
    john     300            137
    whistle  125            119
    ...

我在Postgres做这个。最好的方法是什么？

我知道通过这样做，我可以获得存在重复项的名称：

select lower(nick) as nick, totalgames, count(*) 
from stats 
group by lower(nick), totalgames
having count(*) > 1;

我想到了这样的事情：

update stats
set totalgames = totalgames + s.totalgames
from (that query up there) s
where lower(nick) = s.nick

除此之外不能正常工作。我似乎仍然无法删除包含重复名称的其他重复行。我能做什么？有什么建议吗？

Answer 1

SQL Fiddle

这是您的更新：

 UPDATE stats
 SET totalgames = x.games, wins = x.wins
 FROM (SELECT LOWER(nick) AS nick, SUM(totalgames) AS games, SUM(wins) AS wins
     FROM stats
      GROUP BY LOWER(nick) ) AS x
 WHERE LOWER(stats.nick) = x.nick;

以下是删除重复行的删除：

 DELETE FROM stats USING stats s2
 WHERE lower(stats.nick) = lower(s2.nick) AND stats.nick < s2.nick;

（请注意，'update ... from'和'delete ... using'语法是Postgres特有的，并且从this answer和this answer无耻地被盗。）

您可能还希望运行此命令以使所有名称都包含在内：

 UPDATE STATS SET nick = lower(nick);

Aaaand在'nick'的小写版本上抛出一个唯一索引（或者向该列添加一个约束以禁止非小写值）：

CREATE UNIQUE INDEX ON stats (LOWER(nick));

Answer 2

可以使用RETURNING在一个语句中完成。

-- The data
CREATE TABLE stats
        ( nick VARCHAR PRIMARY KEY
        , totalgames INTEGER NOT NULL DEFAULT 0
        , wins INTEGER NOT NULL DEFAULT 0
        );

INSERT INTO stats(nick, totalgames,wins) VALUES
 ( 'John', 100, 40) ,( 'john', 200, 97)
,( 'Whistle', 50, 47) ,( 'wHiStLe', 75, 72)
, ( 'Single', 42, 13 ) -- this person has only one record
        ;
SELECT * FROM stats;

-- The query:
WITH upd AS (
        UPDATE stats dst
        SET totalgames = src.totalgames
                , wins = src.wins
        FROM ( SELECT MIN(nick) AS nick -- pick the "lowest" nick as the canonical nick
                , SUM(totalgames) AS totalgames
                , SUM(wins) AS wins
                FROM stats
                GROUP BY lower(nick)
                ) src
        WHERE dst.nick = src.nick
        RETURNING dst.nick -- only the records that have been updated
        )
-- Delete the records that were NOT updated.
DELETE FROM stats del
WHERE NOT EXISTS (
        SELECT * FROM upd
        WHERE upd.nick = del.nick
        )
        ;

SELECT * FROM stats;

输出：

INSERT 0 5
  nick   | totalgames | wins 
---------+------------+------
 John    |        100 |   40
 john    |        200 |   97
 Whistle |         50 |   47
 wHiStLe |         75 |   72
 Single  |         42 |   13
(5 rows)

DELETE 2
  nick   | totalgames | wins 
---------+------------+------
 wHiStLe |        125 |  119
 john    |        300 |  137
 Single  |         42 |   13
(3 rows)

Answer 3

我认为在一个查询中最简单的方法是使用common table expressions：

with cte as (
    delete from stats
    where lower(nick) in (
      select lower(nick) from stats group by lower(nick) having count(*) > 1
    )
    returning *
)
insert into stats(nick, totalgames, wins)
select lower(nick), sum(totalgames), sum(wins)
from cte
group by lower(nick);

如您所见，在cte中我删除重复项并返回已删除的行，然后将已分组的已删除数据插回到表中。

请参阅sql fiddle demo

Answer 4

UPDATE stats SET totalgames=s.totalgames, wins=s.wins
FROM (SELECT lower(nick) AS nick,SUM(totalgames) AS totalgames,SUM(wins) AS wins FROM stats GROUP BY lower(nick))s WHERE lower(nick)=s.nick; DELETE FROM stats WHERE lower(nick) IN (SELECT lower(nick) FROM stats GROUP BY lower(nick) HAVING COUNT(*)>1) AND NOT lower(nick) IN (SELECT first(nick) FROM stats GROUP BY lower(nick)应该有效。

SQL：如何合并不区分大小写的重复项

4 个答案: