如何选择一组随机记录，其中一列是唯一的？

Question

我今天一直在努力解决这个 SQL 查询要求，我想知道是否有人可以帮助我。

我有一张运动问题表。其中一列是与问题相关的团队。我的要求是返回一组独特的随机问题。

假设我们有下表并想要 5 个问题：

Question        Answer        Team
-----------------------------------
question 1      answer 1      team A
question 2      answer 2      team B
question 3      answer 3      team B
question 4      answer 3      team D
question 5      answer 3      team A
question 6      answer 3      team C
question 7      answer 3      team F
question 8      answer 3      team C
question 9      answer 3      team G
question 10     answer 3      team D

会返回一个有效的结果：

question 1      answer 1      team A
question 2      answer 2      team B
question 4      answer 3      team D
question 6      answer 3      team C
question 7      answer 3      team F

我觉得应该可以通过巧妙地使用 Distinct 和 Take 将其作为一个干净的 SQL 语句来完成，但我还不能做到这一点。

目前最好的解决方案来自Mladen Prajdic。我刚刚稍微更新了它以改善它的随机性：

SELECT TOP 10 * 
FROM    (SELECT ROW_NUMBER() OVER(PARTITION BY Team ORDER BY Team, NEWID()) AS RN, *
    FROM Question
    ) teams
WHERE   RN = 2
ORDER BY NEWID()

Answer 1

对于 sql 2005，您可以这样做：

select top 5 * 
from    (
            select ROW_NUMBER() over(partition by team order by team) as RN, *
            from @t 
        ) t
where RN = 1
order by NEWID()

Answer 2

这应该可以满足您的需求，在 oracle 中；对于不同的数据库，您显然需要使用他们的随机数源。可能有更好的方法；让我们希望其他人会向我们指出：p

select question, answer, team
from
(
select question, answer, team, r
from
(
select 
    question, 
    answer, 
    team,
    rank() over (partition by team order by dbms_random.value) r 
from questions
)
where r = 1
order by dbms_random.value
) where rownum<=5;

测试代码：

create table questions(question varchar2(16), answer varchar2(16), team varchar2(16));

insert into questions(question, answer, team)
values ('question 1',      'answer 1',      'team A');

insert into questions(question, answer, team)
values ('question 2',      'answer 2',      'team B');

insert into questions(question, answer, team)
values ('question 3',      'answer 3',      'team B');

insert into questions(question, answer, team)
values ('question 4',      'answer 3',      'team D');

insert into questions(question, answer, team)
values ('question 5',      'answer 3',      'team A');

insert into questions(question, answer, team)
values ('question 6',      'answer 3',      'team C');

insert into questions(question, answer, team)
values ('question 7',      'answer 3',      'team F');

insert into questions(question, answer, team)
values ('question 8',      'answer 3',      'team C');

insert into questions(question, answer, team)
values ('question 9',      'answer 3',      'team G');

insert into questions(question, answer, team)
values ('question 10',    'answer 3',      'team D');

commit;

Answer 3

在 PostgreSQL（有不同的 on）中，我可能会做这样的事情：

select distinct on (Team) Question, Answer, Team from test order by Team, random() limit 5;

刚刚测试过。似乎有效。