優化sql，原sql：
select distinct q.* from aws_question q inner join aws_topic_question tq
on ( tq.topic_id=12564 or tq.topic_id=4558 or tq.topic_id=934 ) and q.question_id = tq.question_id
where q.category_id = 1 and q.status = 'CHECKED' and q.add_time <= '1496905791' and q.fenlei_id in (100,101)
order by q.add_time desc limit 0,5
條件: 除了兩張表的主鍵索引，給aws_quesrion表 添加索引 add_time_fenlei_id（add_time和fenlei_id的結合）
(1) exists : 5 rows in set (0.01 sec)
select q.* from aws_question q
where exists (
select tq.question_id from aws_topic_question tq
where tq.topic_id in (12564,4558, 934) and tq.question_id=q.question_id
group by question_id
)
and q.category_id = 1 and q.status = 'CHECKED' and q.add_time <= '1496905791' and q.fenlei_id in (100,101)
order by q.add_time desc limit 0,5
=》執行結果： 5 rows in set (0.01 sec)
(2)in : 5 rows in set (46.91 sec)
select q.* from aws_question q
where q.question_id in (
select tq.question_id from aws_topic_question tq
where tq.topic_id in (12564,4558, 934)
group by question_id
)
and q.category_id = 1 and q.status = 'CHECKED' and q.add_time <= '1496905791' and q.fenlei_id in (100,101)
order by q.add_time desc limit 0,5
=》執行結果： 5 rows in set (46.91 sec)
三條sql的explain分析：
最后結論： exists效率最高，望高手指教！！！