Question

我正在尝试做的是加快此查询速度。目前它的速度大约是15秒以上。 r.email 和 v.sent 以及 r.id ， r.viewed 上都有索引和 r.saved 。 EXPLAIN显示两个键都在使用中，但我发现它正在使用Using where; Using temporary; Using filesort。

有关改进此查询的任何想法吗？

SELECT r.id, r.top, r.email
FROM request as r
WHERE r.viewed = 0 AND r.saved = 0 AND r.email NOT IN ( 
    SELECT v.sent FROM viewing as v WHERE v.sent = r.email
)
GROUP BY r.email 
ORDER BY r.top desc, r.date
LIMIT 100;

Answer 1

MySQL在IN子句中为多个记录使用临时表。尝试

SELECT r.id, r.top, r.email
FROM request as r
left outer join viewing v on v.sent = r.email
WHERE r.viewed = 0 AND r.saved = 0 AND v.sent is null
GROUP BY r.email 
ORDER BY r.top desc, r.date
LIMIT 100;

Answer 2

使用LEFT JOIN代替NOT IN：为什么......？见here

SELECT r.id, r.top, r.email
FROM request as r
     LEFT JOIN viewing v
        ON r.email= v.sent
WHERE r.viewed = 0 AND 
      r.saved = 0 AND 
      v.sent IS NULL
GROUP BY r.email
ORDER BY r.top DESC, r.date
LIMIT 100;

为获得最佳性能，请考虑在表格上添加以下索引，最好是covering indexes：

ALTER TABLE request ADD INDEX ix1 (email, viewed, saved, top, date);

ALTER TABLE viewing ADD INDEX ix1 (sent);

Answer 3

SELECT
  r.id,
  r.top,
  r.email
FROM
  request r
LEFT OUTER JOIN
  viewing v
ON
  v.sent = r.email
WHERE
  r.viewed = 0
AND
  r.saved = 0
AND
  v.sent IS NULL
GROUP BY
  r.email
ORDER BY
  r.top DESC, r.date
LIMIT 100

Answer 4

显然MySQL有subquery performance issues。

幸运的是，有办法解决这个问题。具体来说，将子查询转换为联接，或者在特定情况下转换为exclusion join。

加快MySQL查询中的速度

4 个答案: