mysql中order by 和limit一起使用不当会导致效率极慢的4种优化方法-白红宇

今天从慢查询发现一条语句查询时间达6秒。结果只查出一条记录。

原语句如下

SELECT

biz_order_id, buyer_id, buyer_nick, gmt_create, gmt_modified, attributeCc, seller_id

FROM trade.biz_order

WHERE shop_id=20484 AND STATUS=4 AND gmt_create >= '2017-10-30 16:34:42' AND order_type = 6

ORDER BY gmt_create DESC, biz_order_id DESC

LIMIT 0,100;

执行计划

shop_id都有索引可却走了时间gmt_create的索引，rows=861665

优化方法3种：

1：强制走shop_id索引

SELECT

biz_order_id, buyer_id, buyer_nick, gmt_create, gmt_modified, attributeCc, seller_id

FROM trade.biz_order force index(idx_shop_id)

WHERE shop_id=20484 AND STATUS=4 AND gmt_create >= '2017-10-30 16:34:42' AND order_type = 6

ORDER BY gmt_create DESC, biz_order_id DESC

LIMIT 0,100;

2：用子查询：

select * from ( SELECT biz_order_id, buyer_id, buyer_nick, gmt_create, gmt_modified, attributeCc, seller_id

FROM trade.biz_order

WHERE shop_id=20484 AND STATUS=4 AND gmt_create >= '2017-10-30 16:34:42' AND order_type = 6

ORDER BY gmt_create DESC, biz_order_id DESC) t

LIMIT 0,100;执行计划如上一样

3：调换order by中的两个条件顺序

ORDER BY biz_order_id DESC ,gmt_create DESC limit 0,100;换成这样。我发现这样执行计划的rows=1.9万效果更好。

4：还有一种方法删除

gmt_create列的索引，原理和方法3差不多。

总结：mysql中的order ,limit一起使用时的顺序是这样的和oracle不一样

order -->limit-->where条件

而常规一般是where-->order-->limit。