1. Create a table
CREATE TABLE IF NOT EXISTS `article` (
`id` BIGINT(10) NOT NULL AUTO_INCREMENT,
`author_id` INT(10) NOT NULL,
`category_id` INT(10) NOT NULL,
`views` INT(10) NOT NULL,
`comments` INT(10) NOT NULL,
`title` VARCHAR(10) COLLATE utf8_unicode_ci NOT NULL,
`content` TEXT COLLATE utf8_unicode_ci NOT NULL,
PRIMARY KEY (`id`)
) ENGINE=INNODB AUTO_INCREMENT=4 DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci;
2. Add data
INSERT INTO article(author_id,category_id,views,comments,title, content)
VALUES (1,1,1,1,'1','1'), (2,2,2,2,'2','2'),(1,1,3,3,'3','3');
3. Query
SELECT * FROM article;
4. Demand
Query category_id case 1 is greater than 1 and comments, views the most article_id.
5. SQL
SELECT id, author_id FROM article WHERE category_id =1 AND comments>1 ORDER BY views DESC LIMIT 1;
6. index optimization analysis
type = ALL: full table scan
key = NULL: do not use the index,
Extra also appeared Using filesort, had a secondary sorting
Conclusion: garbage, it needs to be optimized.
(1) for the first time to create an index
First View original article index
It is a primary key primary index. And where, order by using the column did not have any relationship, so do not take the index to normal.
The following create a multi-column index based on where, order by field
create index idx_c_c_v on article(category_id,comments, views);
View index again
Note that the index inx_c_c_v, Seq_in_index indicates the index column for the order, more than, for example, said in use inx_c_c_v, first find category_id, find comments, and finally find views.
After creating the index, we then analyze SELECT id, author_id FROM article WHERE category_id = 1 AND comments> 1 ORDER BY views DESC LIMIT 1 of the implementation plan.
type = range: scan range, than the previous full table scan type = ALL higher efficiency.
key = inx_c_c_v: Use the index was created. OK,
Extra: Using filesort,. . . . . . The pit cargo is still there. . . .
Then we look at the following execution plan
type = ref: non-unique index scan, the efficiency is higher than the above range of the scanning range it
key = inx_c_v_v: use of the index
ref = const,const : 两个常量,优秀!
Extra ,干掉了Using filesorting
通过对比,我们不难发现,inx_c_c_v不变的情况下,仅是由于查询语句的不同,直接造成执行计划的巨大差异。 其根本原因是comment> 1是个type=range范围查询,它会导致该索引列之后索引列失效,即是(category --√--- comments -----×--views)
所以,index_c_v_v这个索引不行呀,都是因为comments造成的, 所以我们建索引时,不要它,试试!
drop index idx_c_c_v on article;
(2) 第二次创建索引
create index idx_c_v on article(category_id,views);
查看表索引。。。
最后来看一下explain SELECT id, author_id FROM article WHERE category_id =1 AND comments>1 ORDER BY views DESC LIMIT 1\G
type = ref : 完美
ref = const : 完美
Extra ,没有Using filesort, 也算完美!
总之,还可以吧!
7. 总结
相同的索引 ,select 语句的差别也会造成不同的执行计划,性能差别距大
创建索引时,范围查询需要 特别注意。