为什么向用于连接两个表的两个键之一添加索引会使执行时间比根本不使用任何索引要慢？

Question

I am executing queries where two tables authors and work_authors are being joined. I am using MySQL. authors and work_authors have a one-to-many relationship and authors contains around 7 million records and work_authors contains around 21 million records.

SELECT  t2.author_id, t1.work_id, t2.name 
FROM    work_author AS t1, 
        authors AS t2 
WHERE   t1.author_id = t2.id AND
        t2.name LIKE '%Tolkien%';

Firstly, there is no index at all in the database, ie no primary key or other indexes. Then the query on average takes 24 seconds on my machine. When adding index ie making the id-column of authors primary key, but not any index or primary key for the work_author table, the query takes on average 64 seconds on my machine.

Secondly, when I add an non-unique index to column author_id the work_author table the query on average takes 4 seconds on my machine.

How can this be the case? Why is the query taking longer when adding index to one of the attributes used to join than having no index at all? And then when adding index to the both join keys it executes faster than the other two alternatives. Anyone who can explain this?

Answer 1

No index can help LIKE '%Tolkien%' . However, ...

Consider adding a FULLTEXT index to authors.name and using

MATCH(name) AGAINST('+Tolkein' IN BOOLEAN MODE)

This will run a lot faster.

PRIMARY KEYs are important.

Using InnoDB is important.

Do use the JOIN...ON syntax instead of commajoin.

A many-to-many table is optimized by having two indexes as discussed here: http://mysql.rjweb.org/doc.php/index_cookbook_mysql#many_to_many_mapping_table