MySQL索引定义，有什么区别？

Question

What is the difference between these two indexes? I'm referring to the difference between both columns being under the 'PRIMARY' and the other way..

Answer 1

第一个意味着组合user_id, wallet_type是唯一的，第二个意味着只将visitid定义为唯一。

Answer 2

The first one is known as a multi-column, compound, or composite key. In this case, the primary key constraint is the combination of the two columns, in the listed order (user_id, then wallet_type). The two columns, combined, must be unique for each row.

So, the following data is okay:

user_id(pk)  wallet_type(pk)
-------------------------
1            1
1            2
2            2

Notice that rows can have duplicate user_id's or wallet_type's, but not both:

user_id(pk)  wallet_type(pk)
-------------------------
1            1
1            2
2            2
1            1      // Duplicate key

The second example is just a regular single-column key, where the just value in the visitid column must be unique:

visitid(pk)    session
-------------------------
1              1
1              2   // Duplicate key

There are many implications with this.

The multi-column primary key is larger, so it uses more storage and is slightly slower for MySQL to traverse.

Also, with InnoDB, inserts are negatively affected because primary keys are clustered indexes, which means they are stored in physical order. A larger index takes longer to update. Appending rows to the end is fine, however.

There are some benefits, such as the ability to use the multi-column index as a covering index.

For example, the following queries would be optimized with the multi-column primary key:

SELECT user_id, wallet_type FROM table1 WHERE user_id = 1
SELECT user_id, wallet_type FROM table1 WHERE user_id = 1 AND wallet_type = 2
SELECT user_id, wallet_type FROM table1 ORDER BY user_id, wallet_type
SELECT user_id, wallet_type FROM table1 WHERE user_id = 1 ORDER BY wallet_type
SELECT user_id, wallet_type FROM table1 WHERE user_id > 1 AND wallet_type > 2
SELECT user_id, wallet_type FROM table1 ORDER BY user_id, wallet_type

However, since columns are left prefixed, this query would not be optimized:

SELECT user_id, wallet_type FROM table1 WHERE wallet_type = 2

This is because user_id is not referenced in the WHERE clause, but user_id is the first column in the key. You can't skip a column.