INSERT ... ON DUPLICATE KEY (do nothing)

Yes, use INSERT ... ON DUPLICATE KEY UPDATE id=id (it won't trigger row update even though id is assigned to itself).

If you don't care about errors (conversion errors, foreign key errors) and autoincrement field exhaustion (it's incremented even if the row is not inserted due to duplicate key), then use INSERT IGNORE like this:

INSERT IGNORE INTO <table_name> (...) VALUES (...)

HOW TO IMPLEMENT 'insert if not exist'?

1. REPLACE INTO

pros:

1. simple.

cons:

1. too slow.

2. auto-increment key will CHANGE(increase by 1) if there is entry matches unique key or primary key, because it deletes the old entry then insert new one.

2. INSERT IGNORE

pros:

1. simple.

cons:

1. auto-increment key will not change if there is entry matches unique key or primary key but auto-increment index will increase by 1

2. some other errors/warnings will be ignored such as data conversion error.

3. INSERT ... ON DUPLICATE KEY UPDATE

pros:

1. you can easily implement 'save or update' function with this

cons:

1. looks relatively complex if you just want to insert not update.

2. auto-increment key will not change if there is entry matches unique key or primary key but auto-increment index will increase by 1

4. Any way to stop auto-increment key increasing if there is entry matches unique key or primary key?

As mentioned in the comment below by @toien: "auto-increment column will be effected depends on innodb_autoinc_lock_mode config after version 5.1" if you are using innodb as your engine, but this also effects concurrency, so it needs to be well considered before used. So far I'm not seeing any better solution.