Self adjustment process of MySQL index tree when adding data

At first, after you build a watch , Just one data page , Is part of the clustered index , And it's empty . If you insert data , Is to insert... Directly into this data page , There's no need to get him an index page ：

The initial data page is a root page , Each data page has a page directory based on the primary key by default , So it's OK to search according to the primary key , Directly find... In the only data page according to the page directory .

But with more and more data in the table , Your data page is full , A new data page is needed , Copy all the data in your root page , At the same time, make a new data page , Move according to the size of your primary key value , Let two new data pages sort according to the primary key value , The primary key value of the second data page is greater than that of the first data page ：

At this point, the root page is upgraded to the index page , In this root page ：

• Page number of two data pages

• The smallest primary key value in them

So it seems that the root page becomes the index page , Two data pages are referenced .

As you keep plugging in data , Then the data page keeps splitting , Split up more data pages . Your only index page , That is, there are more and more data page index entries stored in the root page , I can't even fit your index page , Then split an index page into two index pages , Then the root page continues to go up one level , Reference two index pages .

More and more data pages , The index page that the root page points to keeps splitting , Split more index pages , When you have too many index pages in the lower layer , It will cause your root page to point to too many index pages , The root page continues to split into multiple index pages , The root page moves up one level again . This is when you add, delete and modify , The whole process of cluster index maintenance .

Secondary indexes

Empathy , Like you name Field has an index , So at the beginning, you insert data , On the one hand, insert... In the unique data page of the cluster index , On the one hand name Index of field B+ Insert... Into the only data page of the tree .

With more and more follow-up data ,name The only data page in the index tree will also split , The whole process of division is the same as above , So when you insert data , The index tree itself will be maintained automatically .

Yours name Index of field B+ In the index page in the tree , In fact, in addition to storing the page number and the smallest name Other than field value , Each index page will also store the smallest name The primary key value corresponding to the field value . Because sometimes there will be the smallest number of the lower page pointed to by multiple index pages name Field values are the same , You have to judge according to the primary key .

For example, insert a new name value , He needs to be based on name The root page of the index tree starts , Find your new... Layer by layer name Which data page the value should be inserted into the leaf node .

In case of different index pages in the same layer pointing to different lower page numbers , but name Have the same value , You have to compare according to the primary key value ： new name The value is inserted into the data page with a larger primary key value ！