B+ tree index uses (7) matching column prefix, matching value range (19)

The cost of indexing the last article ,b+ Trees take up a lot of space , Add, delete and correct b+ The index sorting of each node of the tree also has a great impact , It takes a long time , So there is no need to build indexes , It also introduces the leftmost principle of index and full value query .

B+ Tree index uses (6) The principle of left --mysql From entry to mastery （ eighteen )

Match column prefix

innoDB Add secondary indexes to other columns , Will sort him by column , Whether it's two-way linked list sorting between pages , Or one-way list sorting of data slot points in the page , They are all arranged by column value . such as name Column , If the name has TOM,Anny,Aone etc. ,1） First sort him alphabetically .2） The initials are A equally , Then sort him by the second letter .3） By analogy , Just sort him in the back . So when we use fuzzy query where name like ‘a%’; This percentage number can trigger index queries after it , But if where name like ‘%a’; Because the first letter cannot be found , This kind is not sorted well , As a result, you can only scan the whole disk , Discard index query .

for instance , When we have field columns url, There are data ,www.baidu.com,www.google,www.gov.com,www.wto.com.

At this time, we want to inquire com At the end of the , If you use where url like ‘%com’; Then the efficiency will be very low , Unable to trigger index query . In order to improve efficiency , Use index to query data , You can sort the data of the table in reverse , Change it to ,moc.udiab.www,moc.elgoog.www,nc.vog.www,gro.otw.www, It can be used at this time where url like ‘com%’; To trigger index queries .

Matching value range

Let's see idx_name_birthday_phone Indexes b+ Sketch Map , All records are sorted by index from small to large , For example, we use where name > ‘Anny’ and name < ‘barlow’; So this query process ：1） First in b+ Tree leaf node found name Greater than Anny Secondary index records for , Read the primary key , Using cluster index to query back to the table , Get all the user record data of the cluster index and send it to the client .2） According to the record found in the previous step , Along the position of the linked list （ Two way linked list association between pages , In page data , One way linked list Association ） Find the next secondary index record , Judge whether the record <barlow, If meet , Return to client .3） repeat 2 The operation of , Until it doesn't meet .

therefore , At this time, index query will be used , But the point needs attention , Be careful , Be careful ( Say important things three times )： If you perform a range query on multiple columns , Only the leftmost column of the index will be used when querying b+ Query the index of the tree .

such as where name > ‘Anny’ and name < ‘barlow’and birthday > ‘1990-01-01’; At this time, the query process ：1） Same as before , First name stay b+ The tree leaf node finds the column >Anny Primary key of , Tree back to table operation of querying clustered index with primary key , After the query , Query whether the next one meets <barlow.2） For the top name Different values , It's going on birhday Conduct birthday>’1990-01-01’ Filter .

In this way, queries are important for federated indexes , It's just name Index sorting is used in , And because biryhday The sorting condition is that name Only when the sorting is the same , Obtained at this moment name It's all different , So in birthday Index cannot be triggered during range query .

Accurate to a certain column , Range another column

When we use where name = ‘Anny’ and birthday > ‘1990-01-01’ and phone > ‘13200000000’; At this time, our query process is ：1）name Definitely use b+ The secondary index of the tree first queries the column value of the leaf node plus the primary key , The clustered index back to table operation returns all the data of the clustered index leaf nodes .2） because name In the same case ,birthday Will trigger index query , First in b+ Tree leaf node found >’1990-01-01’ Column value and primary key , Query all data through the primary key back table 3） because phone The premise of using index query is birthday identical , The front one is different birthday, Indexes phone Cannot use index query .