题主假定按照主键检索。我们假定是等值查询。范围查询和表遍历情形可以在文末经推导得出。

primary key

A set of columns—and by implication, the index based on this set of columns—that can uniquely identify every row in a table. As such, it must be a unique index that does not contain any NULL values.

InnoDB requires that every table has such an index (also called the clustered index or cluster index), and organizes the table storage based on the column values of the primary key.

以上说明存储引擎是按照主键聚簇存储的。

With the exception of spatial indexes,InnoDBindexes are B-tree data structures. Spatial indexes use R-trees, which are specialized data structures for indexing multi-dimensional data. Index records are stored in the leaf pages of their B-tree or R-tree data structure. The default size of an index page is 16KB.

以上说明数据是按照B+树组织的。主键按照逻辑索引被存储在叶子结点上。所以我们只能继续去那里找数据。

An InnoDB page has seven parts:

Fil Header

Page Header

Infimum + Supremum Records

User Records

Free Space

Page Directory

Fil Trailer

以上说明物理页中没有现成的主键数据簇。

Similarly,InnoDBdoes not want to insert new rows according to the B-tree's key order (that would involve expensive shifting of large amounts of data), so it inserts new rows right after the end of the existing rows (at the top of the Free Space part) or wherever there's space left by a deleted row.

以上说明用户数据物理上在页内形成一个堆。

But by definition the records of a B-tree must be accessible in order by key value, so there is a record pointer in each record (the "next" field in the Extra Bytes) which points to the next record in key order. In other words, the records are a one-way linked list. SoInnoDBcan access rows in key order when searching.

以上说明这个物理堆形成一个逻辑有序的链表。假设这个查询可以命中。(细节) 存储引擎通过在页目录中二分查找到一个大致定位。并且最终找到了这条记录。接下来，我们要看看如何在物理记录中找到主键列数据。

The chart below shows the three parts of a physical record.

NameSizeField Start Offsets(F*1) or (F*2) bytesExtra Bytes6 bytesField Contentsdepends on content

Legend: The letter 'F' stands for 'Number Of Fields'.

The meaning of the parts is as follows:

The FIELD START OFFSETS is a list of numbers containing the information "where a field starts".

The EXTRA BYTES is a fixed-size header.

The FIELD CONTENTS contains the actual data.

以上说明一个物理记录分为三部分。乍一看我们也许需要继续查找主键在什么位置。

Clustered indexes

The clustered key (PRIMARY KEY) has one of the more complex record structures:

Cluster Key Fields: The cluster key fields, concatenated together (literally). InnoDB just concatenates the raw bytes of its internal storage formats per column type together into a single byte stream.

以上说明主键在物理记录中聚簇，可以直接提取。至此题主应该了解等值查询究竟都经历了些什么。至于范围查询和表遍历，找到第一个记录之后按照链表结构依次提交给游标即可。

所以主键的遍历也仅仅是不用在非键列中再查询一次而已。当然在索引和聚簇数据结构的帮助下，主键的查询投影开销一定是比较小的。

References: