PostgreSQL技术内幕6：PostgreSQL索引技术

0. 简介

本文主要介绍PG的索引技术，包含PG支持的索引类型，语法，查看方式，以及其中B-Tree索引的原理解析和源码解读。

1.PG索引类型介绍

PG支持多种索引类型：B-tree、Hash、GiST、SP-GiST 、GIN 和 BRIN。不同的索引类型使用不同的算法来适应不同类型的查询，下面是其具体适用情况：
1）B-tree索引：是一种自平衡树，支持O(logn)的插入，删除，访问。
2）Hash索引：通过hash运算查找，只支持等于查找，不支持范围。
3）Gist索引：Gist是通用搜索树（generalized search tree）的缩写，和之前介绍的btree类似（一种平衡树）。但是它和btree不同的是，btree索引常常用来进行例如大于、小于、等于这些操作中，而在实际生活中很多数据其实不适用这种场景，例如地理数据、图像等等。因为Gist索引允许定义规则来将任意类型的数据分布到一个平衡的树中，并且允许定义一个方法使用此表示形式来让某些运算符访问。例如，对于空间数据，GiST索引可以使用 R树，以支持相对位置运算符（位于左侧，右侧，包含等），而对于树形图，R树可以支持相交或包含运算符。
4）SP-GiST索引：SP-GiST 代表空间分区 GiST，主要用于 GIS、多媒体、电话路由以及 IP 路由等数据的索引。
5）GIN索引: 倒排索引，主要用于搜索特定值是不是存在。
6）BRIN索引：BRIN 代表块区间索引（block range indexes），存储了连续物理范围区间内的数据摘要信息。BRIN 也相比 B-树索引要小很多，维护也更容易。对于不进行水平分区就无法使用 B-树索引的超大型表，可以考虑 BRIN。

2. PG创建索引说明及索引属性查看

2.1 创建说明

CREATE [ UNIQUE ] INDEX [ CONCURRENTLY ] [ [ IF NOT EXISTS ] name ] ON [ ONLY ] table_name [ USING method ]
    ( { column_name | ( expression ) } [ COLLATE collation ] [ opclass [ ( opclass_parameter = value [, ... ] ) ] ] [ ASC | DESC ] [ NULLS { FIRST | LAST } ] [, ...] )
    [ INCLUDE ( column_name [, ...] ) ]
    [ WITH ( storage_parameter [= value] [, ... ] ) ]
    [ TABLESPACE tablespace_name ]
[ WHERE predicate ]

主要参数说明：
UNIQUE:唯一索引，创建索引的列数据不能重复。
CONNCURRENTLY:构建索引时不会阻塞该表正在进行的插入，更新，删除。
METHOD:要使用的索引方法，如btree,hash等。
ASC：升序。
DESC：降序。

2.2 查看方式

2.2.1 查看PG默认支持的索引及对应的Handler类型

select * , obj_description(oid,'pg_am') from pg_am order by 1;

2.2.2 查看B树索引属性

select a.amname, p.name, pg_indexam_has_property(a.oid, p.name) from pg_am a, unnest(array['can_order','can_unique','can_multi_col','can_exclude']) p(name) where a.amname='btree' order by a.amname;

3. 索引选择

索引选择可以分两步进行考虑：1.是否建立索引：主要考虑索引的资源占用，对插入和更新的影响以及备份恢复的影响；2.索引类型选择：考虑创建索引以及使用查询的速度，索引大小，索引支持的类型等。

3.1 查看索引情况

1）查看索引

\di

2）查看所有和磁盘占用

select relname, pg_size_pretty(pg_relation_size(oid)) finsertrom pg_class where relname like 't\_%' or relname='t1' order by relname;

3）查看索引支持的类型

select opfname from pg_opfamily, pg_am where opfmethod = pg_am.oid and amname='btree' order by 1;

4）查看所有支持的操作符

select amop.amopopr::regoperator as opfamily_operator, amop.amopstrategy from pg_am am, pg_opfamily opf, pg_amop amop where opf.opfmethod = am.oid and amop.amopfamily =opf.oid and am.amname='btree' and opf.opfname='bool_ops' order by amopstrategy;

4.PG中B-Tree索引原理

PG中的BTree来源于论文《Efficient locking for concurrent operations on B-trees》，论文中是一种B+树的变形，增加了非叶子节点的右侧的连接，同时引入了引入了“High Key”（下述HK）用于描述当前节点子节点的最大值，PG在此基础上，增加了左侧兄弟节点的连接，对于并发更加友好（并发控制在后续并发控制章节介绍），其结构和特点如图：

1）树是平衡的
2）支持范围和等值查询以及排序操作
3）是多分支的，深度不会太深，大表4-5层就足够
4）双向互联，可以内部遍历，不需要回到根节点

4.1 页存储结构

PG的索引存储结构和其他页面存储结构一致：

linp用于索引itup，其存储了每个itup在页面中的实际位置。根据PostgreSQL中对BTree索引结构的描述，分为当前节点是否是最右节点两种类型。由于非最右节点需要一个字段来保存HK，故当对一个页面进行填充时，存在着以下两种方式：
（1）当前节点为非最右节点

1.将首先将itup3（最大的索引元组）复制到当前节点的右兄弟节点，然后将linp0指向itup3（HK）。
2.去掉linp3。使用linp0来指向页面中的HK。

（2）当前节点为最右节点

最右节点不需要HK，所以每个linp减一，linp3不需要使用

整体结构

（1）对于非叶子节点，itup指向下一个节点，而对于叶子节点，itup指向实际物理存储的位置。

（2）Special space中，实现了两个指针，分配用于指向左右兄弟节点。

（3）根据BTree的特性，索引元组为有序，第一个叶子节点中itup3实际为最大索引元组，即HK，第二个叶子节点中itup1实际为最小索引元组，两者相同，故指向了同一物理存储位置。

5.索引代码分析

5.1 不同索引结构解析

5.1.1 索引的Handler结构

每种索引会初始化不同的handler，定义其属性和行为，如创建时的操作，插入时的操作，新加一种索引可以定义不同的hanlder，这也体现了PG的良好的可扩展性。

typedef struct IndexAmRoutine
{
  NodeTag    type;

  /*
   * Total number of strategies (operators) by which we can traverse/search
   * this AM.  Zero if AM does not have a fixed set of strategy assignments.
   */
  uint16    amstrategies;
  /* total number of support functions that this AM uses */
  uint16    amsupport;
  /* opclass options support function number or 0 */
  uint16    amoptsprocnum;
  /* does AM support ORDER BY indexed column's value? */
  bool    amcanorder;
  /* does AM support ORDER BY result of an operator on indexed column? */
  bool    amcanorderbyop;
  /* does AM support backward scanning? */
  bool    amcanbackward;
  /* does AM support UNIQUE indexes? */
  bool    amcanunique;
  /* does AM support multi-column indexes? */
  bool    amcanmulticol;
  /* does AM require scans to have a constraint on the first index column? */
  bool    amoptionalkey;
  /* does AM handle ScalarArrayOpExpr quals? */
  bool    amsearcharray;
  /* does AM handle IS NULL/IS NOT NULL quals? */
  bool    amsearchnulls;
  /* can index storage data type differ from column data type? */
  bool    amstorage;
  /* can an index of this type be clustered on? */
  bool    amclusterable;
  /* does AM handle predicate locks? */
  bool    ampredlocks;
  /* does AM support parallel scan? */
  bool    amcanparallel;
  /* does AM support parallel build? */
  bool    amcanbuildparallel;
  /* does AM support columns included with clause INCLUDE? */
  bool    amcaninclude;
  /* does AM use maintenance_work_mem? */
  bool    amusemaintenanceworkmem;
  /* does AM store tuple information only at block granularity? */
  bool    amsummarizing;
  /* OR of parallel vacuum flags.  See vacuum.h for flags. */
  uint8    amparallelvacuumoptions;
  /* type of data stored in index, or InvalidOid if variable */
  Oid      amkeytype;

  /*
   * If you add new properties to either the above or the below lists, then
   * they should also (usually) be exposed via the property API (see
   * IndexAMProperty at the top of the file, and utils/adt/amutils.c).
   */

  /* interface functions */
  ambuild_function ambuild;
  ambuildempty_function ambuildempty;
  aminsert_function aminsert;
  aminsertcleanup_function aminsertcleanup;
  ambulkdelete_function ambulkdelete;
  amvacuumcleanup_function amvacuumcleanup;
  amcanreturn_function amcanreturn;  /* can be NULL */
  amcostestimate_function amcostestimate;
  amoptions_function amoptions;
  amproperty_function amproperty; /* can be NULL */
  ambuildphasename_function ambuildphasename; /* can be NULL */
  amvalidate_function amvalidate;
  amadjustmembers_function amadjustmembers;  /* can be NULL */
  ambeginscan_function ambeginscan;
  amrescan_function amrescan;
  amgettuple_function amgettuple; /* can be NULL */
  amgetbitmap_function amgetbitmap;  /* can be NULL */
  amendscan_function amendscan;
  ammarkpos_function ammarkpos;  /* can be NULL */
  amrestrpos_function amrestrpos; /* can be NULL */

  /* interface functions to support parallel index scans */
  amestimateparallelscan_function amestimateparallelscan; /* can be NULL */
  aminitparallelscan_function aminitparallelscan; /* can be NULL */
  amparallelrescan_function amparallelrescan; /* can be NULL */
} IndexAmRoutine;

下面简单看btree的handler初始化

Datum
bthandler(PG_FUNCTION_ARGS)
{
  IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);

  amroutine->amstrategies = BTMaxStrategyNumber;
  amroutine->amsupport = BTNProcs;
  amroutine->amoptsprocnum = BTOPTIONS_PROC;
  amroutine->amcanorder = true;
  amroutine->amcanorderbyop = false;
  amroutine->amcanbackward = true;
  amroutine->amcanunique = true;
  amroutine->amcanmulticol = true;
  amroutine->amoptionalkey = true;
  amroutine->amsearcharray = true;
  amroutine->amsearchnulls = true;
  amroutine->amstorage = false;
  amroutine->amclusterable = true;
  amroutine->ampredlocks = true;
  amroutine->amcanparallel = true;
  amroutine->amcanbuildparallel = true;
  amroutine->amcaninclude = true;
  amroutine->amusemaintenanceworkmem = false;
  amroutine->amsummarizing = false;
  amroutine->amparallelvacuumoptions =
    VACUUM_OPTION_PARALLEL_BULKDEL | VACUUM_OPTION_PARALLEL_COND_CLEANUP;
  amroutine->amkeytype = InvalidOid;

  amroutine->ambuild = btbuild;
  amroutine->ambuildempty = btbuildempty;
  amroutine->aminsert = btinsert;
  amroutine->aminsertcleanup = NULL;
  amroutine->ambulkdelete = btbulkdelete;
  amroutine->amvacuumcleanup = btvacuumcleanup;
  amroutine->amcanreturn = btcanreturn;
  amroutine->amcostestimate = btcostestimate;
  amroutine->amoptions = btoptions;
  amroutine->amproperty = btproperty;
  amroutine->ambuildphasename = btbuildphasename;
  amroutine->amvalidate = btvalidate;
  amroutine->amadjustmembers = btadjustmembers;
  amroutine->ambeginscan = btbeginscan;
  amroutine->amrescan = btrescan;
  amroutine->amgettuple = btgettuple;
  amroutine->amgetbitmap = btgetbitmap;
  amroutine->amendscan = btendscan;
  amroutine->ammarkpos = btmarkpos;
  amroutine->amrestrpos = btrestrpos;
  amroutine->amestimateparallelscan = btestimateparallelscan;
  amroutine->aminitparallelscan = btinitparallelscan;
  amroutine->amparallelrescan = btparallelrescan;

  PG_RETURN_POINTER(amroutine);
}

对于不同索引对应的函数和属性在系统初始化时，创建到pg_am、pg_opfamily等系统表中

# Index access method handlers
{ oid => '330', descr => 'btree index access method handler',
  proname => 'bthandler', provolatile => 'v', prorettype => 'index_am_handler',
  proargtypes => 'internal', prosrc => 'bthandler' },
{ oid => '331', descr => 'hash index access method handler',
  proname => 'hashhandler', provolatile => 'v',
  prorettype => 'index_am_handler', proargtypes => 'internal',
  prosrc => 'hashhandler' },
{ oid => '332', descr => 'gist index access method handler',
  proname => 'gisthandler', provolatile => 'v',
  prorettype => 'index_am_handler', proargtypes => 'internal',
  prosrc => 'gisthandler' },

5.2 BTree关键流程解析

5.2.1 构造函数btbuild

5.2.2 插入流程btinsert