openGauss内核分析(十):数据库索引的创建过程

211 阅读4分钟

数据库索引可以提高数据的访问速度,openGauss支持唯一索引、多字段索引、部分索引和表达式索引。行存表(ASTORE存储引擎)支持的索引类型:btree(行存表缺省值)、hash、gin、gist。行存表(USTORE存储引擎)支持的索引类型:ubtree。列存表支持的索引类型:Psort(列存表缺省值)、btree、gin。全局临时表不支持GIN索引和Gist索引。

如上一篇分析数据库表的创建过程,standard_ProcessUtility函数会根据nodeTag(parsetree)的值来确定sql的操作类型,create table一般都是进入T_CreateStmt分支调用CreateCommand函数。create index则进入T_IndexStmt分支调用DefineIndex函数。在调用DefineIndex前会首先执行函数transformIndexStmt,如果语句没有指定索引类型则会使用缺省值。

if (stmt->accessMethod == NULL) {

        if (!isColStore) {

            /* row store using btree index by default */

            if (!RelationIsUstoreFormat(rel)) {

                stmt->accessMethod = DEFAULT_INDEX_TYPE; // 行存表ASTORE存储引擎缺省值btree

            } else {

                stmt->accessMethod = DEFAULT_USTORE_INDEX_TYPE; // 行存表USTORE存储引擎缺省值ubtree

            }

……

            } else {

                /* column store using psort index by default */

                stmt->accessMethod = DEFAULT_CSTORE_INDEX_TYPE; // 列存表缺省值psort

            }

        }
#define DEFAULT_INDEX_TYPE	"btree"
#define DEFAULT_HASH_INDEX_TYPE "hash"
#define DEFAULT_CSTORE_INDEX_TYPE "psort"
#define DEFAULT_GIST_INDEX_TYPE	"gist"
#define CSTORE_BTREE_INDEX_TYPE "cbtree"
#define DEFAULT_GIN_INDEX_TYPE "gin"
#define CSTORE_GINBTREE_INDEX_TYPE "cgin"
#define DEFAULT_USTORE_INDEX_TYPE "ubtree"

普通表索引

DefineIndex函数

DefineIndex为创建索引主入口函数。通常创建索引以Share锁锁定表,允许并发查询,但禁上对表进行修改。如果创建索引时指定关键字CONCURRENTLY以不阻塞DML的方式创建索引,即允许读取和更新表,以ShareUpdateExclusiveLock锁锁定表

image.png

lockmode = concurrent ? ShareUpdateExclusiveLock : ShareLock;

rel = heap_open(relationId, lockmode);

如果没有指定索引名,ChooseIndexName根据规则生成索引名

/*

     * Select name for index if caller didn't specify.

     */

    indexRelationName = stmt->idxname;

    if (indexRelationName == NULL) {

        indexRelationName = ChooseIndexName(RelationGetRelationName(rel),

            namespaceId,

            indexColNames,

            stmt->excludeOpNames,

            stmt->primary,

            stmt->isconstraint);

为index_create函数构造参数IndexInfo结构体

/*

     * Prepare arguments for index_create, primarily an IndexInfo structure.

     * Note that ii_Predicate must be in implicit-AND format.

     */

    indexInfo = makeNode(IndexInfo);

Index_create函数创建索引

/*

     * Make the catalog entries for the index, including constraints. Then, if

     * not skip_build || concurrent, actually build the index.

     */

indexRelationId = index_create(rel,

……

关闭表并返回索引表id

heap_close(rel, NoLock);
return indexRelationId;

Index_create函数

打开系统表pg_class

pg_class = heap_open(RelationRelationId, RowExclusiveLock);

heap_create创建relcache和索引物理文件

/*

     * create the index relation's relcache entry and physical disk file. (If

     * we fail further down, it's the smgr's responsibility to remove the disk

     * file again.)

     */

    StorageType storage_type = RelationGetStorageType(heapRelation);

    indexRelation = heap_create(indexRelationName, namespaceId, tableSpaceId, indexRelationId, relFileNode,

        RELATION_CREATE_BUCKET(heapRelation) ? heapRelation->rd_bucketoid : InvalidOid, indexTupDesc, relKind,

        relpersistence, isLocalPart, false, shared_relation, mapped_relation, allow_system_table_mods,

        REL_CMPRS_NOT_SUPPORT, (Datum)reloptions, heapRelation->rd_rel->relowner, skip_create_storage,

        isUstore ? TAM_USTORE : TAM_HEAP, /* XXX: Index tables are by default HEAP Table Type */

        relindexsplit, storage_type, extra->crossBucket, accessMethodObjectId);

将索引表元信息存入系统表pg_class

/*

     * store index's pg_class entry

     */

    InsertPgClassTuple(

        pg_class, indexRelation, RelationGetRelid(indexRelation), (Datum)0, reloptions, relKind, NULL);

 

    /* done with pg_class */

    heap_close(pg_class, RowExclusiveLock);

将索引表元信息存入系统表pg_index

UpdateIndexRelation(indexRelationId,

        heapRelationId,

        indexInfo,

……

Index_build建立索引

} else if (extra && (!extra->isPartitionedIndex || extra->isGlobalPartitionedIndex)) {

        /* support regular index or GLOBAL partition index */

        index_build(heapRelation, NULL, indexRelation, NULL, indexInfo, isprimary, false, PARTITION_TYPE(extra));

    }

index_build

index_build调用index_build_storage,如果创建的是btree索引最终调用btbuild,如果是hash索引最终调用hashbuild,如果是psort则最终调用psortbuild,更多索引访问方法的信息可查看系统表pg_am。

stats = index_build_storage(targetHeapRelation, targetIndexRelation, indexInfo);

static IndexBuildResult* index_build_storage(Relation heapRelation, Relation indexRelation, IndexInfo* indexInfo)

{

    RegProcedure procedure = indexRelation->rd_am->ambuild;

    Assert(RegProcedureIsValid(procedure));

 

    IndexBuildResult* stats = (IndexBuildResult*)DatumGetPointer(OidFunctionCall3(

        procedure, PointerGetDatum(heapRelation), PointerGetDatum(indexRelation), PointerGetDatum(indexInfo)));

    Assert(PointerIsValid(stats));

    if (RELPERSISTENCE_UNLOGGED == heapRelation->rd_rel->relpersistence) {

        index_build_init_fork(heapRelation, indexRelation);

    }

    return stats;

}

btree索引的procedure为btbuild。

#0  btbuild (fcinfo=0x7fb4f9c63920) at nbtree.cpp:63

#1  0x00000000011fc07d in OidFunctionCall3Coll (functionId=338, collation=0, arg1=140415366247440, arg2=140415366237480, arg3=140415402419864) at fmgr.cpp:1857

#2  0x0000000000c16b77 in index_build_storage (heapRelation=0x7fb50006b410, indexRelation=0x7fb500068d28, indexInfo=0x7fb5022ea698) at index.cpp:2475

#3  0x0000000000c18097 in index_build (heapRelation=0x7fb50006b410, heapPartition=0x0, indexRelation=0x7fb500068d28, indexPartition=0x0, indexInfo=0x7fb5022ea698, isPrimary=false, isreindex=false,

partitionType=INDEX_CREATE_NONE_PARTITION, parallel=true) at index.cpp:2834

以上函数调用栈如下

#0  index_build_storage 
#1  index_build
#2  index_create
#3  DefineIndex
#4  standard_ProcessUtility 
#5  gsaudit_ProcessUtility_hook
#6  pgaudit_ProcessUtility 
#7  hypo_utility_hook 
#8  ProcessUtility 
#9  PortalRunUtility
#10  PortalRun 
#11  exec_simple_query 
#12  PostgresMain

分区表索引

创建普通表索引语法如下

CREATE [ UNIQUE ] INDEX [ CONCURRENTLY ] [ [schema_name.]index_name ]

    { ON table_name [ USING method ] | [ USING method ] ON table_name }

    ({ { column_name | ( expression ) } [ COLLATE collation ] [ opclass ] [ ASC | DESC ] [ NULLS { FIRST | LAST } ] }[, ...] )

    [ index_option ]

    [ WHERE predicate ];

创建分区表索引语法

CREATE [ UNIQUE ] INDEX [ [schema_name.]index_name ]
    { ON table_name [ USING method ] | [ USING method ] ON table_name }
    ( {{ column_name | ( expression ) } [ COLLATE collation ] [ opclass ] [ ASC | DESC ] [ NULLS LAST ] }[, ...] )
    [ LOCAL [ ( { PARTITION index_partition_name [ TABLESPACE index_partition_tablespace ] } [, ...] ) ] | GLOBAL ]
    [ index_option ]


两者执行流程基本一致,分区表索引在DefineIndex中是遍历每个分区调用partition_index_create。