2021SC@SDUSC

B-Tree建立过程

上次我们讲了关于nbtree的相关算法原理和相关的数据结构，这次我们具体分析nbtree的关于B-Tree索引创建流程的源码。

IndexAmRoutine

我们在第二篇文章讲过索引的使用是由下层接口函数和上层统一操作函数来进行的，在nbtree.c中bthanlder()函数就是返回一个IndexAmRoutine，统一管理下层接口函数，为上层操作函数统一调度。

//Datum
bthandler(PG_FUNCTION_ARGS)
{IndexAmRoutine *amroutine = makeNode(IndexAmRoutine);//返回IndexAmRoutine的对象amroutine//下面是对amroutine相关参数和处理函数进行赋值amroutine->amstrategies = BTMaxStrategyNumber;amroutine->amsupport = BTNProcs;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->amcaninclude = true;amroutine->amkeytype = InvalidOid;
//关于bttree索引建立，插入，删除，扫描等处理函数进行赋予具体的操作函数，
//这些函数均在之后的代码中有声明amroutine->ambuild = btbuild;amroutine->ambuildempty = btbuildempty;amroutine->aminsert = btinsert;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->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);
}

BTBuildState

当建立索引时为了保存索引元组建立了BTBuildState的数据结构，其代码如下

typedef struct BTBuildState
{bool       isunique;//是否为唯一索引bool      havedead;//是否有标记删除的元组Relation   heap;//相关的堆BTSpool    *spool;//保存生成的元组BTSpool    *spool2;//保存"dead tuple"对应的元组double      indtuples;//保存的元组的总和BTLeader   *btleader;
} BTBuildState;

BTWriteState

typedef struct BTWriteState
{Relation   heap;Relation   index;BTScanInsert inskey;      /* generic insertion scankey */bool     btws_use_wal;   /* 是否开启wal日志 */BlockNumber btws_pages_alloced; /* #申请的页面的块号 */BlockNumber btws_pages_written; /* #写入的页面的块号*/Page        btws_zeropage;  /* 需要填充0值得页面*/
} BTWriteState;

btbuild()

建立索引的入口函数为btbuild()函数，它的代码如下

IndexBuildResult *
btbuild(Relation heap, Relation index, IndexInfo *indexInfo)
{IndexBuildResult *result;BTBuildState buildstate;double        reltuples;#ifdef BTREE_BUILD_STATSif (log_btree_build_stats)ResetUsage();
#endif
//设置BTBuildState的初值buildstate.isunique = indexInfo->ii_Unique;buildstate.havedead = false;buildstate.heap = heap;buildstate.spool = NULL;buildstate.spool2 = NULL;buildstate.indtuples = 0;buildstate.btleader = NULL;/*希望精确调用索引关系一次*/if (RelationGetNumberOfBlocks(index) != 0)elog(ERROR, "index \"%s\" already contains data",RelationGetRelationName(index));reltuples = _bt_spools_heapscan(heap, index, &buildstate, indexInfo);
//表元组封装成索引元组_bt_leafbuild(buildstate.spool, buildstate.spool2);//spool文件的排序_bt_spooldestroy(buildstate.spool);//清除spool和它的子结构if (buildstate.spool2)_bt_spooldestroy(buildstate.spool2);//清除spool2if (buildstate.btleader)_bt_end_parallel(buildstate.btleader);//结束并行扫描result = (IndexBuildResult *) palloc(sizeof(IndexBuildResult));result->heap_tuples = reltuples;result->index_tuples = buildstate.indtuples;#ifdef BTREE_BUILD_STATSif (log_btree_build_stats){ShowUsage("BTREE BUILD STATS");ResetUsage();}
#endif                          /* BTREE_BUILD_STATS */return result;
}

_bt_leafbuild

btbuild会调用关于索引建立时生成的索引结构函数为_bt_leafbuild，会对索引元组进行排序

static void
_bt_leafbuild(BTSpool *btspool, BTSpool *btspool2)
{BTWriteState wstate;//建立BTWriteState的数据结构#ifdef BTREE_BUILD_STATSif (log_btree_build_stats){ShowUsage("BTREE BUILD (Spool) STATISTICS");ResetUsage();}
#endif                          /* BTREE_BUILD_STATS */pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,PROGRESS_BTREE_PHASE_PERFORMSORT_1);tuplesort_performsort(btspool->sortstate);//进行排序if (btspool2){pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,PROGRESS_BTREE_PHASE_PERFORMSORT_2);tuplesort_performsort(btspool2->sortstate);//如果存在唯一索引也会进行排序}wstate.heap = btspool->heap;//对于wstate赋值wstate.index = btspool->index;wstate.inskey = _bt_mkscankey(wstate.index, NULL);/*在WAL进行操作*/wstate.btws_use_wal = XLogIsNeeded() && RelationNeedsWAL(wstate.index);/* reserve the metapage */wstate.btws_pages_alloced = BTREE_METAPAGE + 1;//元页加一wstate.btws_pages_written = 0;//写入的页号为0wstate.btws_zeropage = NULL;    /* until needed */pgstat_progress_update_param(PROGRESS_CREATEIDX_SUBPHASE,PROGRESS_BTREE_PHASE_LEAF_LOAD);_bt_load(&wstate, btspool, btspool2);
}

_bt_load

_bt_load函数是在 _bt_leafbuild中被调用，将排好序的spool依次取出。源代码如下：

static void
_bt_load(BTWriteState *wstate, BTSpool *btspool, BTSpool *btspool2)
{BTPageState *state = NULL;bool        merge = (btspool2 != NULL);IndexTuple itup,itup2 = NULL;bool     load1;TupleDesc tupdes = RelationGetDescr(wstate->index);int            i,keysz = IndexRelationGetNumberOfKeyAttributes(wstate->index);SortSupport sortKeys;int64       tuples_done = 0;if (merge)//如果spool2元组为非空{/*需要将spool和spool2进行归并，按照偏序关系依次取出索引元组*//* 归并前准备 */itup = tuplesort_getindextuple(btspool->sortstate, true);itup2 = tuplesort_getindextuple(btspool2->sortstate, true);/* 为每一列准备SortSupport data  */sortKeys = (SortSupport) palloc0(keysz * sizeof(SortSupportData));for (i = 0; i < keysz; i++){SortSupport sortKey = sortKeys + i;ScanKey       scanKey = wstate->inskey->scankeys + i;int16        strategy;sortKey->ssup_cxt = CurrentMemoryContext;sortKey->ssup_collation = scanKey->sk_collation;sortKey->ssup_nulls_first =(scanKey->sk_flags & SK_BT_NULLS_FIRST) != 0;sortKey->ssup_attno = scanKey->sk_attno;/*不支持缩写 */sortKey->abbreviate = false;AssertState(sortKey->ssup_attno != 0);strategy = (scanKey->sk_flags & SK_BT_DESC) != 0 ?BTGreaterStrategyNumber : BTLessStrategyNumber;PrepareSortSupportFromIndexRel(wstate->index, strategy, sortKey);}//按照偏序关系决定次序，load1为true时为spool中元组，为false为spool2元组//判断顺序后调用bt_buildadd函数for (;;){load1 = true;     /* load BTSpool next ? */if (itup2 == NULL){if (itup == NULL)break;}else if (itup != NULL){int32       compare = 0;for (i = 1; i <= keysz; i++){SortSupport entry;Datum        attrDatum1,attrDatum2;bool      isNull1,isNull2;entry = sortKeys + i - 1;attrDatum1 = index_getattr(itup, i, tupdes, &isNull1);attrDatum2 = index_getattr(itup2, i, tupdes, &isNull2);compare = ApplySortComparator(attrDatum1, isNull1,attrDatum2, isNull2,entry);if (compare > 0){load1 = false;break;}else if (compare < 0)break;}/*如果compare值相等，需要比较ItemPointer*/if (compare == 0){compare = ItemPointerCompare(&itup->t_tid, &itup2->t_tid);Assert(compare != 0);if (compare > 0)load1 = false;}}elseload1 = false;/* When we see first tuple, create first index page */if (state == NULL)state = _bt_pagestate(wstate, 0);if (load1){_bt_buildadd(wstate, state, itup);itup = tuplesort_getindextuple(btspool->sortstate, true);}else{_bt_buildadd(wstate, state, itup2);itup2 = tuplesort_getindextuple(btspool2->sortstate, true);}/* Report progress */pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE,++tuples_done);}pfree(sortKeys);}else    //如果spool2不存在，直接依次读取spool元组调用bt_buildadd函数{while ((itup = tuplesort_getindextuple(btspool->sortstate,true)) != NULL){if (state == NULL)state = _bt_pagestate(wstate, 0);_bt_buildadd(wstate, state, itup);pgstat_progress_update_param(PROGRESS_CREATEIDX_TUPLES_DONE,++tuples_done);}}

bt_buildadd

_bt_load函数会调用bt_buildadd函数，由于bt_buildadd函数代码很长，在这里大概讲一下该函数的原理，bt_buildadd函数会在页面上直接进行插入，而如果页面不够时会申请新的页面作为它的右兄弟节点，并设置其highkey。并同时查找相应的父节点，如果没有父节点则会新建一个，并赋予最小值，实现了递归调用。该函数的流程图如下

总结

以上便是B-Tree索引的构建的基本流程，由bt_build函数进入，调用bt_leafbuild函数对于spool元组排序，它在调用bt_load函数将排好序的spool元组取出，最后由bt_buildadd函数进行索引结构的实现。可见pg中的B-Tree索引实现层层递进，结构清晰，安全性很高。

postgreSQL源码分析——索引的建立与使用——B-Tree索引(2)相关推荐

PostgreSQL源码分析
PostgreSQL源码结构 PostgreSQL的使用形态 PostgreSQL采用C/S(客户机/服务器)模式结构.应用层通过INET或者Unix Socket利用既定的协议与数据库服务器进行通信 ...
postgreSQL源码分析——索引的建立与使用——GIST索引（2）
2021SC@SDUSC 本篇博客主要讲解GiST索引创建以及删除的相关函数这里写目录标题 GIST创建相关数据结构 GISTBuildState GISTInsertStack gistbuil ...
postgreSQL源码分析——索引的建立与使用——Hash索引(3)
2021SC@SDUSC 上一篇博客讲了关于Hash索引创建与插入的相关函数,这一篇博客讲述关于溢出页的操作函数以及Hash表的扩展相关的函数. 目录溢出页的分配和回收 _hash_addovflp ...
postgreSQL源码分析——索引的建立与使用——Hash索引(2)
2021SC@SDUSC 目录 Hash索引创建 hashbuild函数 _hash_init函数 Hash索引的插入 hashinsert函数 _hash_doinsert函数总结 Hash索引创 ...
postgreSQL源码分析——索引的建立与使用——各种索引类型的管理和操作（2）
2021SC@SDUSC 目录上层操作函数 index_open index_beginscan() index_create() indexcmd.c 下层接口函数 IndexScanDescDa ...
postgreSQL源码分析——索引的建立与使用——Hash索引(1)
2021SC@SDUSC 目录 Hash索引 Hash索引原理 Hash表 Hash索引结构 Hash的页面结构元页桶页,溢出页,位图页和B-Tree相比的优缺点优点缺点总结 Hash索引 ...
postgreSQL源码分析——索引的建立与使用——各种索引类型的管理和操作（1）
2021SC@SDUSC 目录概述管理索引的系统表记录索引相关的系统表与索引系统表相关的后端源码索引的操作函数上层操作函数下层接口函数概述索引是指按表中某些关键属性或表达式建立元组的 ...
postgreSQL源码分析——索引的建立与使用——总结篇
2021SC@SDUSC 在小组中我负责索引的建立与使用的相关部分,在此一共写了16篇相关的分析报告,着重分析各种索引的操作和管理方法,以及分析了PG中四种最重要的索引B-Tree索引,Hash索引, ...
postgreSQL源码分析——索引的建立与使用——B-Tree索引(3)
2021SC@SDUSC 目录 B-Tree的插入 bt_insert _bt_doinsert BTInsertStateData _bt_search函数 _bt_moveright函数 B-Tr ...

postgreSQL源码分析——索引的建立与使用——B-Tree索引(2)

目录