PostgreSQL在何处处理 sql查询之五十九

本文涉及的产品
云原生数据库 PolarDB MySQL 版,Serverless 5000PCU 100GB
简介:

由于用单纯的SQL语句来探查代码,看得还是不够清楚。

所以我再采用如下的方法:

复制代码
postgres=# explain select dept.no_emps,emp.age from dept,emp where emp.name = dept.mgr and dept.dept_name = 'shoe';
                            QUERY PLAN                            
------------------------------------------------------------------
 Hash Join  (cost=19.30..45.07 rows=23 width=8)
   Hash Cond: ((emp.name)::text = (dept.mgr)::text)
   ->  Seq Scan on emp  (cost=0.00..21.30 rows=1130 width=42)
   ->  Hash  (cost=19.25..19.25 rows=4 width=42)
         ->  Seq Scan on dept  (cost=0.00..19.25 rows=4 width=42)
               Filter: ((dept_name)::text = 'shoe'::text)
(6 rows)

postgres=# 
复制代码

通过对代码的跟踪,可以看到 ExecInitNode被执行了四次。

每次都运行时的 NodeTag依次是:

124--T_HashJoin

109--T_SeqScan

131--T_Hash

109--T_SeqScan

正好和用 explain看到的顺序相同。

下面要进一步看代码中相关的结构如何变化。

看这段源代码:

复制代码
/* ----------------------------------------------------------------
 *        ExecInitHashJoin
 *
 *        Init routine for HashJoin node.
 * ----------------------------------------------------------------
 */
HashJoinState *
ExecInitHashJoin(HashJoin *node, EState *estate, int eflags)
{
    HashJoinState *hjstate;
    Plan       *outerNode;
    Hash       *hashNode;
    List       *lclauses;
    List       *rclauses;
    List       *hoperators;
    ListCell   *l;

    /* check for unsupported flags */
    Assert(!(eflags & (EXEC_FLAG_BACKWARD | EXEC_FLAG_MARK)));

    /*
     * create state structure
     */
    hjstate = makeNode(HashJoinState);
    hjstate->js.ps.plan = (Plan *) node;
    hjstate->js.ps.state = estate;

    /*
     * Miscellaneous initialization
     *
     * create expression context for node
     */
    ExecAssignExprContext(estate, &hjstate->js.ps);

    /*
     * initialize child expressions
     */
    hjstate->js.ps.targetlist = (List *)
        ExecInitExpr((Expr *) node->join.plan.targetlist,
                     (PlanState *) hjstate);
    hjstate->js.ps.qual = (List *)
        ExecInitExpr((Expr *) node->join.plan.qual,
                     (PlanState *) hjstate);
    hjstate->js.jointype = node->join.jointype;
    hjstate->js.joinqual = (List *)
        ExecInitExpr((Expr *) node->join.joinqual,
                     (PlanState *) hjstate);
    hjstate->hashclauses = (List *)
        ExecInitExpr((Expr *) node->hashclauses,
                     (PlanState *) hjstate);

    /*
     * initialize child nodes
     *
     * Note: we could suppress the REWIND flag for the inner input, which
     * would amount to betting that the hash will be a single batch.  Not
     * clear if this would be a win or not.
     */
    outerNode = outerPlan(node);
    hashNode = (Hash *) innerPlan(node);

    outerPlanState(hjstate) = ExecInitNode(outerNode, estate, eflags);
    innerPlanState(hjstate) = ExecInitNode((Plan *) hashNode, estate, eflags);

    /*
     * tuple table initialization
     */
    ExecInitResultTupleSlot(estate, &hjstate->js.ps);
    hjstate->hj_OuterTupleSlot = ExecInitExtraTupleSlot(estate);

    /* set up null tuples for outer joins, if needed */
    switch (node->join.jointype)
    {
        case JOIN_INNER:
        case JOIN_SEMI:
            break;
        case JOIN_LEFT:
        case JOIN_ANTI:
            hjstate->hj_NullInnerTupleSlot =
                ExecInitNullTupleSlot(estate,
                                 ExecGetResultType(innerPlanState(hjstate)));
            break;
        case JOIN_RIGHT:
            hjstate->hj_NullOuterTupleSlot =
                ExecInitNullTupleSlot(estate,
                                 ExecGetResultType(outerPlanState(hjstate)));
            break;
        case JOIN_FULL:
            hjstate->hj_NullOuterTupleSlot =
                ExecInitNullTupleSlot(estate,
                                 ExecGetResultType(outerPlanState(hjstate)));
            hjstate->hj_NullInnerTupleSlot =
                ExecInitNullTupleSlot(estate,
                                 ExecGetResultType(innerPlanState(hjstate)));
            break;
        default:
            elog(ERROR, "unrecognized join type: %d",
                 (int) node->join.jointype);
    }

    /*
     * now for some voodoo.  our temporary tuple slot is actually the result
     * tuple slot of the Hash node (which is our inner plan).  we can do this
     * because Hash nodes don't return tuples via ExecProcNode() -- instead
     * the hash join node uses ExecScanHashBucket() to get at the contents of
     * the hash table.    -cim 6/9/91
     */
    {
        HashState  *hashstate = (HashState *) innerPlanState(hjstate);
        TupleTableSlot *slot = hashstate->ps.ps_ResultTupleSlot;

        hjstate->hj_HashTupleSlot = slot;
    }

    /*
     * initialize tuple type and projection info
     */
    ExecAssignResultTypeFromTL(&hjstate->js.ps);
    ExecAssignProjectionInfo(&hjstate->js.ps, NULL);

    ExecSetSlotDescriptor(hjstate->hj_OuterTupleSlot,
                          ExecGetResultType(outerPlanState(hjstate)));

    /*
     * initialize hash-specific info
     */
    hjstate->hj_HashTable = NULL;
    hjstate->hj_FirstOuterTupleSlot = NULL;

    hjstate->hj_CurHashValue = 0;
    hjstate->hj_CurBucketNo = 0;
    hjstate->hj_CurSkewBucketNo = INVALID_SKEW_BUCKET_NO;
    hjstate->hj_CurTuple = NULL;

    /*
     * Deconstruct the hash clauses into outer and inner argument values, so
     * that we can evaluate those subexpressions separately.  Also make a list
     * of the hash operator OIDs, in preparation for looking up the hash
     * functions to use.
     */
    lclauses = NIL;
    rclauses = NIL;
    hoperators = NIL;
    foreach(l, hjstate->hashclauses)
    {
        FuncExprState *fstate = (FuncExprState *) lfirst(l);
        OpExpr       *hclause;

        Assert(IsA(fstate, FuncExprState));
        hclause = (OpExpr *) fstate->xprstate.expr;
        Assert(IsA(hclause, OpExpr));
        lclauses = lappend(lclauses, linitial(fstate->args));
        rclauses = lappend(rclauses, lsecond(fstate->args));
        hoperators = lappend_oid(hoperators, hclause->opno);
    }
    hjstate->hj_OuterHashKeys = lclauses;
    hjstate->hj_InnerHashKeys = rclauses;
    hjstate->hj_HashOperators = hoperators;
    /* child Hash node needs to evaluate inner hash keys, too */
    ((HashState *) innerPlanState(hjstate))->hashkeys = rclauses;

    hjstate->js.ps.ps_TupFromTlist = false;
    hjstate->hj_JoinState = HJ_BUILD_HASHTABLE;
    hjstate->hj_MatchedOuter = false;
    hjstate->hj_OuterNotEmpty = false;

    return hjstate;
}
复制代码

将之简化:

复制代码
/* ----------------------------------------------------------------
 *        ExecInitHashJoin
 *
 *        Init routine for HashJoin node.
 * ----------------------------------------------------------------
 */
HashJoinState *
ExecInitHashJoin(HashJoin *node, EState *estate, int eflags)
{
    HashJoinState *hjstate;
    Plan       *outerNode;
    Hash       *hashNode;
    List       *lclauses;
    List       *rclauses;
    List       *hoperators;
    ListCell   *l;

    ...

    /*
     * initialize child nodes
     *
     * Note: we could suppress the REWIND flag for the inner input, which
     * would amount to betting that the hash will be a single batch.  Not
     * clear if this would be a win or not.
     */
    outerNode = outerPlan(node);
    hashNode = (Hash *) innerPlan(node);

    outerPlanState(hjstate) = ExecInitNode(outerNode, estate, eflags);
    innerPlanState(hjstate) = ExecInitNode((Plan *) hashNode, estate, eflags);

    ...

    return hjstate;
}
复制代码

可以看到其实 

outerNode = outerPlan(node) 就是: outerNode = (((Plan *)(node))->lefttree)
hashNode = (Hash *) innerPlan(node) 就是:hashNode = (((Plan *)(node))->righttree)

outerPlanState(hjstate) 就是 (((PlanState *)(hjstate))->lefttree)

innerPlanState(hjstate就是 (((PlanState *)(hjstate))->righttree)

或者说,在对 Hash 节点进行处理的时候,要分别处理左节点和右节点。

由于计划树结构比较复杂,借鉴explain.c中的代码来观察如何读取其中的数据,是一个可行的办法。






本文转自健哥的数据花园博客园博客,原文链接:http://www.cnblogs.com/gaojian/archive/2013/06/11/3131778.html,如需转载请自行联系原作者

相关实践学习
使用PolarDB和ECS搭建门户网站
本场景主要介绍基于PolarDB和ECS实现搭建门户网站。
阿里云数据库产品家族及特性
阿里云智能数据库产品团队一直致力于不断健全产品体系,提升产品性能,打磨产品功能,从而帮助客户实现更加极致的弹性能力、具备更强的扩展能力、并利用云设施进一步降低企业成本。以云原生+分布式为核心技术抓手,打造以自研的在线事务型(OLTP)数据库Polar DB和在线分析型(OLAP)数据库Analytic DB为代表的新一代企业级云原生数据库产品体系, 结合NoSQL数据库、数据库生态工具、云原生智能化数据库管控平台,为阿里巴巴经济体以及各个行业的企业客户和开发者提供从公共云到混合云再到私有云的完整解决方案,提供基于云基础设施进行数据从处理、到存储、再到计算与分析的一体化解决方案。本节课带你了解阿里云数据库产品家族及特性。
目录
相关文章
|
16天前
|
SQL
sql语句加正则 简化查询
sql语句加正则 简化查询
14 0
sql语句加正则 简化查询
|
1月前
|
SQL
sql server链接查询
sql server链接查询
17 1
|
1月前
|
SQL
sql server简单查询
sql server简单查询
14 1
|
1月前
|
关系型数据库 分布式数据库 数据库
PolarDB常见问题之加了索引但是查询没有使用如何解决
PolarDB是阿里云推出的下一代关系型数据库,具有高性能、高可用性和弹性伸缩能力,适用于大规模数据处理场景。本汇总囊括了PolarDB使用中用户可能遭遇的一系列常见问题及解答,旨在为数据库管理员和开发者提供全面的问题指导,确保数据库平稳运行和优化使用体验。
|
24天前
|
SQL 关系型数据库 MySQL
mysql一条sql查询出多个统计结果
mysql一条sql查询出多个统计结果
14 0
|
1月前
|
SQL
sql高级查询
sql高级查询
12 0
|
1天前
|
SQL Java 数据库连接
Java从入门到精通:2.3.2数据库编程——了解SQL语言,编写基本查询语句
Java从入门到精通:2.3.2数据库编程——了解SQL语言,编写基本查询语句
|
8天前
|
SQL 存储 Oracle
关系型数据库查询数据的语句
本文介绍了关系型数据库中的基本SQL查询语句,包括选择所有或特定列、带条件查询、排序、分组、过滤分组、表连接、限制记录数及子查询。SQL还支持窗口函数、存储过程等高级功能,是高效管理数据库的关键。建议深入学习SQL及相应数据库系统文档。
9 2
|
1月前
|
SQL 数据库
sql server高级查询,看这篇文章就够了
sql server高级查询,看这篇文章就够了
24 0
|
1月前
|
SQL
T-SQL 语句查询
T-SQL 语句查询
55 0