这篇文章主要为大家展示了“PostgreSQL中如何生成子链接执行计划、使用Param替换上层Vars以及转换表达式为隐式AND格式”，内容简而易懂，条理清晰，希望能够帮助大家解决疑惑，下面让小编带领大家一起研究并学习一下“PostgreSQL中如何生成子链接执行计划、使用Param替换上层Vars以及转换表达式为隐式AND格式”这篇文章吧。
一、主函数
主函数preprocess_expression先前章节也介绍过，在此函数中调用了生成子链接执行计划、使用Param替换上层Vars以及转换表达式为隐式AND格式（implicit-AND format）等相关子函数。
preprocess_expression
static Node *
preprocess_expression(PlannerInfo *root, Node *expr, int kind)
{
//...
if (root->parse->hasSubLinks)//扩展子链接为子计划
expr = SS_process_sublinks(root, expr, (kind == EXPRKIND_QUAL));
if (root->query_level > 1)
expr = SS_replace_correlation_vars(root, expr);//使用Param节点替换上层的Vars
if (kind == EXPRKIND_QUAL)//转换为隐式AND格式
expr = (Node *) make_ands_implicit((Expr *) expr);
return expr;
}二、生成子链接执行计划
先前的章节已介绍了上拉子链接的相关处理过程,对于不能上拉的子链接,PG会生成子执行计划.对于会生成常量的子链接,则会把生成的常量记录在Param中,在需要的时候由父查询使用.
例1:以下的子链接,PG会生成子计划,并把子链接的结果物化(Materialize)提升整体性能.
testdb=# explain verbose select * from t_dwxx where dwbh > all (select b.dwbh from t_grxx b);
QUERY PLAN
---------------------------------------------------------------------------------
Seq Scan on public.t_dwxx (cost=0.00..1498.00 rows=80 width=474)
Output: t_dwxx.dwmc, t_dwxx.dwbh, t_dwxx.dwdz
Filter: (SubPlan 1)
SubPlan 1
-> Materialize (cost=0.00..17.35 rows=490 width=38)
Output: b.dwbh
-> Seq Scan on public.t_grxx b (cost=0.00..14.90 rows=490 width=38)
Output: b.dwbh
(8 rows)例2:以下的子链接,PG会把生成的常量记录在Param中(注意生成的参数:$0)
testdb=# explain verbose select * from t_dwxx a where exists (select max(b.dwbh) from t_grxx b);
QUERY PLAN
---------------------------------------------------------------------------------
Result (cost=16.14..27.73 rows=160 width=474)
Output: a.dwmc, a.dwbh, a.dwdz
One-Time Filter: $0
InitPlan 1 (returns $0)
-> Aggregate (cost=16.12..16.14 rows=1 width=32)
Output: max((b.dwbh)::text)
-> Seq Scan on public.t_grxx b (cost=0.00..14.90 rows=490 width=38)
Output: b.dwbh, b.grbh, b.xm, b.nl
-> Seq Scan on public.t_dwxx a (cost=16.14..27.73 rows=160 width=474)
Output: a.dwmc, a.dwbh, a.dwdz
(10 rows)源代码如下:
SS_process_sublinks
Node *
SS_process_sublinks(PlannerInfo *root, Node *expr, bool isQual)
{
process_sublinks_context context;
context.root = root;
context.isTopQual = isQual;
return process_sublinks_mutator(expr, &context);//调用XX_mutator函数遍历并处理
}
static Node *
process_sublinks_mutator(Node *node, process_sublinks_context *context)
{
process_sublinks_context locContext;
locContext.root = context->root;
if (node == NULL)
return NULL;
if (IsA(node, SubLink))//子链接
{
SubLink *sublink = (SubLink *) node;
Node *testexpr;
locContext.isTopQual = false;
testexpr = process_sublinks_mutator(sublink->testexpr, &locContext);
return make_subplan(context->root,
(Query *) sublink->subselect,
sublink->subLinkType,
sublink->subLinkId,
testexpr,
context->isTopQual);//生成子执行计划,与整体的执行计划类似
}
if (IsA(node, PlaceHolderVar))
{
if (((PlaceHolderVar *) node)->phlevelsup > 0)
return node;
}
else if (IsA(node, Aggref))
{
if (((Aggref *) node)->agglevelsup > 0)
return node;
}
Assert(!IsA(node, SubPlan));
Assert(!IsA(node, AlternativeSubPlan));
Assert(!IsA(node, Query));
if (and_clause(node))//AND语句
{
List *newargs = NIL;
ListCell *l;
locContext.isTopQual = context->isTopQual;
foreach(l, ((BoolExpr *) node)->args)
{
Node *newarg;
newarg = process_sublinks_mutator(lfirst(l), &locContext);
if (and_clause(newarg))
newargs = list_concat(newargs, ((BoolExpr *) newarg)->args);
else
newargs = lappend(newargs, newarg);
}
return (Node *) make_andclause(newargs);
}
if (or_clause(node))//OR语句
{
List *newargs = NIL;
ListCell *l;
locContext.isTopQual = context->isTopQual;
foreach(l, ((BoolExpr *) node)->args)
{
Node *newarg;
newarg = process_sublinks_mutator(lfirst(l), &locContext);
if (or_clause(newarg))
newargs = list_concat(newargs, ((BoolExpr *) newarg)->args);
else
newargs = lappend(newargs, newarg);
}
return (Node *) make_orclause(newargs);
}
locContext.isTopQual = false;
return expression_tree_mutator(node,
process_sublinks_mutator,
(void *) &locContext);
}三、使用Param替换上层变量
SQL例子参考"二、生成子链接执行计划"中的例2,这也是使用Param替代Var的一个例子.
源代码如下:
Node *
SS_replace_correlation_vars(PlannerInfo *root, Node *expr)
{
//调用XX_mutator遍历处理
return replace_correlation_vars_mutator(expr, root);
}
static Node *
replace_correlation_vars_mutator(Node *node, PlannerInfo *root)
{
if (node == NULL)
return NULL;
if (IsA(node, Var))//Var
{
if (((Var *) node)->varlevelsup > 0)
return (Node *) replace_outer_var(root, (Var *) node);//使用Param替换
}
if (IsA(node, PlaceHolderVar))
{
if (((PlaceHolderVar *) node)->phlevelsup > 0)
return (Node *) replace_outer_placeholdervar(root,
(PlaceHolderVar *) node);
}
if (IsA(node, Aggref))
{
if (((Aggref *) node)->agglevelsup > 0)
return (Node *) replace_outer_agg(root, (Aggref *) node);
}
if (IsA(node, GroupingFunc))
{
if (((GroupingFunc *) node)->agglevelsup > 0)
return (Node *) replace_outer_grouping(root, (GroupingFunc *) node);
}
return expression_tree_mutator(node,
replace_correlation_vars_mutator,
(void *) root);
}
static Param *
replace_outer_var(PlannerInfo *root, Var *var)//构造Param替换Var
{
Param *retval;
int i;
Assert(var->varlevelsup > 0 && var->varlevelsup < root->query_level);
i = assign_param_for_var(root, var);
retval = makeNode(Param);
retval->paramkind = PARAM_EXEC;
retval->paramid = i;
retval->paramtype = var->vartype;
retval->paramtypmod = var->vartypmod;
retval->paramcollid = var->varcollid;
retval->location = var->location;
return retval;
}四、转换表达式为隐式AND格式
源码如下:
List *
make_ands_implicit(Expr *clause)
{
if (clause == NULL)//如为NULL,返回空指针
return NIL;
else if (and_clause((Node *) clause))//AND语句,直接返回AND中的args参数
return ((BoolExpr *) clause)->args;
else if (IsA(clause, Const) &&
!((Const *) clause)->constisnull &&
DatumGetBool(((Const *) clause)->constvalue))
return NIL;
else
return list_make1(clause);//返回List
}以上是“PostgreSQL中如何生成子链接执行计划、使用Param替换上层Vars以及转换表达式为隐式AND格式”这篇文章的所有内容，感谢各位的阅读！相信大家都有了一定的了解，希望分享的内容对大家有所帮助，如果还想学习更多知识，欢迎关注亿速云行业资讯频道！
