2

我有以下表格:

CREATE TABLE element (
  element_id serial PRIMARY KEY,
  local_id integer,
  name varchar,
  CONSTRAINT fk_element_local_id FOREIGN KEY (local_id)
      REFERENCES local (local_id) MATCH SIMPLE
      ON UPDATE NO ACTION ON DELETE NO ACTION
);

CREATE TABLE local (
  local_id serial PRIMARY KEY,
  parent_id integer,
  name varchar,
  CONSTRAINT fk_local_parent_id_local_id FOREIGN KEY (parent_id)
      REFERENCES local (local_id) MATCH SIMPLE
      ON UPDATE CASCADE ON DELETE SET NULL
);

CREATE TABLE category (
  category_id serial PRIMARY KEY,
  name varchar
);

CREATE TABLE action (
  action_id serial PRIMARY KEY,
  local_id integer,
  category_id integer,
  CONSTRAINT fk_action_local_id FOREIGN KEY (local_id)
      REFERENCES local (local_id) MATCH SIMPLE
      ON UPDATE NO ACTION ON DELETE NO ACTION,
  CONSTRAINT fk_action_element_id FOREIGN KEY (element_id)
      REFERENCES element (element_id) MATCH SIMPLE
      ON UPDATE NO ACTION ON DELETE NO ACTION
);

我想从一个动作中选择所有元素。如果元素的本地是动作的本地的后代,它也应该出现。
例子:

local

|local_id | parent_id | name |
|---------+-----------+------|
|1        |NULL       |A     |
|2        |1          |B     |
|3        |1          |C     |
|4        |3          |D     |
|5        |NULL       |E     |
|6        |5          |F     |
|_________|___________|______|

category

| category_id | name |
|-------------+------|
|1            |A     |
|2            |B     |
|2            |C     |
|_____________|______|

element

|element_id | local_id | name | category_id |
|-----------+----------+------+-------------|
|1          |1         |A     | 1           |
|2          |2         |B     | 2           |
|3          |2         |C     | 1           |
|4          |4         |D     | 2           |
|5          |5         |E     | 2           |
|6          |6         |F     | 1           |
|7          |6         |G     | 1           |
|___________|__________|______|_____________|

action

|action_id | local_id | category_id |
|----------+----------+-------------|
| 1        | 1        | 2           |
| 2        | 3        | 1           |
| 3        | 5        | 1           |
| 4        | 6        | 1           |
|__________|__________|_____________|

我想要的查询结果:

CASE: action_id = 1
return: element_id: 2,4

CASE: action_id = 2
return: element_id: null

CASE: action_id = 3
return: element_id: 6,7

我已经创建了一个函数,它返回包括实际节点在内的所有后代,但由于调用函数数千次时的性能,我遇到了困难。我的功能如下所示:

CREATE OR REPLACE FUNCTION fn_local_get_childs(_parent_id integer)
  RETURNS SETOF integer AS
$BODY$
DECLARE
   r integer;
BEGIN
   FOR r IN SELECT local_id FROM local WHERE local_id IN ( 
      (WITH RECURSIVE parent AS
      (
         SELECT local_id , parent_id  from local WHERE local_id = _parent_id
         UNION ALL 
         SELECT t.local_id , t.parent_id FROM parent
         INNER JOIN local t ON parent.local_id =  t.parent_id
      )
      SELECT local_id FROM  parent
      ) 
   )
   LOOP
      RETURN NEXT r;
   END LOOP;
   RETURN;        
END;
$BODY$
  LANGUAGE plpgsql VOLATILE
  COST 100
  ROWS 1000;

我的超慢查询如下所示:

select e.element_id, a.action_id
from action a
join element e on (
                   e.local_id=any(select fn_local_get_childs(a.local_id)) AND 
                   e.category_id=a.category_id)

有没有办法在单个查询中组合函数中使用的递归?

4

1 回答 1

1

集成查询

在多处改进逻辑,您可以将整个操作集成到一个查询中。包装成 SQL 函数是可选的:

CREATE OR REPLACE FUNCTION f_elems(_action_id integer)
  RETURNS SETOF integer AS
$func$
   WITH RECURSIVE l AS (
      SELECT a.category_id, l.local_id
      FROM   action a
      JOIN   local  l USING (local_id)
      WHERE  a.action_id = $1

      UNION ALL 
      SELECT l.category_id, c.local_id
      FROM   l
      JOIN   local c ON c.parent_id = l.local_id  -- c for "child"
      )
   SELECT e.element_id
   FROM   l
   JOIN   element e USING (category_id, local_id);
$func$  LANGUAGE sql STABLE;

检索element_id给定的相同和子本地人的所有内容action_id

称呼:

SELECT * FROM f_elem(3);

element_id
-----------
6
7

db<>fiddle here
sqlfiddle

由于几个原因,这应该已经大大加快了。最明显的是:

  • 用纯 SQL 代替 plpgsql 中的慢循环。
  • 缩小递归查询的起始集。
  • 删除不必要的和臭名昭著的缓慢IN构造。

我调用 withSELECT * FROM ...而不是 just SELECT,即使该行只有一列,以获取我在函数标题中声明的OUT参数 ( ) 的列名。element_id

更快,但

指数

索引action.action_id由主键提供。

但是您可能错过了local.parent_id. 在此期间,将其作为第一个元素和第二个元素的覆盖多列索引(Postgres 9.2+)。如果桌子很大,这应该会有很大帮助。对于一张小桌子来说没有那么多或根本没有:parent_idlocal_idlocal

CREATE INDEX l_mult_idx ON local(parent_id, local_id);

为什么?看:

最后,表上的多列索引element应该有更多帮助:

CREATE INDEX e_mult_idx ON element (category_id, local_id, element_id);

第三列element_id仅用于使其成为覆盖索引。如果您的查询从表中检索到更多列element,您可能希望将更多列添加到索引或删除element_id。要么会让它更快。

物化视图

如果您的表收到很少或没有更新,则提供共享同一类别的所有对的预先计算集的物化视图(action_id, element_id)将使这种闪电变得快速。使(action_id, element_id)(按此顺序)主键。

于 2013-05-25T22:31:39.667 回答