删除具有两个或多个左连接表的重复行

Yan*_*hon 8 postgresql join postgresql-9.4

像这样的查询

SELECT a.id, a.name,
       COALESCE( json_agg(b.*), '[]'::json ),
       COALESCE( json_agg(c.*), '[]'::json ),
  FROM a
  LEFT JOIN b ON a.id = b.a_id
  LEFT JOIN c ON a.id = c.a_id
 GROUP BY a.id, a.name;
Run Code Online (Sandbox Code Playgroud)

执行时,cb将彼此相乘并在 JSON 数组对象中产生重复的条目。

我尝试将查询更改为使用 2 个子查询,但出现各种错误和警告,例如“子查询必须只返回一列”等。

我也尝试使用LEFT OUTER JOIN,但我想我还没有掌握连接表的工作原理,因为它仅适用于b并且c仍然相乘并包含重复项。

编辑DISTINCTCOALESCE函数上使用带有“无法识别 json 类型的相等运算符”的错误。

如何修复此查询并仅聚合唯一行?


编辑 2

我需要指定表bc实际上都是VIEWs,而且它们都至少有一json_agg列,所以我不能只使用json_agg(DISTINCT b.*). 这太容易了。


编辑 3

这是一个重现问题的小片段:

--DROP TABLE IF EXISTS tbl_a CASCADE;
--DROP TABLE IF EXISTS tbl_b CASCADE;
--DROP TABLE IF EXISTS tbl_c CASCADE;

CREATE TABLE tbl_a (
  id bigserial NOT NULL,
  name character varying(16),
  CONSTRAINT "PK_tbl_a" PRIMARY KEY (id)
) WITH ( OIDS=FALSE );

CREATE TABLE tbl_b (
  a_id bigint NOT NULL,
  foo json NOT NULL DEFAULT '{}'::json,
  CONSTRAINT "FK_tbl_b_a" FOREIGN KEY (a_id)
      REFERENCES tbl_a (id) MATCH SIMPLE
      ON UPDATE CASCADE ON DELETE CASCADE
) WITH ( OIDS=FALSE );

CREATE TABLE tbl_c (
  a_id bigint NOT NULL,
  bar json NOT NULL DEFAULT '{}'::json,
  CONSTRAINT "FK_tbl_c_a" FOREIGN KEY (a_id)
      REFERENCES tbl_a (id) MATCH SIMPLE
      ON UPDATE CASCADE ON DELETE CASCADE
) WITH ( OIDS=FALSE );

INSERT INTO tbl_a (id,name) VALUES (1, 'Test');
INSERT INTO tbl_b (a_id, foo) VALUES (1, '{"foo":"Hello"}'::json), (1, '{"foo":"World"}'::json);
INSERT INTO tbl_c (a_id, bar) VALUES (1, '{"bar":"abc"}'::json), (1, '{"bar":"def"}'::json);

SELECT tbl_a.id, tbl_a.name,
       COALESCE(json_agg(tbl_b.*), '{}'::json),
       COALESCE(json_agg(tbl_c.*), '{}'::json)
  FROM tbl_a
  LEFT JOIN tbl_b ON tbl_a.id = tbl_b.a_id
  LEFT JOIN tbl_c ON tbl_a.id = tbl_c.a_id
 GROUP BY tbl_a.id, tbl_a.name;
Run Code Online (Sandbox Code Playgroud)

退货

id  name    coalesce                              coalesce
--  ------  ------------------------------------  ----------------------
 1  "Test"  "[{"a_id":1,"foo":{"foo":"World"}},   "[{"a_id":1,"bar":{"bar":"abc"}},
             {"a_id":1,"foo":{"foo":"Hello"}},    {"a_id":1,"bar":{"bar":"abc"}}, 
             {"a_id":1,"foo":{"foo":"World"}},    {"a_id":1,"bar":{"bar":"def"}}, 
             {"a_id":1,"foo":{"foo":"Hello"}}]"   {"a_id":1,"bar":{"bar":"def"}}]"
Run Code Online (Sandbox Code Playgroud)

Yan*_*hon 6

我找到了解决办法。我不确定它是否是最佳的,但它有效。

SELECT tbl_a.id, tbl_a.name,
       COALESCE( ( SELECT json_agg(tbl_b.*)
                     FROM tbl_b 
                    WHERE tbl_b.a_id = tbl_a.id ), '{}'::json),
       COALESCE( ( SELECT json_agg(tbl_c.*)
                     FROM tbl_c 
                    WHERE tbl_c.a_id = tbl_a.id ), '{}'::json)
  FROM tbl_a;
Run Code Online (Sandbox Code Playgroud)

哪个正确返回

id  name    coalesce                             coalesce
--  ------  -----------------------------------  ----------------------
 1  "Test"  "[{"a_id":1,"foo":{"foo":"World"}},  "[{"a_id":1,"bar":{"bar":"abc"}},
              {"a_id":1,"foo":{"foo":"Hello"}},    {"a_id":1,"bar":{"bar":"def"}}]"
Run Code Online (Sandbox Code Playgroud)