Postgres | 插入 cte 后执行 delete cte

Question

我有一张桌子：

CREATE TABLE product
(
    id      bigint GENERATED ALWAYS AS IDENTITY PRIMARY KEY,
    from_id bigint NOT NULL,
    to_id   bigint NOT NULL,
    comments text NOT NULL,
    data    jsonb  NOT NULL
);

CREATE UNIQUE INDEX product_unique_idx ON product(from_id, to_id, comments);

使用以下数据：

insert into product(from_id, to_id, comments, data) values
(1, 2, 'bla', '{}'),
(2, 3, 'bla', '{}'),
(1, 3, 'bla', '{}'),
(3, 2, 'bla', '{}'),
(2, 1, 'bla', '{}'),
(3, 1, 'bla', '{}');

现在，我想插入将使用给定集合更新 from_id 和 to_id 的新记录，例如，将所有 from_id 和 to_ids [1,2] 替换为 3（也删除 where from_id == to_id），因为我使用了UNIQUE INDEX它通过insert然后delete所有其他行：

with
   insert_stmt_to_id AS (
      insert into product
         (from_id, to_id, comments, data)
         (select from_id,3,comments,data from product
           where to_id in (1,2))
         ON CONFLICT (from_id, to_id, comments) DO NOTHING),
   insert_stmt_from_id AS (
      insert into product
         (from_id, to_id, comments, data)
         (select 3,to_id,comments,data from product
           where from_id in (1,2))
         ON CONFLICT (from_id, to_id, comments) DO NOTHING),
   delete_stmt AS (DELETE from product where to_id in (1,2) or from_id in (1,2) RETURNING *)
select * from delete_stmt 

但是在select * from product之后，我得到（在 from+it\to_id 中有 [1, 2]）：

如何？

Answer 1

所以我设法做到了：

with
   delete_duplicates as (
      delete from product where
         (from_id in (1,2) and to_id in (1,2)) or
         (from_id = 3 and to_id in (1,2)) or
         (from_id in (1,2) and to_id  = 3)
      RETURNING id),
   insert_stmt_from_id AS (
      insert into product
         (from_id, to_id, comments, data)
         (select 3,to_id,comments,data from product
           where from_id in (1,2) and not id in (select id from delete_duplicates))
       ON CONFLICT (from_id, to_id, comments) DO NOTHING RETURNING id),
   insert_stmt_to_id AS (
      insert into product
         (from_id, to_id, comments, data)
         (select from_id,3,comments,data from product
           where to_id in (1,2) and not id in (select id from delete_duplicates))
       ON CONFLICT (from_id, to_id, comments) DO NOTHING RETURNING id),
   delete_leftovers_stmt AS (DELETE from product where from_id in (1,2) or to_id in (1,2) RETURNING id)
select id from delete_duplicates union select id from delete_leftovers_stmt;

Answer 2

你需要一个事务，以便所有命令都在一个额外的状态下运行，它会给你一个答案，结果只有一个

BEGIN; insert into product (from_id, to_id, comments, data) (select from_id,3,comments,data from product where to_id in (1,2)) ON CONFLICT (from_id, to_id, comments) DO NOTHING; insert into product (from_id, to_id, comments, data) (select 3,to_id,comments,data from product where from_id in (1,2)) ON CONFLICT (from_id, to_id, comments) DO NOTHING; DELETE from product where to_id in (1,2) or from_id in (1,2); COMMIT;

 ✓

 1 行受影响

 ✓

 6 行受影响

 ✓

SELECT * FROM product

编号 |  from_id |  to_id | 评论 | 数据
 -: |  ------: |  ----: |  :------- |  :---
  8 |  3 |  3 | 布拉|  {}  

db<> 在这里摆弄