postgresql：将 table1 复制到 table2 和 table3，但使用返回的 table2 值插入到 table3

Question

我有三张桌子：

table1(id, name, foo)
表 2（ID，名称）
table3(id, id2, name, foo, bar)

我想将table1中的数据复制到表2和3，但table3.id2必须对应table2中的行。我想我需要对 RETURNING 命令做点什么，但我一直没有成功。

WITH oldstuff AS (
    SELECT name, foo, 'some value' AS bar
    FROM table1
  ),
  newstuff AS (
    INSERT INTO table2 (name)
    SELECT name FROM oldstuff
    RETURNING id AS id2 /* but i also need oldstuff.name, oldstuff.foo,and oldstuff.bar */
  )
INSERT INTO table3 (id2, name, foo, bar)
SELECT * FROM newstuff; /* I cant just do a join here because the fields arent unique */

Answer 1

从概念上讲，你会这样做：

WITH oldstuff AS (
      SELECT name, foo, 'some value' AS bar
      FROM table1
     ),
     newstuff AS (
      INSERT INTO table2 (name)
          SELECT name
          FROM oldstuff
          RETURNING *
    )
INSERT INTO table3 (id2, name, foo, bar)
    SELECT ns.id, ns.name, os.foo, os.bar
    FROM newstuff ns join
         oldstuff os
         on ns.name = os.name;

根据您的说法，这并不完全符合您的要求，因为newstuff 中没有唯一的名称。要么为原始表生成一个唯一的 id，要么只在其中插入唯一的数据：

      INSERT INTO table2 (name)
          SELECT DISTINCT name
          FROM oldstuff
          RETURNING *;

嗯。 . .可能有一种笨拙的方式：

WITH oldstuff AS (
      SELECT name, foo, 'some value' AS bar
      FROM table1
     ),
     newstuff AS (
      INSERT INTO table2 (name)
          SELECT name
          FROM oldstuff
          RETURNING *
    )
INSERT INTO table3 (id2, name, foo, bar)
    SELECT ns.id, ns.name, os.foo, os.bar
    FROM (SELECT ns.*, ROW_NUMBER() OVER (PARTITION BY name ORDER BY name) as seqnum
          FROM newstuff ns
         ) ns JOIN
         (SELECT os.*, ROW_NUMBER() OVER (PARTITION BY name ORDER BY name) as seqnum
          FROM oldstuff os
         ) os
         on ns.name = os.name and ns.seqnum = os.seqnum;

这将适用于重复的名称，并且您在决赛桌中得到一个匹配项。