Question

我的数据库中有一张包含友谊记录的表格。表中有两列：user1_id和user2_id，表示user1是user2的朋友。它被设置为user1_id总是＆lt; user2_id。我正在尝试编写一个查询来返回按共享朋友数量排序的用户对。但是，我希望在结果中只包含不是朋友的对。到目前为止，这是我的查询：

SELECT af.1, bf.1, count(*) as count
FROM (SELECT *
      FROM friends
      UNION
      SELECT user2_id, user1_id
      FROM friends ) af,
     (SELECT *
      FROM friends
      UNION
      SELECT user2_id, user1_id
      FROM friends ) bf
WHERE af.1 != bf.1 
AND af.2 = bf.2
GROUP BY af.1, bf.1
ORDER BY count desc

这应该给我所有用户排序的用户组合。但是，它不会省略包含实际上是朋友的对的元组。我可以在java中做到这一点，但我更愿意在数据库方面这样做。有没有好办法呢？

Answer 1

SELECT af.1, bf.1, count(*) as count
FROM (SELECT *
      FROM friends
      UNION
      SELECT user2_id, user1_id
      FROM friends ) af,
     (SELECT *
      FROM friends
      UNION
      SELECT user2_id, user1_id
      FROM friends ) bf
WHERE af.1 != bf.1 
AND af.2 = bf.2
AND NOT EXISTS (SELECT 1
                FROM friends
                WHERE user1_id = af.1 AND user2_id = bf.1
                OR user1_id = bf.1 AND user2_id = af.1)
GROUP BY af.1, bf.1
ORDER BY count desc

Answer 2

您可能还想考虑使用CONNECT BY clause，尽管这可能有点过分。

根据其他元组中的条件省略元组

2 个答案: