Question

我正在处理一个典型的博客应用，并有一个返回以下数据的视图：

| post_id | title | publish_on | tag_id | tag_name |

| 1 | Why is Postgres awesome                | 2014-09-02 | 1    | tech |
| 1 | Why is Postgres awesome                | 2014-09-02 | 2    | postgres |
| 2 | How to ask a question on stackoverflow | 2014-09-10 | 1    | tech |
| 2 | How to ask a question on stackoverflow | 2014-09-10 | 2    | postgres |
| 2 | How to ask a question on stackoverflow | 2014-09-10 | 3    | guide |
| 3 | This is a draft                        | null       | null | null |
| 4 | This is something else without a tag   | 2014-10-10 | null | null |
| 5 | This question is also published on 9/2 | 2014-09-02 | null | null |
| 6 | And so is this                         | 2014-09-02 | 1    | tech |
| 7 | But this one is on 9/10                | 2014-09-10 | 3    | guide|
| 8 | This is on 10/10                       | 2014-10-10 | null | null |
| 9 | And so is this                         | 2014-10-10 | 2    | postgres |
| 10| This is another draft                  | null       | null | null |

我希望按发布日期对帖子进行分组，然后为每个广告素材选择前3个帖子（这将显示在信息中心内，以便用户可以知道今天将发布哪些帖子，下周某个时间后来）现在我尝试these solutions使用了类似的东西：

ROW_NUMBER() OVER (PARTITION BY publish_on ORDER BY publish_on DESC)

但由于多个标记可以复制行，因此这些查询会失败。我还尝试了各种PARTION BY条件的组合，但我想我不太了解它们以使其正常运行。

任何帮助/指示赞赏！

更新：预期输出

对于每个publish_on日期，我希望预计在该日期发布N（3）个帖子。

| 1 | Why is Postgres awesome                | 2014-09-02 | 1    | tech |
| 1 | Why is Postgres awesome                | 2014-09-02 | 2    | postgres |
| 5 | This question is also published on 9/2 | 2014-09-02 | null | null |
| 6 | And so is this                         | 2014-09-02 | 1    | tech |

| 2 | How to ask a question on stackoverflow | 2014-09-10 | 1    | tech |
| 2 | How to ask a question on stackoverflow | 2014-09-10 | 2    | postgres |
| 2 | How to ask a question on stackoverflow | 2014-09-10 | 3    | guide |
| 7 | But this one is on 9/10                | 2014-09-10 | 3    | guide|

| 4 | This is something else without a tag   | 2014-10-10 | null | null |
| 8 | This is on 10/10                       | 2014-10-10 | null | null |
| 9 | And so is this                         | 2014-10-10 | 2    | postgres |

| 3 | This is a draft                        | null       | null | null |
| 10| This is another draft                  | null       | null | null |

希望这可以使问题更清楚地理解。

Answer 1

这是你在找什么？ SQL Fiddle

SELECT * 
FROM (SELECT ROW_NUMBER() OVER (PARTITION BY tag_name order by publish_on DESC) AS r,
    t.*
    from blog t ) x
where x.r <= 3

解释和问题：

我假设“每个桶”是指tag_name（或tag_id）。那么你只需要“每个桶”中最近的3个帖子。如果帖子被多次标记，那么您希望如何对待它们 - 每个标记出现一次 - 或者每个结果集只出现一次？

修改

现在可以按预期显示结果。 SQL Fiddle for this here.

SELECT DISTINCT x.Post_id, y.title, x.Publish_on, y.tag_id, y.tag_name FROM blog y INNER JOIN (SELECT ROW_NUMBER() OVER (PARTITION BY publish_on order by publish_on DESC) AS r, t.post_id, t.publish_on from (SELECT DISTINCT s.post_id, s.publish_on FROM blog s) t ) x ON x.post_id = y.post_id where x.r <= 3 ORDER BY x.publish_on

增加复杂性的主要问题是表结构未规范化。实际上这应该是3个表，因此描述和日期不会在不同的行中重复，即

CREATE TABLE blog (post_id int not null, title varchar(50) not null, publish_on date) CREATE TABLE blog_tag (post_id int not null, tag_ig int not null) CREATE TABLE tag (tag_id int not null, tag_name varchar(10) not null)

然后可以用see full SQL Fidle for this here.
替换SQL
SELECT x.Post_id, x.title, x.Publish_on, t.tag_id, t.tag_name FROM (SELECT ROW_NUMBER() OVER (PARTITION BY publish_on order by publish_on DESC) AS r, b.* from blog b) x LEFT JOIN blog_tag bt ON bt.post_id = x.post_id LEFT JOIN tag t ON t.tag_id = bt.tag_id WHERE x.r <= 3 ORDER BY x.publish_on, x.post_id, t.tag_id

PostgreSQL找到按日期分组的前N行

1 个答案: