如何根据 SQL 中的时间戳查找特定事件的下一个序列？

Posted 2023-03-31

【中文标题】如何根据 SQL 中的时间戳查找特定事件的下一个序列？

【英文标题】：How to find the next sequence of a particular event based on timestamp in SQL?

【发布时间】：2021-01-09 19:14:42

【问题描述】：

我有 2 个表：- views_tbl 表（页面浏览量）和 play_tbl 表（记录客户播放特定视频的时间）。来自views_tbl 的nowPlaying page_type 将在play_tbl 中具有相同的时间戳。因此时间戳是唯一可用于连接这两个表的列。

play_ts = event_ts when page_type = 'nowPlaying';

我正在尝试根据以下标准查找导致成功播放事件的browseFind 计数（nowPlaying page_type 来自views_tbl）。

如果browseFind 后跟browseSearch，那么这种情况将被忽略，（browseFind 之后应该没有browseSearch 事件，但可以有其他事件）。

例如对于第一个browseFind 发生在2021-01-07 00:36:57.321，我们有一个立即的browseSearch，因此将被忽略。 对于browseFind 在2021-01-07 11:17:27.286 的下一次出现，在以下nowPlaying page_type 之间没有立即的browseSearch，因此将在计数中考虑这一点。还有来自play_tbl 的duration_seconds 应该大于 30 才能计算在内。如果持续时间小于 30 秒，则忽略此情况。

我试图通过将两个表合并为一个联合来做到这一点，但 LEAD 和 LAG 没有给我任何结果。

select
tbl_name,
play_ts,
ctrl_nm,
genre,
album_id,
song_id,
duration_seconds
 FROM play_tbl
union
select tbl_name, event_ts, page_type, NULL, NULL, NULL,NULL from views_tbl  ;

然后尝试使用LEAD 和LAG。我使用 Redshift 作为我的数据库。我无法获得正确的行组合。

输出： 导致 nowPlaying 持续时间大于 30 秒且它们之间没有 browseSearch 事件的 browseFind 计数。

任何帮助将不胜感激。我不知道如何编写可以逐步或逐行查找正确条件的sql。

表格 DDL：

create table play_tbl (tbl_name varchar(10), play_ts timestamp, ctrl_nm varchar(10), genre varchar(100), album_id varchar(50), song_id varchar(50), duration_seconds int);

insert into play_tbl values ('play',  '2021-01-07 03:17:51.474',  'umse',  'browseGema',  'B089999999',  'B99FPD1MCJ',  212);
insert into play_tbl values ('play',  '2021-01-07 11:17:37.228',  'umse',  'browseGema',  'B089999999',  'B99K2XSNY3',  175);
insert into play_tbl values ('play',  '2021-01-07 11:48:19.136', 'umse',  'browseGema',  'B089999999',  'B99328YJBW',  155);
insert into play_tbl values ('play',  '2021-01-07 11:49:51.419',  'umse',  'browseGema',  'B089999999',  'B999PR4XRS',  48);
insert into play_tbl values ('play',  '2021-01-07 11:50:26.264',  'umse',  'browseGema',  'B089999999',  'B99C98DB5T',  89);
insert into play_tbl values ('play',  '2021-01-07 11:52:56.433',  'umse',  'browseGema',  'B089999999',  'B99L88RQZS',  12);
insert into play_tbl values ('play',  '2021-01-07 19:14:53.865',  'umse',  'browseGema',  'B089999999',  'B99GWTZZ3H',  23);
insert into play_tbl values ('play',  '2021-01-07 19:40:46.806',  'umse',  'browseGema',  'B089999999',  'B99NCVV16G',  185);
insert into play_tbl values ('play',  '2021-01-07 19:48:47.708',  'umse',  'browseGema',  'B089999999',  'B99BVYG1S6',  48);
insert into play_tbl values ('play',  '2021-01-07 21:30:03.102',  'umse',  'browseGema',  'B089999999',  'B99C9KDQW6',  69);
insert into play_tbl values ('play',  '2021-01-07 11:17:33.655',  'umse',  'browseGema',  'B089999999',  'B99FPH4GD1',  232);
insert into play_tbl values ('play',  '2021-01-07 21:30:05.931',  'umse',  'browseGema',  'B089999999',  'B99GFBC4V5',   2);



create table views_tbl (tbl_name varchar(10), event_ts timestamp, page_type varchar(150));

insert into views_tbl values ('view',  '2021-01-07 00:36:55.33','detail-userPlaylist');       
insert into views_tbl values ('view',  '2021-01-07 00:36:52.328','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 00:36:57.321','browseFind');        
insert into views_tbl values ('view',  '2021-01-07 00:37:03.871','browseSearch');        
insert into views_tbl values ('view',  '2021-01-07 03:17:42.541','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 03:17:45.78','detail-userPlaylist');      
insert into views_tbl values ('view',  '2021-01-07 03:17:51.474','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 11:17:25.38','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 11:17:27.286','browseFind');        
insert into views_tbl values ('view',  '2021-01-07 11:17:29.048','browseGema');        
insert into views_tbl values ('view',  '2021-01-07 11:17:32.342','browseGema');        
insert into views_tbl values ('view',  '2021-01-07 11:17:31.363','detail-playlist');        
insert into views_tbl values ('view',  '2021-01-07 11:17:34.221','detail-playlist');        
insert into views_tbl values ('view',  '2021-01-07 11:17:37.228','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 11:48:04.055','browseGema');        
insert into views_tbl values ('view',  '2021-01-07 11:48:04.796','browseFind');        
insert into views_tbl values ('view',  '2021-01-07 11:48:05.359','browseSearch');        
insert into views_tbl values ('view',  '2021-01-07 11:48:08.778','browseSearch');        
insert into views_tbl values ('view',  '2021-01-07 11:48:19.136','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 11:48:12.066','detail-playlist');        
insert into views_tbl values ('view',  '2021-01-07 11:49:51.419','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 11:50:12.778','browseSearch');        
insert into views_tbl values ('view',  '2021-01-07 11:50:17.936','browseSearch');        
insert into views_tbl values ('view',  '2021-01-07 11:50:26.264','nowPlaying');               
insert into views_tbl values ('view',  '2021-01-07 11:52:21.502','browseFind');        
insert into views_tbl values ('view',  '2021-01-07 11:52:26.201','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 11:52:27.375','browseHome');        
insert into views_tbl values ('view',  '2021-01-07 11:52:36.111','detail-playlist');        
insert into views_tbl values ('view',  '2021-01-07 11:52:42.604','browseHome');      
insert into views_tbl values ('view',  '2021-01-07 11:52:41.909','detail-playlist');        
insert into views_tbl values ('view',  '2021-01-07 11:52:45.771','browseFind');               
insert into views_tbl values ('view',  '2021-01-07 11:52:53.719','detail-album');      
insert into views_tbl values ('view',  '2021-01-07 11:52:56.433','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 11:59:38.747','browseSearch');        
insert into views_tbl values ('view',  '2021-01-07 11:59:39.718','browseFind');        
insert into views_tbl values ('view',  '2021-01-07 11:59:41.481','browseHome');        
insert into views_tbl values ('view',  '2021-01-07 11:59:43.998','detail-playlist');        
insert into views_tbl values ('view',  '2021-01-07 11:59:47.427','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 12:00:58.284','browseHome');      
insert into views_tbl values ('view',  '2021-01-07 12:24:15.929','browseHome');        
insert into views_tbl values ('view',  '2021-01-07 18:34:15.191','browseHome');        
insert into views_tbl values ('view',  '2021-01-07 19:14:47.426','browseHome');        
insert into views_tbl values ('view',  '2021-01-07 19:14:50.187','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 19:14:52.098','detail-userPlaylist');     
insert into views_tbl values ('view',  '2021-01-07 19:14:53.865','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 19:40:45.267','detail-userPlaylist');     
insert into views_tbl values ('view',  '2021-01-07 19:40:43.486','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 19:40:46.806','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 19:48:42.942','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 19:48:46.362','detail-userPlaylists');    
insert into views_tbl values ('view',  '2021-01-07 19:48:47.708','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 21:29:54.488','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 21:29:56.973','activity-feed');        
insert into views_tbl values ('view',  '2021-01-07 21:30:01.206','detail-userPlaylist');     
insert into views_tbl values ('view',  '2021-01-07 21:29:59.795','cloudLibrary-playlists');   
insert into views_tbl values ('view',  '2021-01-07 21:30:03.102','nowPlaying');        
insert into views_tbl values ('view',  '2021-01-07 21:30:05.931','nowPlaying');

【问题讨论】：

Redshift 还是 mysql？这是两个非常不同的数据库。

想要的结果是什么

@a_horse_with_no_name 数据库是 Redshift，但如果需要使用仅在 MySQL 中可用的功能，我可以将数据复制到 MySQL。谢谢。

@Strawberry: 期望的输出是 2

【参考方案1】：

如果我正确理解了您的问题（您的数据的答案是 4），那么我认为问题在于如何将 browseSearch 和 browseFind 视为视图事件中的特殊情况。在下面的 SQL 中，我使用 DECODE 进行了此操作（如果您愿意，请使用 CASE）并将 IGNORE NULLS 添加到 LAG()。我在子选择中留下了额外的列，以防您想自己运行每个级别以适应正在发生的事情。如果不完全正确，它应该会为您提供根据您的确切需求进行调整所需的提示。

SELECT COUNT(DECODE(prev_browse,'browseFind',duration_seconds,NULL)) AS cnt
FROM (SELECT event_ts,
             LAG(browse_type IGNORE NULLS) OVER (ORDER BY event_ts) AS prev_browse,
             page_type,
             duration_seconds
      FROM (SELECT event_ts,
                   page_type,
                   DECODE(page_type IN ('browseFind', 'browseSearch'),
                          TRUE, page_type,
                          NULL
                   ) AS browse_type,
                   duration_seconds
            FROM views_tbl v
              LEFT JOIN play_tbl p
                     ON v.event_ts = p.play_ts
                    AND v.page_type = 'nowPlaying'
            WHERE page_type IN ('nowPlaying','browseFind','browseSearch')
            AND   (p.duration_seconds >= 30 OR p.duration_seconds IS NULL)));

【讨论】：

感谢您的提示，正确答案应该是2。我想我可以修改查询以适应条件。

很高兴听到。一切顺利。