文章目录
某音短视频
描述
用户-视频互动表tb_user_video_log
id | uid | video_id | start_time | end_time | if_follow | if_like | if_retweet | comment_id |
---|---|---|---|---|---|---|---|---|
1 | 101 | 2001 | 2021-10-01 10:00:00 | 2021-10-01 10:00:20 | 0 | 1 | 1 | NULL |
2 | 102 | 2001 | 2021-10-01 10:00:00 | 2021-10-01 10:00:15 | 0 | 0 | 1 | NULL |
3 | 103 | 2001 | 2021-10-01 11:00:50 | 2021-10-01 11:01:15 | 0 | 1 | 0 | 1732526 |
4 | 102 | 2002 | 2021-09-10 11:00:00 | 2021-09-10 11:00:30 | 1 | 0 | 1 | NULL |
5 | 103 | 2002 | 2021-10-01 10:59:05 | 2021-10-01 11:00:05 | 1 | 0 | 0 | NULL |
(uid-用户ID, video_id-视频ID, start_time-开始观看时间, end_time-结束观看时间, if_follow-是否关注, if_like-是否点赞, if_retweet-是否转发, comment_id-评论ID)
短视频信息表tb_video_info
id | video_id | author | tag | duration | release_time |
---|---|---|---|---|---|
1 | 2001 | 901 | 影视 | 30 | 2021-01-01 07:00:00 |
2 | 2002 | 901 | 美食 | 60 | 2021-01-01 07:00:00 |
3 | 2003 | 902 | 旅游 | 90 | 2020-01-01 07:00:00 |
(video_id-视频ID, author-创作者ID, tag-类别标签, duration-视频时长, release_time-发布时间)
SQL 156 各个视频的平均完播率
问题:计算2021年里有播放记录的每个视频的完播率(结果保留三位小数),并按完播率降序排序
注:视频完播率是指完成播放次数占总播放次数的比例。简单起见,结束观看时间与开始播放时间的差>=视频时长时,视为完成播放。
输出示例:
示例数据的结果如下:
video_id | avg_comp_play_rate |
---|---|
2001 | 0.667 |
2002 | 0.000 |
解释:
视频2001在2021年10月有3次播放记录,观看时长分别为30秒、24秒、34秒,视频时长30秒,因此有两次是被认为完成播放了的,故完播率为0.667;
视频2002在2021年9月和10月共2次播放记录,观看时长分别为42秒、30秒,视频时长60秒,故完播率为0.000。
数据示例
输入:
DROP TABLE IF EXISTS tb_user_video_log, tb_video_info;
CREATE TABLE tb_user_video_log (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
uid INT NOT NULL COMMENT '用户ID',
video_id INT NOT NULL COMMENT '视频ID',
start_time datetime COMMENT '开始观看时间',
end_time datetime COMMENT '结束观看时间',
if_follow TINYINT COMMENT '是否关注',
if_like TINYINT COMMENT '是否点赞',
if_retweet TINYINT COMMENT '是否转发',
comment_id INT COMMENT '评论ID'
) CHARACTER SET utf8 COLLATE utf8_bin;
CREATE TABLE tb_video_info (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
video_id INT UNIQUE NOT NULL COMMENT '视频ID',
author INT NOT NULL COMMENT '创作者ID',
tag VARCHAR(16) NOT NULL COMMENT '类别标签',
duration INT NOT NULL COMMENT '视频时长(秒数)',
release_time datetime NOT NULL COMMENT '发布时间'
)CHARACTER SET utf8 COLLATE utf8_bin;
INSERT INTO tb_user_video_log(uid, video_id, start_time, end_time, if_follow, if_like, if_retweet, comment_id) VALUES
(101, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:30', 0, 1, 1, null),
(102, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:24', 0, 0, 1, null),
(103, 2001, '2021-10-01 11:00:00', '2021-10-01 11:00:34', 0, 1, 0, 1732526),
(101, 2002, '2021-09-01 10:00:00', '2021-09-01 10:00:42', 1, 0, 1, null),
(102, 2002, '2021-10-01 11:00:00', '2021-10-01 11:00:30', 1, 0, 1, null);
INSERT INTO tb_video_info(video_id, author, tag, duration, release_time) VALUES
(2001, 901, '影视', 30, '2021-01-01 7:00:00'),
(2002, 901, '美食', 60, '2021-01-01 7:00:00'),
(2003, 902, '旅游', 90, '2021-01-01 7:00:00');
复制
输出:
2001|0.667
2002|0.000
解答
select a.video_id,
round(sum(if(end_time-start_time>=duration,1,0)) /count(start_time),3) as avg_play_progress
from tb_user_video_log a
left join tb_video_info tvi on a.video_id = tvi.video_id
and year(start_time) = 2021
group by a.video_id
order by avg_play_progress DESC ;
涉及知识点
if xxx 1,0 以及时间戳相加减得到的结果为数字,可以直接做比较
SQL157 平均播放进度大于60%的视频类别
问题:计算各类视频的平均播放进度,将进度大于60%的类别输出。
注:
- 播放进度=播放时长÷视频时长*100%,当播放时长大于视频时长时,播放进度均记为100%。
- 结果保留两位小数,并按播放进度倒序排序。
输出示例:
示例数据的输出结果如下:
tag | avg_play_progress |
---|---|
影视 | 90.00% |
美食 | 75.00% |
解释:
影视类视频2001被用户101、102、103看过,播放进度分别为:30秒(100%)、21秒(70%)、30秒(100%),平均播放进度为90.00%(保留两位小数);
美食类视频2002被用户102、103看过,播放进度分别为:30秒(50%)、60秒(100%),平均播放进度为75.00%(保留两位小数);
数据示例
输入:
DROP TABLE IF EXISTS tb_user_video_log, tb_video_info;
CREATE TABLE tb_user_video_log (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
uid INT NOT NULL COMMENT '用户ID',
video_id INT NOT NULL COMMENT '视频ID',
start_time datetime COMMENT '开始观看时间',
end_time datetime COMMENT '结束观看时间',
if_follow TINYINT COMMENT '是否关注',
if_like TINYINT COMMENT '是否点赞',
if_retweet TINYINT COMMENT '是否转发',
comment_id INT COMMENT '评论ID'
) CHARACTER SET utf8 COLLATE utf8_bin;
CREATE TABLE tb_video_info (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
video_id INT UNIQUE NOT NULL COMMENT '视频ID',
author INT NOT NULL COMMENT '创作者ID',
tag VARCHAR(16) NOT NULL COMMENT '类别标签',
duration INT NOT NULL COMMENT '视频时长(秒数)',
release_time datetime NOT NULL COMMENT '发布时间'
)CHARACTER SET utf8 COLLATE utf8_bin;
INSERT INTO tb_user_video_log(uid, video_id, start_time, end_time, if_follow, if_like, if_retweet, comment_id) VALUES
(101, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:30', 0, 1, 1, null),
(102, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:21', 0, 0, 1, null),
(103, 2001, '2021-10-01 11:00:50', '2021-10-01 11:01:20', 0, 1, 0, 1732526),
(102, 2002, '2021-10-01 11:00:00', '2021-10-01 11:00:30', 1, 0, 1, null),
(103, 2002, '2021-10-01 10:59:05', '2021-10-01 11:00:05', 1, 0, 1, null);
INSERT INTO tb_video_info(video_id, author, tag, duration, release_time) VALUES
(2001, 901, '影视', 30, '2021-01-01 7:00:00'),
(2002, 901, '美食', 60, '2021-01-01 7:00:00'),
(2003, 902, '旅游', 90, '2020-01-01 7:00:00');
复制
输出:
影视|90.00%
美食|75.00%
解答
select tag,
concat(round((sum(case when TIMESTAMPDIFF(SECOND,start_time,end_time) >= duration then 1
else TIMESTAMPDIFF(second ,start_time,end_time)/duration end)/count(*))*100,2),'%') as avg_comp_play_rate
from tb_user_video_log a
left join tb_video_info b on a.video_id = b.video_id
group by tag
having SUBSTRING_INDEX(avg_comp_play_rate,'%',1)>60
order by avg_comp_play_rate desc ;
涉及知识点
TIMESTAMPDIFF函数
TIMESTAMPDIFF(unit,begin,end),其中begin和end是DATE或DATETIME表达式。TIMESTAMPDIFF函数允许其参数具有混合类型,例如,begin是DATE值,end可以是DATETIME值。 如果使用DATE值,则TIMESTAMPDIFF函数将其视为时间部分为“00:00:00”的DATETIME值。
unit参数是确定(end-begin)的结果的单位,表示为整数。 以下是有效单位:
MICROSECOND 微秒
SECOND 秒
MINUTE 分钟
HOUR 小时
DAY 天
WEEK 周
MONTH 月份
QUARTER
YEAR 年份
示例
select TIMESTAMPDIFF(day ,'2020-01-02','2020-01-04') as result;
结果为
2
SUBSTRING_INDEX 函数
SUBSTRING_INDEX(STR,DELIM,COUNT)
str:要处理的字符串 delim:分隔符 count:计数
示例
select SUBSTRING_INDEX('2020-02-01','-',1) as result;
select SUBSTRING_INDEX('2020-02-01','-',2) as result;
结果分别为i
2020
2020-02
SQL158 每类视频近一个月的转发量/率
问题:统计在有用户互动的最近一个月(按包含当天在内的近30天算,比如10月31日的近30天为10.2~10.31之间的数据)中,每类视频的转发量和转发率(保留3位小数)。
注:转发率=转发量÷播放量。结果按转发率降序排序。
输出示例:
示例数据的输出结果如下
tag | retweet_cut | retweet_rate |
---|---|---|
影视 | 2 | 0.667 |
美食 | 1 | 0.500 |
解释:
由表tb_user_video_log的数据可得,数据转储当天为2021年10月1日。近30天内,影视类视频2001共有3次播放记录,被转发2次,转发率为0.667;美食类视频2002共有2次播放记录,1次被转发,转发率为0.500。
数据示例
DROP TABLE IF EXISTS tb_user_video_log, tb_video_info;
CREATE TABLE tb_user_video_log (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
uid INT NOT NULL COMMENT '用户ID',
video_id INT NOT NULL COMMENT '视频ID',
start_time datetime COMMENT '开始观看时间',
end_time datetime COMMENT '结束观看时间',
if_follow TINYINT COMMENT '是否关注',
if_like TINYINT COMMENT '是否点赞',
if_retweet TINYINT COMMENT '是否转发',
comment_id INT COMMENT '评论ID'
) CHARACTER SET utf8 COLLATE utf8_bin;
CREATE TABLE tb_video_info (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
video_id INT UNIQUE NOT NULL COMMENT '视频ID',
author INT NOT NULL COMMENT '创作者ID',
tag VARCHAR(16) NOT NULL COMMENT '类别标签',
duration INT NOT NULL COMMENT '视频时长(秒数)',
release_time datetime NOT NULL COMMENT '发布时间'
)CHARACTER SET utf8 COLLATE utf8_bin;
INSERT INTO tb_user_video_log(uid, video_id, start_time, end_time, if_follow, if_like, if_retweet, comment_id) VALUES
(101, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:20', 0, 1, 1, null)
,(102, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:15', 0, 0, 1, null)
,(103, 2001, '2021-10-01 11:00:50', '2021-10-01 11:01:15', 0, 1, 0, 1732526)
,(102, 2002, '2021-09-10 11:00:00', '2021-09-10 11:00:30', 1, 0, 1, null)
,(103, 2002, '2021-10-01 10:59:05', '2021-10-01 11:00:05', 1, 0, 0, null);
INSERT INTO tb_video_info(video_id, author, tag, duration, release_time) VALUES
(2001, 901, '影视', 30, '2021-01-01 7:00:00')
,(2002, 901, '美食', 60, '2021-01-01 7:00:00')
,(2003, 902, '旅游', 90, '2020-01-01 7:00:00');
复制
输出:
影视|2|0.667
美食|1|0.500
解答
select tag,
sum(if(if_retweet =1,1,0)) as retweet_cut,
round(sum(if(if_retweet =1,1,0)) /count(if_retweet),3) as retweet_rate
from tb_user_video_log a
left join tb_video_info b on a.video_id = b.video_id
where datediff(date((select max(start_time) from tb_user_video_log)),date(start_time)) <= 29
group by tag
order by retweet_rate desc ;
涉及知识点
datediff函数
由于本题提出了前提,要用户互动的最近一个月求出转发率与播放率,所以我们必须要对条件进行限制
datediff函数 返回两个日期时间之差
datediff(date1,date2)
SELECT DATEDIFF('2010-6-30','2010-6-26') AS DiffDate
结果为
4
SQL 159 每个创作者每月的涨粉率及截止当前的总粉丝量
用户-视频互动表tb_user_video_log
id | uid | video_id | start_time | end_time | if_follow | if_like | if_retweet | comment_id |
---|---|---|---|---|---|---|---|---|
1 | 101 | 2001 | 2021-09-01 10:00:00 | 2021-09-01 10:00:20 | 0 | 1 | 1 | NULL |
2 | 105 | 2002 | 2021-09-10 11:00:00 | 2021-09-10 11:00:30 | 1 | 0 | 1 | NULL |
3 | 101 | 2001 | 2021-10-01 10:00:00 | 2021-10-01 10:00:20 | 1 | 1 | 1 | NULL |
4 | 102 | 2001 | 2021-10-01 10:00:00 | 2021-10-01 10:00:15 | 0 | 0 | 1 | NULL |
5 | 103 | 2001 | 2021-10-01 11:00:50 | 2021-10-01 11:01:15 | 1 | 1 | 0 | 1732526 |
6 | 106 | 2002 | 2021-10-01 10:59:05 | 2021-10-01 11:00:05 | 2 | 0 | 0 | NULL |
(uid-用户ID, video_id-视频ID, start_time-开始观看时间, end_time-结束观看时间, if_follow-是否关注, if_like-是否点赞, if_retweet-是否转发, comment_id-评论ID)
短视频信息表tb_video_info
id | video_id | author | tag | duration | release_time |
---|---|---|---|---|---|
1 | 2001 | 901 | 影视 | 30 | 2021-01-01 07:00:00 |
2 | 2002 | 901 | 美食 | 60 | 2021-01-01 07:00:00 |
3 | 2003 | 902 | 旅游 | 90 | 2020-01-01 07:00:00 |
4 | 2004 | 902 | 美女 | 90 | 2020-01-01 08:00:00 |
(video_id-视频ID, author-创作者ID, tag-类别标签, duration-视频时长, release_time-发布时间)
问题:计算2021年里每个创作者每月的涨粉率及截止当月的总粉丝量
注:
- 涨粉率=(加粉量 - 掉粉量) / 播放量。结果按创作者ID、总粉丝量升序排序。
- if_follow-是否关注为1表示用户观看视频中关注了视频创作者,为0表示此次互动前后关注状态未发生变化,为2表示本次观看过程中取消了关注。
输出示例:
示例数据的输出结果如下
author | month | fans_growth_rate | total_fans |
---|---|---|---|
901 | 2021-09 | 0.500 | 1 |
901 | 2021-10 | 0.250 | 2 |
解释:
示例数据中表tb_user_video_log里只有视频2001和2002的播放记录,都来自创作者901,播放时间在2021年9月和10月;其中9月里加粉量为1,掉粉量为0,播放量为2,因此涨粉率为0.500(保留3位小数);其中10月里加粉量为2,掉份量为1,播放量为4,因此涨粉率为0.250,截止当前总粉丝数为2。
数据示例
输入:
DROP TABLE IF EXISTS tb_user_video_log, tb_video_info;
CREATE TABLE tb_user_video_log (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
uid INT NOT NULL COMMENT '用户ID',
video_id INT NOT NULL COMMENT '视频ID',
start_time datetime COMMENT '开始观看时间',
end_time datetime COMMENT '结束观看时间',
if_follow TINYINT COMMENT '是否关注',
if_like TINYINT COMMENT '是否点赞',
if_retweet TINYINT COMMENT '是否转发',
comment_id INT COMMENT '评论ID'
) CHARACTER SET utf8 COLLATE utf8_bin;
CREATE TABLE tb_video_info (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
video_id INT UNIQUE NOT NULL COMMENT '视频ID',
author INT NOT NULL COMMENT '创作者ID',
tag VARCHAR(16) NOT NULL COMMENT '类别标签',
duration INT NOT NULL COMMENT '视频时长(秒数)',
release_time datetime NOT NULL COMMENT '发布时间'
)CHARACTER SET utf8 COLLATE utf8_bin;
INSERT INTO tb_user_video_log(uid, video_id, start_time, end_time, if_follow, if_like, if_retweet, comment_id) VALUES
(101, 2001, '2021-09-01 10:00:00', '2021-09-01 10:00:20', 0, 1, 1, null)
,(105, 2002, '2021-09-10 11:00:00', '2021-09-10 11:00:30', 1, 0, 1, null)
,(101, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:20', 1, 1, 1, null)
,(102, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:15', 0, 0, 1, null)
,(103, 2001, '2021-10-01 11:00:50', '2021-10-01 11:01:15', 1, 1, 0, 1732526)
,(106, 2002, '2021-10-01 10:59:05', '2021-10-01 11:00:05', 2, 0, 0, null);
INSERT INTO tb_video_info(video_id, author, tag, duration, release_time) VALUES
(2001, 901, '影视', 30, '2021-01-01 7:00:00')
,(2002, 901, '影视', 60, '2021-01-01 7:00:00')
,(2003, 902, '旅游', 90, '2020-01-01 7:00:00')
,(2004, 902, '美女', 90, '2020-01-01 8:00:00');
复制
输出:
901|2021-09|0.500|1
901|2021-10|0.250|2
解法1
select author,substring(start_time,1,7) month,
round(sum(case when if_follow =1 then 1
when if_follow =2 then -1
else 0 end ) /count(t1.author),3) as fans_growth_rate,
sum(sum(case when if_follow =1 then 1
when if_follow =2 then -1
else 0 end )) over (partition by author order by substring(start_time,1,7)) fans_total
from tb_video_info t1
left join tb_user_video_log t2 on t1.video_id = t2.video_id
where substring(start_time,1,4) = 2021
and substring(end_time,1,4) = 2021
group by author,month
order by author,fans_total;
解法2
SELECT B.AUTHOR AS AUTHOR,DATE_FORMAT(A.start_time,'%Y-%m') AS MONTH ,
ROUND((COUNT(CASE WHEN A.if_follow=1 THEN 1 END ) - COUNT(CASE WHEN A.if_follow=2 THEN 1 END ))/COUNT(1) ,3) AS FANS_GROWTH_RATE,
sum(sum(case when A.if_follow = 1 then 1
when A.if_follow = 2 then -1
else 0 end) ) over (partition by B.author order by date_format(A.start_time,'%Y-%m')) fans_total
FROM tb_user_video_log A
LEFT JOIN tb_video_info B
ON A.VIDEO_ID=B.video_id
WHERE year(A.start_time)=2021
and year(A.end_time)=2021
GROUP BY B.AUTHOR,DATE_FORMAT(A.start_time,'%Y-%m')
ORDER BY AUTHOR,fans_total
涉及知识点
date_format(start_time,‘%Y-%m’)
select date_format('2021-09-01 10:00:00','%Y-%m')
结果为:
2021-09
select date_format('2021-09-01 10:00:00','%Y-%m-%d %H:%i:%s')
结果为
2021-09-01 10:00:00
substring前面说过
case when 语法
case when xxx then xx
when xxx then xxx
else xx end
over (partition by)
如果在partition结果上聚合,千万注意聚合函数是逐条累计运行结果的!
而在group by后的结果集上使用聚合函数,会作用在分组下的所有记录上。
因为题目的要求是截止10月的粉丝总数,即包括了9月的粉丝,如果我们用group by 必须要对月份进行分组,那么就只能统计每一个月内的粉丝总数,但如果我们使用开窗函数通过partition by authr进行开窗,再进行累积结果就不会出现上述问题
以下就是group by 与partition by的区别
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-Ptx4DVIo-1670167877867)(E:\资料汇总\图片\605.png)]
[外链图片转存失败,源站可能有防盗链机制,建议将图片保存下来直接上传(img-X73kkKNl-1670167877868)(E:\资料汇总\图片\606.png)]
SQL160 国庆期间每类视频点赞量和转发量
描述
用户-视频互动表tb_user_video_log
id | uid | video_id | start_time | end_time | if_follow | if_like | if_retweet | comment_id |
---|---|---|---|---|---|---|---|---|
1 | 101 | 2001 | 2021-09-24 10:00:00 | 2021-09-24 10:00:20 | 1 | 1 | 0 | NULL |
2 | 105 | 2002 | 2021-09-25 11:00:00 | 2021-09-25 11:00:30 | 0 | 0 | 1 | NULL |
3 | 102 | 2002 | 2021-09-25 11:00:00 | 2021-09-25 11:00:30 | 1 | 1 | 1 | NULL |
4 | 101 | 2002 | 2021-09-26 11:00:00 | 2021-09-26 11:00:30 | 1 | 0 | 1 | NULL |
5 | 101 | 2002 | 2021-09-27 11:00:00 | 2021-09-27 11:00:30 | 1 | 1 | 0 | NULL |
6 | 102 | 2002 | 2021-09-28 11:00:00 | 2021-09-28 11:00:30 | 1 | 0 | 1 | NULL |
7 | 103 | 2002 | 2021-09-29 11:00:00 | 2021-10-02 11:00:30 | 1 | 0 | 1 | NULL |
8 | 102 | 2002 | 2021-09-30 11:00:00 | 2021-09-30 11:00:30 | 1 | 1 | 1 | NULL |
9 | 101 | 2001 | 2021-10-01 10:00:00 | 2021-10-01 10:00:20 | 1 | 1 | 0 | NULL |
10 | 102 | 2001 | 2021-10-01 10:00:00 | 2021-10-01 10:00:15 | 0 | 0 | 1 | NULL |
11 | 103 | 2001 | 2021-10-01 11:00:50 | 2021-10-01 11:01:15 | 1 | 1 | 0 | 1732526 |
12 | 106 | 2002 | 2021-10-02 10:59:05 | 2021-10-02 11:00:05 | 2 | 0 | 1 | NULL |
13 | 107 | 2002 | 2021-10-02 10:59:05 | 2021-10-02 11:00:05 | 1 | 0 | 1 | NULL |
14 | 108 | 2002 | 2021-10-02 10:59:05 | 2021-10-02 11:00:05 | 1 | 1 | 1 | NULL |
15 | 109 | 2002 | 2021-10-03 10:59:05 | 2021-10-03 11:00:05 | 0 | 1 | 0 | NULL |
(uid-用户ID, video_id-视频ID, start_time-开始观看时间, end_time-结束观看时间, if_follow-是否关注, if_like-是否点赞, if_retweet-是否转发, comment_id-评论ID)
短视频信息表tb_video_info
id | video_id | author | tag | duration | release_time |
---|---|---|---|---|---|
1 | 2001 | 901 | 旅游 | 30 | 2020-01-01 07:00:00 |
2 | 2002 | 901 | 旅游 | 60 | 2021-01-01 07:00:00 |
3 | 2003 | 902 | 影视 | 90 | 2020-01-01 07:00:00 |
4 | 2004 | 902 | 美女 | 90 | 2020-01-01 08:00:00 |
(video_id-视频ID, author-创作者ID, tag-类别标签, duration-视频时长, release_time-发布时间)
问题:统计2021年国庆头3天每类视频每天的近一周总点赞量和一周内最大单天转发量,结果按视频类别降序、日期升序排序。假设数据库中数据足够多,至少每个类别下国庆头3天及之前一周的每天都有播放记录。
输出示例:
示例数据的输出结果如下
tag | dt | sum_like_cnt_7d | max_retweet_cnt_7d |
---|---|---|---|
旅游 | 2021-10-01 | 5 | 2 |
旅游 | 2021-10-02 | 5 | 3 |
旅游 | 2021-10-03 | 6 | 3 |
解释:
由表tb_user_video_log里的数据可得只有旅游类视频的播放,2021年9月25到10月3日每天的点赞量和转发量如下:
tag | dt | like_cnt | retweet_cnt |
---|---|---|---|
旅游 | 2021-09-25 | 1 | 2 |
旅游 | 2021-09-26 | 0 | 1 |
旅游 | 2021-09-27 | 1 | 0 |
旅游 | 2021-09-28 | 0 | 1 |
旅游 | 2021-09-29 | 0 | 1 |
旅游 | 2021-09-30 | 1 | 1 |
旅游 | 2021-10-01 | 2 | 1 |
旅游 | 2021-10-02 | 1 | 3 |
旅游 | 2021-10-03 | 1 | 0 |
因此国庆头3天(10.0110.03)里10.01的近7天(9.2510.01)总点赞量为5次,单天最大转发量为2次(9月25那天最大);同理可得10.02和10.03的两个指标。
数据示例
输入:
DROP TABLE IF EXISTS tb_user_video_log, tb_video_info;
CREATE TABLE tb_user_video_log (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
uid INT NOT NULL COMMENT '用户ID',
video_id INT NOT NULL COMMENT '视频ID',
start_time datetime COMMENT '开始观看时间',
end_time datetime COMMENT '结束观看时间',
if_follow TINYINT COMMENT '是否关注',
if_like TINYINT COMMENT '是否点赞',
if_retweet TINYINT COMMENT '是否转发',
comment_id INT COMMENT '评论ID'
) CHARACTER SET utf8 COLLATE utf8_bin;
CREATE TABLE tb_video_info (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
video_id INT UNIQUE NOT NULL COMMENT '视频ID',
author INT NOT NULL COMMENT '创作者ID',
tag VARCHAR(16) NOT NULL COMMENT '类别标签',
duration INT NOT NULL COMMENT '视频时长(秒数)',
release_time datetime NOT NULL COMMENT '发布时间'
)CHARACTER SET utf8 COLLATE utf8_bin;
INSERT INTO tb_user_video_log(uid, video_id, start_time, end_time, if_follow, if_like, if_retweet, comment_id) VALUES
(101, 2001, '2021-09-24 10:00:00', '2021-09-24 10:00:20', 1, 1, 0, null)
,(105, 2002, '2021-09-25 11:00:00', '2021-09-25 11:00:30', 0, 0, 1, null)
,(102, 2002, '2021-09-25 11:00:00', '2021-09-25 11:00:30', 1, 1, 1, null)
,(101, 2002, '2021-09-26 11:00:00', '2021-09-26 11:00:30', 1, 0, 1, null)
,(101, 2002, '2021-09-27 11:00:00', '2021-09-27 11:00:30', 1, 1, 0, null)
,(102, 2002, '2021-09-28 11:00:00', '2021-09-28 11:00:30', 1, 0, 1, null)
,(103, 2002, '2021-09-29 11:00:00', '2021-09-29 11:00:30', 1, 0, 1, null)
,(102, 2002, '2021-09-30 11:00:00', '2021-09-30 11:00:30', 1, 1, 1, null)
,(101, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:20', 1, 1, 0, null)
,(102, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:15', 0, 0, 1, null)
,(103, 2001, '2021-10-01 11:00:50', '2021-10-01 11:01:15', 1, 1, 0, 1732526)
,(106, 2002, '2021-10-02 10:59:05', '2021-10-02 11:00:05', 2, 0, 1, null)
,(107, 2002, '2021-10-02 10:59:05', '2021-10-02 11:00:05', 1, 0, 1, null)
,(108, 2002, '2021-10-02 10:59:05', '2021-10-02 11:00:05', 1, 1, 1, null)
,(109, 2002, '2021-10-03 10:59:05', '2021-10-03 11:00:05', 0, 1, 0, null);
INSERT INTO tb_video_info(video_id, author, tag, duration, release_time) VALUES
(2001, 901, '旅游', 30, '2020-01-01 7:00:00')
,(2002, 901, '旅游', 60, '2021-01-01 7:00:00')
,(2003, 902, '影视', 90, '2020-01-01 7:00:00')
,(2004, 902, '美女', 90, '2020-01-01 8:00:00');
输出:
旅游|2021-10-01|5|2
旅游|2021-10-02|5|3
旅游|2021-10-03|6|3
解法1
select
*
from (
select tag,
time,
SUM(like_cnt) OVER w sum_like_cnt_7d,
MAX(retweet_cnt) OVER w sum_retweet_cnt_7d
from (
select tag,
date_format(start_time, '%Y-%m-%d') time,
sum(if_like) like_cnt,
sum(if_retweet) retweet_cnt
from tb_video_info t1
left join tb_user_video_log t2 on t1.video_id = t2.video_id
where date_format(start_time, '%Y-%m-%d') between '2021-09-25' and '2021-10-03'
group by tag, date_format(start_time, '%Y-%m-%d')
) a1
WINDOW w AS (PARTITION BY tag ORDER BY time DESC ROWS BETWEEN CURRENT ROW AND 6 FOLLOWING)
)a2
where time between '2021-10-01' and '2021-10-03'
group by tag,time
order by tag desc,time asc ;
解法2
select t2.*
from (select t1.tag,t1.d
,sum(t1.if_like_sum)over(partition by t1.tag order by t1.d rows 6 preceding)
,max(t1.if_retweet_sum)over(partition by t1.tag order by t1.d rows 6 preceding)
from (select tag,date(start_time) d
,sum(if_like) if_like_sum
,sum(if_retweet) if_retweet_sum
from tb_user_video_log tvl,tb_video_info tvi
where tvl.video_id=tvi.video_id
group by tag,d) as t1
) as t2
where t2.d between '2021-10-01' and '2021-10-03'
order by t2.tag desc,t2.d
涉及知识点
**1. partition by xx order by xxx rows 6 preceding **
按照xx进行分区 xxx进行排序,preceding代表当前order by的日期往前提6个,即提出一周的总和点赞量
2. window w as (同1)
总体开一个窗口,当我们需要用这个窗口的时候,over w 就可以进行开窗,最后采用相应的聚合函数就ok
3. DATE
select DATE('2020-11-11 11;12;13')
结果为2020-11-11
其有点类似与DATE_FORMAT(time,‘%Y-%m-%d’) 相同,都是提取完整时间戳的年月日
SQL161 近一个月发布的视频中热度最高的top3视频
描述
现有用户-视频互动表tb_user_video_log
id | uid | video_id | start_time | end_time | if_follow | if_like | if_retweet | comment_id |
---|---|---|---|---|---|---|---|---|
1 | 101 | 2001 | 2021-09-24 10:00:00 | 2021-09-24 10:00:30 | 1 | 1 | 1 | NULL |
2 | 101 | 2001 | 2021-10-01 10:00:00 | 2021-10-01 10:00:31 | 1 | 1 | 0 | NULL |
3 | 102 | 2001 | 2021-10-01 10:00:00 | 2021-10-01 10:00:35 | 0 | 0 | 1 | NULL |
4 | 103 | 2001 | 2021-10-03 11:00:50 | 2021-10-03 10:00:35 | 1 | 1 | 0 | 1732526 |
5 | 106 | 2002 | 2021-10-02 11:00:05 | 2021-10-02 11:01:04 | 2 | 0 | 1 | NULL |
6 | 107 | 2002 | 2021-10-02 10:59:05 | 2021-10-02 11:00:06 | 1 | 0 | 0 | NULL |
7 | 108 | 2002 | 2021-10-02 10:59:05 | 2021-10-02 11:00:05 | 1 | 1 | 1 | NULL |
8 | 109 | 2002 | 2021-10-03 10:59:05 | 2021-10-03 11:00:01 | 0 | 1 | 0 | NULL |
9 | 105 | 2002 | 2021-09-25 11:00:00 | 2021-09-25 11:00:30 | 1 | 0 | 1 | NULL |
10 | 101 | 2003 | 2021-09-26 11:00:00 | 2021-09-26 11:00:30 | 1 | 0 | 0 | NULL |
11 | 101 | 2003 | 2021-09-30 11:00:00 | 2021-09-30 11:00:30 | 1 | 1 | 0 | NULL |
(uid-用户ID, video_id-视频ID, start_time-开始观看时间, end_time-结束观看时间, if_follow-是否关注, if_like-是否点赞, if_retweet-是否转发, comment_id-评论ID)
短视频信息表tb_video_info
id | video_id | author | tag | duration | release_time |
---|---|---|---|---|---|
1 | 2001 | 901 | 旅游 | 30 | 2021-09-05 07:00:00 |
2 | 2002 | 901 | 旅游 | 60 | 2021-09-05 07:00:00 |
3 | 2003 | 902 | 影视 | 90 | 2021-09-05 07:00:00 |
4 | 2004 | 902 | 影视 | 90 | 2021-09-05 08:00:00 |
(video_id-视频ID, author-创作者ID, tag-类别标签, duration-视频时长, release_time-发布时间)
问题:找出近一个月发布的视频中热度最高的top3视频。
注:
- 热度=(a视频完播率+b点赞数+c评论数+d转发数)*新鲜度;
- 新鲜度=1/(最近无播放天数+1);
- 当前配置的参数a,b,c,d分别为100、5、3、2。
- 最近播放日期以end_time-结束观看时间为准,假设为T,则最近一个月按[T-29, T]闭区间统计。
- 结果中热度保留为整数,并按热度降序排序。
输出示例:
示例数据的输出结果如下
video_id | hot_index |
---|---|
2001 | 122 |
2002 | 56 |
2003 | 1 |
解释:
最近播放日期为2021-10-03,记作当天日期;近一个月(2021-09-04及之后)发布的视频有2001、2002、2003、2004,不过2004暂时还没有播放记录;
视频2001完播率1.0(被播放次数4次,完成播放4次),被点赞3次,评论1次,转发2次,最近无播放天数为0,因此热度为:(1001.0+53+31+22)/(0+1)=122
同理,视频2003完播率0,被点赞数1,评论和转发均为0,最近无播放天数为3,因此热度为:(1000+51+30+20)/(3+1)=1(1.2保留为整数)。
数据示例
输入:
DROP TABLE IF EXISTS tb_user_video_log, tb_video_info;
CREATE TABLE tb_user_video_log (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
uid INT NOT NULL COMMENT '用户ID',
video_id INT NOT NULL COMMENT '视频ID',
start_time datetime COMMENT '开始观看时间',
end_time datetime COMMENT '结束观看时间',
if_follow TINYINT COMMENT '是否关注',
if_like TINYINT COMMENT '是否点赞',
if_retweet TINYINT COMMENT '是否转发',
comment_id INT COMMENT '评论ID'
) CHARACTER SET utf8 COLLATE utf8_bin;
CREATE TABLE tb_video_info (
id INT PRIMARY KEY AUTO_INCREMENT COMMENT '自增ID',
video_id INT UNIQUE NOT NULL COMMENT '视频ID',
author INT NOT NULL COMMENT '创作者ID',
tag VARCHAR(16) NOT NULL COMMENT '类别标签',
duration INT NOT NULL COMMENT '视频时长(秒数)',
release_time datetime NOT NULL COMMENT '发布时间'
)CHARACTER SET utf8 COLLATE utf8_bin;
INSERT INTO tb_user_video_log(uid, video_id, start_time, end_time, if_follow, if_like, if_retweet, comment_id) VALUES
(101, 2001, '2021-09-24 10:00:00', '2021-09-24 10:00:30', 1, 1, 1, null)
,(101, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:31', 1, 1, 0, null)
,(102, 2001, '2021-10-01 10:00:00', '2021-10-01 10:00:35', 0, 0, 1, null)
,(103, 2001, '2021-10-03 11:00:50', '2021-10-03 11:01:35', 1, 1, 0, 1732526)
,(106, 2002, '2021-10-02 10:59:05', '2021-10-02 11:00:04', 2, 0, 1, null)
,(107, 2002, '2021-10-02 10:59:05', '2021-10-02 11:00:06', 1, 0, 0, null)
,(108, 2002, '2021-10-02 10:59:05', '2021-10-02 11:00:05', 1, 1, 1, null)
,(109, 2002, '2021-10-03 10:59:05', '2021-10-03 11:00:01', 0, 1, 0, null)
,(105, 2002, '2021-09-25 11:00:00', '2021-09-25 11:00:30', 1, 0, 1, null)
,(101, 2003, '2021-09-26 11:00:00', '2021-09-26 11:00:30', 1, 0, 0, null)
,(101, 2003, '2021-09-30 11:00:00', '2021-09-30 11:00:30', 1, 1, 0, null);
INSERT INTO tb_video_info(video_id, author, tag, duration, release_time) VALUES
(2001, 901, '旅游', 30, '2021-09-05 7:00:00')
,(2002, 901, '旅游', 60, '2021-09-05 7:00:00')
,(2003, 902, '影视', 90, '2021-09-05 7:00:00')
,(2004, 902, '影视', 90, '2021-09-05 8:00:00');
复制
输出:
2001|122
2002|56
2003|1
解法1
select
a1.video_id,
round((wanbo_rate*100+like_sum*5+comment_sum*3+retweet_sum*2)/(fresh+1),0) hot_index
from ( select t1.video_id,
sum(if(TIMESTAMPDIFF(SECOND,start_time,end_time)>=duration,1,0))/count(t1.video_id) wanbo_rate,
sum(t1.if_like) like_sum,
sum(if(t1.comment_id is not null,1,0)) comment_sum,
sum(t1.if_retweet) retweet_sum,
if(count(t1.video_id)=0,datediff(date((select max(end_time) from tb_user_video_log)),date(t2.release_time)),
datediff(date((select max(end_time) from tb_user_video_log)),max(date(t1.end_time)))) fresh
from test_1.tb_user_video_log t1
left join tb_video_info t2 on t1.video_id = t2.video_id
where datediff(date((select max(end_time) from tb_user_video_log)),date(t2.release_time)) <=29
group by t1.video_id) a1
order by hot_index desc
limit 3
解法2
SELECT video_id,
ROUND((100 * comp_play_rate + 5 * like_cnt + 3 * comment_cnt + 2 * retweet_cnt)
/ (TIMESTAMPDIFF(DAY, recently_end_date, cur_date) + 1), 0) as hot_index
FROM (
SELECT video_id,
AVG(IF(
TIMESTAMPDIFF(SECOND, start_time, end_time)>=duration, 1, 0
)) as comp_play_rate,
SUM(if_like) as like_cnt,
COUNT(comment_id) as comment_cnt,
SUM(if_retweet) as retweet_cnt,
MAX(DATE(end_time)) as recently_end_date, -- 最近被播放日期
MAX(DATE(release_time)) as release_date, -- 发布日期
MAX(cur_date) as cur_date -- 非分组列,加MAX避免语法错误
FROM tb_user_video_log
JOIN tb_video_info USING(video_id)
LEFT JOIN (
SELECT MAX(DATE(end_time)) as cur_date FROM tb_user_video_log
) as t_max_date ON 1
GROUP BY video_id
HAVING TIMESTAMPDIFF(DAY, release_date, cur_date) < 30
) as t_video_info
ORDER BY hot_index DESC
LIMIT 3;
涉及知识点
在 一开始对评论的统计数sum(),我采用的是!=null,但是并不对,得到的结果是119,上网查询后如下
在SQL中,NULL是一种特有的数据类型,其等价于没知有任何值、是未知数。NULL与0、空道字符串、空格都不同。SQL默认情况下对WHERE XX!= Null的判断会永远返回0行,却不会提示语法错误。内容
非ANSI SQL标准中data=NULL等同于data IS NULL,data<>NULL等同于data IS NOT NULL。
所以:默认情况下做比较条件时使用关键字“is null”和“is not null”。