Hive分析窗口函数(二) NTILE,ROW_NUMBER,RANK,DENSE_RANK

本文中介绍前几个序列函数，NTILE,ROW_NUMBER,RANK,DENSE_RANK
注意： 序列函数不支持WINDOW子句。

数据准备：    

CREATE EXTERNAL TABLE lxw1234 (
    cookieid string,
    createtime string, --day
    pv INT
    ) ROW FORMAT DELIMITED
    FIELDS TERMINATED BY ','
    stored as textfile location '/tmp/lxw11/';
     
    DESC lxw1234;
    cookieid STRING
    createtime STRING
    pv INT
     
    hive> select * from lxw1234;
    OK
    cookie1 2015-04-10 1
    cookie1 2015-04-11 5
    cookie1 2015-04-12 7
    cookie1 2015-04-13 3
    cookie1 2015-04-14 2
    cookie1 2015-04-15 4
    cookie1 2015-04-16 4
    cookie2 2015-04-10 2
    cookie2 2015-04-11 3
    cookie2 2015-04-12 5
    cookie2 2015-04-13 6
    cookie2 2015-04-14 3
    cookie2 2015-04-15 9
    cookie2 2015-04-16 7

NTILE

NTILE(n)，用于将分组数据按照顺序切分成n片，返回当前切片值
NTILE不支持ROWS BETWEEN，比如 NTILE(2) OVER(PARTITION BY cookieid ORDER BY createtime ROWS BETWEEN 3 PRECEDING AND CURRENT ROW)
如果切片不均匀，默认增加第一个切片的分布

SELECT
    cookieid,
    createtime,
    pv,
    NTILE(2) OVER(PARTITION BY cookieid ORDER BY createtime) AS rn1,        --分组内将数据分成2片
    NTILE(3) OVER(PARTITION BY cookieid ORDER BY createtime) AS rn2, --分组内将数据分成3片
    NTILE(4) OVER(ORDER BY createtime) AS rn3 --将所有数据分成4片
    FROM lxw1234
    ORDER BY cookieid,createtime;
     
    cookieid day pv rn1 rn2 rn3
    -------------------------------------------------
    cookie1 2015-04-10 1 1 1 1
    cookie1 2015-04-11 5 1 1 1
    cookie1 2015-04-12 7 1 1 2
    cookie1 2015-04-13 3 1 2 2
    cookie1 2015-04-14 2 2 2 3
    cookie1 2015-04-15 4 2 3 3
    cookie1 2015-04-16 4 2 3 4
    cookie2 2015-04-10 2 1 1 1
    cookie2 2015-04-11 3 1 1 1
    cookie2 2015-04-12 5 1 1 2
    cookie2 2015-04-13 6 1 2 2
    cookie2 2015-04-14 3 2 2 3
    cookie2 2015-04-15 9 2 3 4
    cookie2 2015-04-16 7 2 3 4

–比如，统计一个cookie，pv数最多的前1/3的天

SELECT
    cookieid,
    createtime,
    pv,
    NTILE(3) OVER(PARTITION BY cookieid ORDER BY pv DESC) AS rn
    FROM lxw1234;
     
    --rn = 1 的记录，就是我们想要的结果
     
    cookieid day pv rn
    ----------------------------------
    cookie1 2015-04-12 7 1
    cookie1 2015-04-11 5 1
    cookie1 2015-04-15 4 1
    cookie1 2015-04-16 4 2
    cookie1 2015-04-13 3 2
    cookie1 2015-04-14 2 3
    cookie1 2015-04-10 1 3
    cookie2 2015-04-15 9 1
    cookie2 2015-04-16 7 1
    cookie2 2015-04-13 6 1
    cookie2 2015-04-12 5 2
    cookie2 2015-04-14 3 2
    cookie2 2015-04-11 3 3
    cookie2 2015-04-10 2 3

ROW_NUMBER

ROW_NUMBER() –从1开始，按照顺序，生成分组内记录的序列
–比如，按照pv降序排列，生成分组内每天的pv名次
ROW_NUMBER() 的应用场景非常多，再比如，获取分组内排序第一的记录;获取一个session中的第一条refer等。

SELECT
    cookieid,
    createtime,
    pv,
    ROW_NUMBER() OVER(PARTITION BY cookieid ORDER BY pv desc) AS rn
    FROM lxw1234;
     
    cookieid day pv rn
    -------------------------------------------
    cookie1 2015-04-12 7 1
    cookie1 2015-04-11 5 2
    cookie1 2015-04-15 4 3
    cookie1 2015-04-16 4 4
    cookie1 2015-04-13 3 5
    cookie1 2015-04-14 2 6
    cookie1 2015-04-10 1 7
    cookie2 2015-04-15 9 1
    cookie2 2015-04-16 7 2
    cookie2 2015-04-13 6 3
    cookie2 2015-04-12 5 4
    cookie2 2015-04-14 3 5
    cookie2 2015-04-11 3 6
    cookie2 2015-04-10 2 7

RANK 和 DENSE_RANK

—RANK() 生成数据项在分组中的排名，排名相等会在名次中留下空位
—DENSE_RANK() 生成数据项在分组中的排名，排名相等会在名次中不会留下空位   

SELECT
    cookieid,
    createtime,
    pv,
    RANK() OVER(PARTITION BY cookieid ORDER BY pv desc) AS rn1,
    DENSE_RANK() OVER(PARTITION BY cookieid ORDER BY pv desc) AS rn2,
    ROW_NUMBER() OVER(PARTITION BY cookieid ORDER BY pv DESC) AS rn3
    FROM lxw1234
    WHERE cookieid = 'cookie1';
     
    cookieid day pv rn1 rn2 rn3
    --------------------------------------------------
    cookie1 2015-04-12 7 1 1 1
    cookie1 2015-04-11 5 2 2 2
    cookie1 2015-04-15 4 3 3 3
    cookie1 2015-04-16 4 3 3 4
    cookie1 2015-04-13 3 5 4 5
    cookie1 2015-04-14 2 6 5 6
    cookie1 2015-04-10 1 7 6 7

    rn1: 15号和16号并列第3, 13号排第5
    rn2: 15号和16号并列第3, 13号排第4
    rn3: 如果相等，则按记录值排序，生成唯一的次序，如果所有记录值都相等，或许会随机排吧。