开窗函数_ROW_NUMBER() / RANK() / DENSE_RANK() / NTILE() ------4个排名函数训练_1

排名函数（训练，其实从SQL2005时就已经被引入）

/*
SQL Server 2012从零开始学_7.8  排序函数
*/

--DROP TABLE fruits
GO
Create table fruits(
  s_id int,
  f_name char(20)
)

insert into fruits(s_id,f_name)
                values('','apple'),
                      ('','blackberry'),
                      ('','cherry'),
                      ('','orange'),
                      ('','banana'),
                      ('','grape'),
                      ('','coconut'),
                      ('','apricot'),
                      ('','berry'),
                      ('','lemon'),
                      ('','melon'),
                      ('','blackberry')

select * from fruits

查询结果：

-----------------------------------------------------------

--ROW_NUMBER()       --添加序号
select row_number() over (order by s_id asc) as rowid,s_id,f_name
from fruits
GO
-----------------------------------------------------------
--RANK()                     排序（对关联数据字段的值相同数据，则名次相同，隐形计算用了 sum(RankID)）
SELECT rank() over (order by s_id asc) as RankID,S_ID,f_name
FROM fruits;            ----注意知识点，到第1个RankID=5时，ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS cnt
-----------------------------------------------------------
--DENSE_RANK()             --排名，只是排名不跳顺序号(用递增号)
SELECT DENSE_RANK() OVER(ORDER BY s_id) as DENSEID,s_id,f_name
FROM fruits
-----------------------------------------------------------
--NTILE()
SELECT NTILE(5) OVER (ORDER BY s_id asc) as NTILEID,s_id,f_name
FROM fruits
-----------------------------------------------------------
SELECT NTILE(4) OVER (ORDER BY s_id ASC) AS NTILEID,s_id,f_name
FROM fruits;
-----------------------------------------------------------
--再将上面的组合到一起便于对比理解NTILE()

--函数是将有序分区中的行分发到指定个数的组中，各个组有编号，编号从1开始，类似'分区'一样 ，分为几个区，一个区会有多少个。

SELECT S_ID,f_name,
       row_number() over (order by s_id asc) as rowid,
       rank() over (order by s_id asc) as RankID,
       DENSE_RANK() OVER(ORDER BY s_id) as DENSEID,
       NTILE(5) OVER (ORDER BY s_id asc) as NTILEID,
       NTILE(4) OVER (ORDER BY s_id asc) as NTILEID1,      --分成4组，刚好能被12整除，每组3个，12条记录 = 4组 * 3条/组
       NTILE(6) OVER (ORDER BY s_id asc) as NTILEID2,      --分成6组，刚好能被12整除，每组2个，12条记录 = 6组 * 2条/组
       NTILE(7) OVER (ORDER BY s_id asc) as NTILEID3
FROM fruits;

      --红色虚线框中分别说明了函数的用法（将它们当做排序来理解，其次理解为排序后开窗）：
                      ROW_NUMBER()      ------>只是添加一个递增行序号
                      RANK()                      ------>排序(结合S_ID列来看)，并列的名次（在S_ID不相同时，跳号）
                      DENSE_RANK()         ------>排序(结合S_ID列来看)，并列的名次（在S_ID不相同时，不跳号）            
                      NTILE()                      ------>似乎理解，还不好讲解出来，但后面排序组要比前面的要小或相同               
                                                                 NTILEID1,  NTILEID2,  NTILEID3(只是试用了NTILE + ID作为了标题列头)

------------------------------------------------------------------------------------------------------------------------
下面是CTE、及窗口函数的示例
问题： 酒店每天要对在同一时间段内服务房间最多的一名服务生进行奖励。表 19-1 列出了一天中酒店服务生所服务的房间号和服务时间。

只要第 2 个房间的服务开始时间小于或等于第 1 个房间，并且结束时间大于第 1个房间的开始时间
                 AND W2.start_time <= W1.start_time
                 AND W2.end_time > W1.start_time

/*
学习本篇的目的： 窗口函数（学习前先来做这个例子）
*/

CREATE TABLE Waiters
(
  room_id int,
  waiter_name char(20),
  start_time datetime,
  end_time datetime
);
INSERT INTO Waiters VALUES
  (1,'张岚','2009-02-01 11:30','2009-02-01 13:30'),
  (2,'张岚','2009-02-01 11:40','2009-02-01 13:15'),
  (3,'孙静','2009-02-01 11:45','2009-02-01 14:20'),
  (4,'孙静','2009-02-01 11:39','2009-02-01 13:20'),
  (5,'孙静','2009-02-01 11:49','2009-02-01 14:16'),
  (6,'孙静','2009-02-01 10:37','2009-02-01 11:00'),
  (3,'孙静','2009-02-01 17:45','2009-02-01 18:20'),
  (4,'孙静','2009-02-01 17:39','2009-02-01 18:25'),
  (5,'孙静','2009-02-01 17:49','2009-02-01 18:36'),
  (6,'孙静','2009-02-01 17:37','2009-02-01 18:40'); 

---------------------------------------------------------
select datediff(dd,'2020-05-21','2020-05-12')
---------------------------------------------------------
SELECT MAX(room_id) FROM Waiters

select COUNT(*) as tally FROM Waiters
---------------------------------------------------------
SELECT * FROM Waiters
--显示服务房间最多的数量
select --W1.waiter_name,
       --W1.start_time,
       --W1.end_time,
       MAX(W1.room_id) as room_id 

from Waiters as W1
group by w1.waiter_name
order by w1.waiter_name
---------------------------------------------------------
--显示了服务房间的最大数量
select w1.waiter_name,
       --w1.start_time,
       --w1.end_time,
       MAX(W1.room_id) as room_id

from Waiters as w1
  inner join Waiters as w2
     on w1.waiter_name = w2.waiter_name
group by w1.waiter_name
order by w1.waiter_name
---------------------------------------------------------
--------------------------------------------------------
--显示每个人的服务的房间号、开始时间、结束时间
SELECT * FROM Waiters

select w1.waiter_name,
       w1.start_time,
       w1.end_time,
       MAX(W1.room_id) as room_id,
       COUNT(*) as tally
from Waiters as w1
  inner join Waiters as w2
     on w1.waiter_name = w2.waiter_name
     and w2.start_time <= w1.end_time    --->>>此处理解错误，应该是： and w2.start_time <= w1.start_time
     and w2.end_time > w1.start_time
group by w1.waiter_name,w1.start_time,w1.end_time,w1.room_id
order by w1.waiter_name,w1.start_time,room_id
---------------------------------------------------------
--
SELECT W1.waiter_name,
        W1.start_time,
        W1.end_time,
        MAX(W1.room_id) AS room_id,
        COUNT(*) AS tally
FROM Waiters AS W1
  INNER JOIN Waiters AS W2
    ON W1.waiter_name = W2.waiter_name
      AND W2.start_time <= W1.start_time
      AND W2.end_time > W1.start_time
GROUP BY W1.waiter_name, W1.start_time, W1.end_time, W1.room_id
ORDER BY W1.waiter_name, W1.start_time, room_id;         

---------------------------------------------------------
--得出同一时间服务房间数量最多的个数
select T1.waiter_name,MAX(T1.TALLY) as Tally
FROM
(SELECT W1.waiter_name,
        W1.start_time,
        W1.end_time,
        COUNT(*) AS tally
FROM Waiters AS W1
  INNER JOIN Waiters AS W2
    ON W1.waiter_name = W2.waiter_name
      AND W2.start_time <= W1.start_time
      AND W2.end_time > W1.start_time
GROUP BY W1.waiter_name, W1.start_time, W1.end_time) AS T1
GROUP BY T1.waiter_name;
---------------------------------------------------------
with T1(waiter_name,start_time,end_time,tally)
AS
(
    SELECT  W1.waiter_name,
            W1.start_time,
            W1.end_time,
            COUNT(*) AS tally
    FROM Waiters AS W1
      INNER JOIN Waiters AS W2
        ON W1.waiter_name = W2.waiter_name
          AND W2.start_time <= W1.start_time
          AND W2.end_time > W1.start_time
    GROUP BY W1.waiter_name, W1.start_time, W1.end_time
)
select waiter_name,MAX(tally)
from T1
group by T1.waiter_name;

---------------------------------------------------------
--使用CTE，这个是书本内容，自己独立写一次看看
WITH T1 (waiter_name, start_time, end_time, tally)   -- 定义 CTE 表达式的名称和列
AS
(
   SELECT W1.waiter_name,
             W1.start_time,
           W1.end_time,
            COUNT(*) AS tally
   FROM Waiters AS W1
     INNER JOIN Waiters AS W2
       ON W1.waiter_name = W2.waiter_name
         AND W2.start_time <= W1.start_time
         AND W2.end_time > W1.start_time
   GROUP BY W1.waiter_name, W1.start_time, W1.end_time
)
SELECT waiter_name, MAX(tally) AS tally
FROM T1
GROUP BY T1.waiter_name;
---------------------------------------------------------
--窗口函数使用前的预习
  select waiter_name,start_time as ts
  from Waiters

  select waiter_name,start_time as ts,+1 as type              --如果是开始时间，则 type设置 为  1
  from Waiters

  select waiter_name,end_time as ts,-1 as type                --如果是结束时间，则 type设置 为 -1
  from Waiters

  select *,
           sum(type) over(partition by waiter_name order by ts,type)
---------------------------------------------------------
--
WITH T1 AS
(
  SELECT waiter_name, start_time AS ts, +1 AS type
  FROM Waiters
  UNION ALL
  SELECT waiter_name, end_time, -1 AS type
  FROM Waiters
),
T2 AS
(
  SELECT *, SUM(type) OVER(PARTITION BY waiter_name
                                ORDER BY ts, type
                    ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS cnt
  FROM T1
)
SELECT * FROM T2 

---------------------------------------------------------

--窗口函数使用
with T1 as
(
  select waiter_name,start_time as ts,+1 as type
  from Waiters

  Union all

  select waiter_name,end_time as ts,-1 as type
  from Waiters
),
T2 as
(
    SELECT *, SUM(type) OVER(PARTITION BY waiter_name                             --sum(type)，相当于在求和
                                ORDER BY ts, type
                    ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS cnt      --此处cnt修改成另外一个别名便于理解： CountX
    FROM T1
)
select * from T2

运行结果：
    

---------------------------------------------------------

--最终完整的SQL

WITH T1 AS
(
  SELECT waiter_name, start_time AS ts, 1 AS type
  FROM Waiters
  UNION ALL
  SELECT waiter_name, end_time, -1 AS type
  FROM Waiters
),
T2 AS
(
  SELECT *, SUM(type) OVER(PARTITION BY waiter_name
                           ORDER BY ts, type
               ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW) AS cnt
  FROM T1
)
SELECT waiter_name, MAX(cnt)  AS tally
FROM T2
GROUP BY waiter_name; 

运行结果：
    waiter_name            tally
      孙静                    4
      张岚                    2

---------------------------------------------------------

---------------------------------------------------------

---------------------------------------------------------

参考书目《SQL Server 2012从零开始学》7.8  排序函数
                《锋利的SQL_第2版_张洪举_王晓文_著》