PLSQL==>分析函数

1.创建示例数据：

drop table sales_fact;
CREATE table sales_fact AS 
SELECT  country_name country,country_subRegion region, prod_name product, calendar_year year, calendar_week_number week, 
SUM(amount_sold) sale, 
sum(amount_sold*
  ( case   
         when mod(rownum, 10)=0 then 1.4
         when mod(rownum, 5)=0 then 0.6
         when mod(rownum, 2)=0 then 0.9
         when mod(rownum,2)=1 then 1.2
         else 1
    end )) receipts
FROM sales, times, customers, countries, products
WHERE sales.time_id = times.time_id AND
sales.prod_id = products.prod_id AND
sales.cust_id = customers.cust_id AND
customers.country_id = countries.country_id
GROUP BY 
country_name,country_subRegion, prod_name, calendar_year, calendar_week_number; 

分析函数具有3个基本组成部分：分区子句，排序子句和开窗子句。

rows BETWEEN unbounded preceding AND CURRENT ROW开窗子句：

SELECT YEAR,
       week,
       sale,
       SUM(sale) over(PARTITION BY product, country, region, YEAR ORDER BY week rows BETWEEN unbounded preceding AND CURRENT ROW) running_sum_ytd
  FROM sales_fact
 WHERE country IN ('Australia')
  7     AND product = 'Xtend Memory'
  8   ORDER BY product, country, YEAR, week;


      YEAR WEEK   SALE RUNNING_SUM_YTD
---------- ---------- ---------- ---------------
      1998    1  58.15   58.15
      1998    2  29.39   87.54
      1998    3  29.49  117.03
      1998    4  29.49  146.52
      1998    5   29.8  176.32
      1998    6  58.78   235.1
      1998    9  58.78  293.88
      1998   10 117.76  411.64
      1998   12   59.6  471.24
      1998   14  58.78  530.02
      1998   15  58.78   588.8

求出sale列的最大值：

SELECT YEAR,
       week,
       sale,
       MAX(sale) over(PARTITION BY product, country, region, YEAR ORDER BY week rows BETWEEN unbounded preceding AND unbounded following) max_sale
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
  8   ORDER BY product, country, YEAR, week;


      YEAR WEEK   SALE   MAX_SALE
---------- ---------- ---------- ----------
      1998    1  58.15     172.56
      1998    2  29.39     172.56
      1998    3  29.49     172.56
      1998    4  29.49     172.56
      1998    5   29.8     172.56
      1998    6  58.78     172.56
      1998    9  58.78     172.56
      1998   10 117.76     172.56
      1998   12   59.6     172.56
      1998   14  58.78     172.56
      1998   15  58.78     172.56

细粒度窗口声明

计算2周前和本周及后2周的情况：

SELECT YEAR,
       week,
       sale,
       MAX(sale) over(PARTITION BY product, country, region, YEAR ORDER BY week rows BETWEEN unbounded preceding AND unbounded following) max_sale
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
  8   ORDER BY product, country, YEAR, week;


      YEAR WEEK   SALE   MAX_SALE
---------- ---------- ---------- ----------
      1998    1  58.15     172.56
      1998    2  29.39     172.56
      1998    3  29.49     172.56
      1998    4  29.49     172.56
      1998    5   29.8     172.56
      1998    6  58.78     172.56
      1998    9  58.78     172.56
      1998   10 117.76     172.56
      1998   12   59.6     172.56
      1998   14  58.78     172.56
      1998   15  58.78     172.56

默认窗口的声明子句是rows between unbounded preceding and current now.

LEAD和LAG

LAG和LEAD 函数提供了跨行引用的能力。LAG提供了访问结果集中前面的的行的能力，Lead函数允许访问结果集后面的行。

SELECT YEAR,
       week,
       sale,
       MAX(sale) over(PARTITION BY product, country, region, YEAR ORDER BY week rows BETWEEN unbounded preceding AND unbounded following) max_sale
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
  8   ORDER BY product, country, YEAR, week;


      YEAR WEEK   SALE   MAX_SALE
---------- ---------- ---------- ----------
      1998    1  58.15     172.56
      1998    2  29.39     172.56
      1998    3  29.49     172.56
      1998    4  29.49     172.56
      1998    5   29.8     172.56
      1998    6  58.78     172.56
      1998    9  58.78     172.56
      1998   10 117.76     172.56
      1998   12   59.6     172.56
      1998   14  58.78     172.56
      1998   15  58.78     172.56

LEAD

SELECT YEAR,
       week,
       sale,
       lead(sale, 1, sale) over(PARTITION BY product, country, region ORDER BY YEAR, week) prior_wk_sales
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
  8   ORDER BY product, country, YEAR, week;

FIRST_VALUE 和LAST_VALUE:

SELECT YEAR,
       week,
       sale,
       first_value(sale) over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC rows BETWEEN unbounded preceding AND unbounded following) top_sale_value,
       first_value(week) over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC rows BETWEEN unbounded preceding AND unbounded following) top_week_sale
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
  9   ORDER BY product, country, YEAR, week;


      YEAR WEEK   SALE TOP_SALE_VALUE TOP_WEEK_SALE
---------- ---------- ---------- -------------- -------------
      1998    1  58.15 172.56   48
      1998    2  29.39 172.56   48
      1998    3  29.49 172.56   48
      1998    4  29.49 172.56   48
      1998    5   29.8 172.56   48
      1998    6  58.78 172.56   48
      1998    9  58.78 172.56   48
      1998   10 117.76 172.56   48
      1998   12   59.6 172.56   48
      1998   14  58.78 172.56   48
      1998   15  58.78 172.56   48

SELECT YEAR,
       week,
       sale,
       last_value(sale) over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC rows BETWEEN unbounded preceding AND unbounded following) low_value
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
  8   ORDER BY product, country, YEAR, week;


      YEAR WEEK   SALE  LOW_VALUE
---------- ---------- ---------- ----------
      1998    1  58.15      28.76
      1998    2  29.39      28.76
      1998    3  29.49      28.76
      1998    4  29.49      28.76
      1998    5   29.8      28.76
      1998    6  58.78      28.76
      1998    9  58.78      28.76
      1998   10 117.76      28.76
      1998   12   59.6      28.76
      1998   14  58.78      28.76
      1998   15  58.78      28.76

SELECT YEAR,
       week,
       sale,
       nth_value(sale, 2) over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC rows BETWEEN unbounded preceding AND unbounded following) top_2nd_value
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
  8   ORDER BY product, country, YEAR, week;


      YEAR WEEK   SALE TOP_2ND_VALUE
---------- ---------- ---------- -------------
      1998    1  58.15 117.76
      1998    2  29.39 117.76
      1998    3  29.49 117.76
      1998    4  29.49 117.76
      1998    5   29.8 117.76
      1998    6  58.78 117.76
      1998    9  58.78 117.76
      1998   10 117.76 117.76
      1998   12   59.6 117.76
      1998   14  58.78 117.76
      1998   15  58.78 117.76

RANK函数：

求出销售前10名的周：

rank跳过并列排名的值：

SELECT *
  FROM (SELECT YEAR,
               week,
               sale,
               rank() over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC) sales_rank
          FROM sales_fact
         WHERE country IN ('Australia')
           AND product = 'Xtend Memory'
         ORDER BY product, country, YEAR, week)
 WHERE sales_rank <= 10
 11   ORDER BY 1, 4;


      YEAR WEEK   SALE SALES_RANK
---------- ---------- ---------- ----------
      1998   48 172.56  1
      1998   10 117.76  2
      1998   18 117.56  3
      1998   23 117.56  3
      1998   26 117.56  3
      1998   38 115.84  6
      1998   42 115.84  6
      1998   39 115.84  6
      1998   34 115.44  9
      1998   52  86.38 10
      1999   17 148.12  1

DENSE_RANK:不会跳过并列排名的值：

SELECT *
  FROM (SELECT YEAR,
               week,
               sale,
               dense_rank() over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC) sales_rank
          FROM sales_fact
         WHERE country IN ('Australia')
           AND product = 'Xtend Memory'
         ORDER BY product, country, YEAR, week)
 WHERE sales_rank <= 10
 11   ORDER BY 1, 4;


      YEAR WEEK   SALE SALES_RANK
---------- ---------- ---------- ----------
      1998   48 172.56  1
      1998   10 117.76  2
      1998   18 117.56  3
      1998   23 117.56  3
      1998   26 117.56  3
      1998   38 115.84  4
      1998   39 115.84  4
      1998   42 115.84  4
      1998   34 115.44  5
      1998   52  86.38  6
      1998   21   59.6  7

ROW_NUMBER:

SELECT YEAR,
       week,
       sale,
       row_number() over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC) sales_rn,
       rank() over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC) sales_rank
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
 ORDER BY product, country, YEAR, sales_rank;

RATIO_TO_REPORT:

SQL> set linesize 100
SELECT YEAR,
       week,
       sale,
       trunc(100 * ratio_to_report(sale)
             over(PARTITION BY product, country, region, YEAR),
             2) sales_yr,
       trunc(100 * ratio_to_report(sale)
             over(PARTITION BY product, country, region),
             2) sales_grod
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
 13   ORDER BY product, country, YEAR, week;


      YEAR WEEK    SALE   SALES_YR SALES_GROD
---------- ---------- ---------- ---------- ----------
      1998     1   58.15       2.26   .43
      1998    2   29.39       1.14   .21
      1998    3   29.49       1.15   .22
      1998   4   29.49       1.15   .22
      1998     5    29.8       1.16   .22
      1998    6   58.78       2.29   .43
      1998    9   58.78       2.29   .43
      1998   10  117.76       4.59   .88
      1998   12    59.6       2.32   .44
      1998   14   58.78       2.29   .43
      1998   15   58.78       2.29   .43

PERCENT_RANK:

SELECT *
  FROM (SELECT YEAR,
               week,
               sale,
               100 * percent_rank() over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC) pr
          FROM sales_fact
         WHERE country IN ('Australia')
           AND product = 'Xtend Memory')
 WHERE pr < 50
 ORDER BY YEAR, sale DESC;

STDDV,计算标准偏差：

SELECT YEAR,
       week,
       sale,
       STDDEV(sale) over(PARTITION BY product, country, region, YEAR ORDER BY sale DESC rows BETWEEN unbounded preceding AND unbounded following) stddv
  FROM sales_fact
 WHERE country IN ('Australia')
   AND product = 'Xtend Memory'
 ORDER BY YEAR, week;

LISTAGG函数：（把国家转换成由逗号分隔的列表）。

SELECT listagg(country, ',') within GROUP(ORDER BY country DESC)
  FROM (SELECT DISTINCT country FROM sales_fact ORDER BY country);