SQL Server Analysis functions and the ranking function

Original: SQL Server analysis functions and the ranking function

Packet-based analysis functions, calculating the aggregate value of the packet data, and a window function is often used in conjunction OVER (), using the analysis function can easily calculate the year and the chain, obtained median, maximum and minimum values obtained packet. Different analysis and aggregate functions, GROUP BY clause is not required, the result of the SELECT clause set by OVER clause packet ().

Insert sample data using the following script:

;with cte_data as 
(
select 'Document Control' as Department,'Arifin' as LastName,17.78 as Rate 
union all 
select 'Document Control','Norred',16.82 
union all 
select 'Document Control','Kharatishvili',16.82
union all 
select 'Document Control','Chai',10.25 
union all 
select 'Document Control','Berge',10.25 
union all 
select 'Information Services','Trenary',50.48
union all 
select 'Information Services','Conroy',39.66 
union all 
select 'Information Services','Ajenstat',38.46
union all 
select 'Information Services','Wilson',38.46
union all 
select 'Information Services','Sharma',32.45
union all 
select 'Information Services','Connelly',32.45
union all 
select 'Information Services','Berg',27.40
union all 
select 'Information Services','Meyyappan',27.40
union all 
select 'Information Services','Bacon',27.40
union all 
select 'Information Services','Bueno ',27.40
)
select Department
    ,LastName
    ,Rate
into #data
from cte_data
go

View Code

First, the analysis functions

Analysis function normally and OVER () function with the use, SQL Server There are four types of analysis functions.

Note: The order of execution is distinct clause after analysis functions.

1，CUME_DIST 和PERCENT_RANK

CUME_DIST logic calculation is: less than or equal to the value of the number of rows / the number of the current packet of rows

PERCENT_RANK calculation logic is a packet in the current row within :( RANK -1) / (total number of lines within packets -1) is the ranking value RANK () function result value sorting.

The following code for calculating cumulative distribution percentage and ranking:

select Department
    ,LastName
    ,Rate
    ,cume_dist() over(partition by Department order by Rate) as CumeDist
    ,percent_rank() over(partition by Department order by Rate) as PtcRank
    ,rank() over(partition by Department order by Rate asc) as rank_number
    ,count(0) over(partition by Department) as count_in_group
from #data
order by DepartMent
    ,Rate desc

2，PERCENTILE_CONT和PERCENTILE_DISC

PERCENTILE_CONT and PERCENTILE_DISC is to calculate percentile values, such as the calculation in a field when a percentile value is.

PERCENTILE_CONT ( numeric_literal )  WITHIN GROUP ( ORDER BY order_by_expression [ ASC | DESC ] ) OVER ( [ <partition_by_clause> ] )
PERCENTILE_DISC ( numeric_literal )  WITHIN GROUP ( ORDER BY order_by_expression [ ASC | DESC ] ) OVER ( [ <partition_by_clause> ] )

The difference between these two functions is that the former is continuous, which is discrete. CONT representative of continuous, successive values, DISC representative of discrete, discrete values. PERCENTILE_CONT is continuous, meaning it is considered interval, the absolute value of intermediate values; and PERCENTILE_DISC is discrete, so it is considered more choice up or down, and does not consider interval.

The following script is used to obtain quantile:

select Department
    ,LastName
    ,Rate
    ,PERCENTILE_CONT(0.5) WITHIN GROUP (ORDER BY Rate) OVER (PARTITION BY Department) AS MedianCont
    ,PERCENTILE_DISC(0.5) WITHIN GROUP (ORDER BY Rate) OVER (PARTITION BY Department) AS MedianDisc
    ,row_number() over(partition by Department order by Rate) as rn
from #data
order by DepartMent
    ,Rate asc

3, LAG and LEAD

In a query, for the same sort fields, Lag function for acquiring the first N rows in the same group, the Lead function for acquiring the N rows in the same group,

LAG (scalar_expression [,offset] [,default])
    OVER ( [ partition_by_clause ] order_by_clause )
LEAD ( scalar_expression [ ,offset ] , [ default ] ) 
    OVER ( [ partition_by_clause ] order_by_clause )

Results date, these two functions are particularly suitable for calculating the year and the chain.

select DepartMent
    ,LastName
    ,Rate
    ,lag(Rate,1,0) over(partition by Department order by LastName) as LastRate
    ,lead(Rate,1,0) over(partition by Department order by LastName) as NextRate
from #data
order by Department
    ,LastName

4，FIRST_VALUE和LAST_VALUE

获取分组内的最大值和最小值，分组内的最大值和最小值是唯一的。

LAST_VALUE ( [scalar_expression ) OVER ( [ partition_by_clause ] order_by_clause rows_range_clause )
FIRST_VALUE ( [scalar_expression ] ) OVER ( [ partition_by_clause ] order_by_clause [ rows_range_clause ] )

二，排名函数

SQL Server的排名函数是对查询的结果进行排名和分组，TSQL共有4个排名函数，分别是：RANK、NTILE、DENSE_RANK和ROW_NUMBER，和OVER()函数搭配使用，按照特定的顺序排名。

1，ROW_NUMBER函数

ROW_NUMBER函数实际上是一个序列，每个分组内都会创建一个序列，序列从1开始，按照顺序依次 +1 递增。

ROW_NUMBER ( ) 
    OVER ( [ PARTITION_BY_clause ] order_by_clause )

分组内序列的最大值就是该分组内的行的数目。

2，RANK函数

RANK函数用于排名时，不会返回连续的整数。RANK函数的语法是：在分组内，按照特定的顺序排名，序号从1依次递增，排名函数以tie为单位，每个tie中的所有行的排名是相同的，排名可能是不连续的。

RANK ( ) OVER ( [ partition_by_clause ] order_by_clause )

排名的算法是：

step1：按照指定的分区字段分组，在每个分组内按照指定的字段排序。
step2：在每个分组内，如果相邻的两行或多行相同在排序字段上的值相同，那么这些行称作一个tie，每个tie中的所有行都会获得相同的排名。
step3：后面的排名会计算每个tie中的行数，RANK函数不总是返回连续的整数，例如，班级中，A，B分数都是100分，C的分数是90分，那么A和B的排名是1，C的排名是3。

3，DENSE_RANK

DENSE_RANK函数用于排名时，会返回连续的整数。每个tie占用一个排名，每个tie中的所有行的排名是相同的。排名值是连续的

DENSE_RANK ( ) OVER ( [ <partition_by_clause> ] < order_by_clause > )

排名的算法是：

step1：按照指定的分区字段分组，在每个分组内按照指定的字段排序。
step2：在每个分组内，如果相邻的两行或多行相同在排序字段上的值相同，那么这些行称作一个tie，每个tie中的所有行都会获得相同的排名。
step3：后面的排名会计算每个tie中的行数，RANK函数总是返回连续的整数，例如，班级中，A，B分数都是100分，C的分数是90分，那么A和B的排名是1，C的排名是2。

4，NTILE

在每个分组中，NTILE按照指定的顺序，把数据行分为N个小组（tile），NTILE返回小组编号。在每个分组内，具有相同的小组编号的数据行，位于同一个小组。注意：小组的编号是按照行数，而不是按照列值。在同一分组内，存在两行的列值相同，而小组编号不同。

NTILE (integer_expression) OVER ( [ <partition_by_clause> ] < order_by_clause > )

如果分区中的行数不能被integer_expression整除，那么会导致小组相差一个成员：较大的小组按OVER子句指定的顺序位于较小的小组之前。例如，如果把8行分为3个小组，前2个小组有3行，后一个小组有2行。

如果分区中的中行数能被integer_expression整除，那么每个小组具有相同的行数。

特别地，NTILE(4) 把一个分组分成4份，叫做Quartile。例如，以下脚本显示各个排名函数的执行结果：

select Department
    ,LastName
    ,Rate
    ,row_number() over(order by Rate) as [row number]
    ,rank() over(order by rate) as rate_rank
    ,dense_rank() over(order by rate) as rate_dense_rank
    ,ntile(4) over(order by rate) as quartile_by_rate
from #data

Reference documents:

Analytic Functions (Transact-SQL)

Ranking Functions (Transact-SQL)