Analytic Functions in Oracle (10g)

Oracle Analytic Functions

Analyctic functions in Oracle are set of functions and clauses used for arriving statistical informations like sum, average. These functions are more flexible to use and efficient.

Examples: Calculate a running total and moving average, Top-N queries

Earlier, these problems were solved using PL/SQL. But wehn we look for better performance, PL/SQL does not benefit much. Analytical functions provides more benifits with no questions on performance.Though this feature has been intrduced from Oracle 8, many developers does not aware of this. Especially for a pivot table query most of the developers still going for the old approach.

Key benefits

• Easier to code
• Faster than a equivalent SQL or PL/SQL
• No need to test or tune for performance as in the case of custom SQL/PL-SQL

Syntax

<Analytic-Function>(parameter1,parameter2,...)

OVER (

<Partition By column1, column2, .. [Order By column1, column2, .. [ASC/DESC] [NULLS FIRST/LAST] ] >

Order By column1, column2, .. [ASC/DESC] [NULLS FIRST/LAST]

[ROWS [UNBOUNDED] v_rows PRECEDING/FOLLOWING] / [RANGE [UNBOUNDED] v_range PRECEDING/FOLLOWING]

)

Note: The ROWS/RANGE keywords are called windowing clauses/windowing functions.

List of Analytic-Functions

AVG, CORR, COVAR_POP, COVAR_SAMP, COUNT, CUME_DIST, DENSE_RANK, FIRST, FIRST_VALUE, LAG, LAST, LAST_VALUE, LEAD, MAX, MIN, NTILE, PERCENT_RANK, PERCENTILE_CONT, PERCENTILE_DISC, RANK, RATIO_TO_REPORT, STDDEV, STDDEV_POP, STDDEV_SAMP, SUM, VAR_POP, VAR_SAMP, VARIANCE.

Below query pulls out top-3 ranked sales figures department wise. Query uses a partition by clause over DENSE_RANK function.

SELECT * FROM

(SELECT sale_date, dept_code, sales_amount,

DENSE_RANK() OVER(PARTITION BY DEPT_CODE ORDER BY SALES_AMOUNT DESC) sale_rank

FROM sales_summary_daily)

WHERE sale_rank <=3;

Below is the resuts of the query. 

SALE_DATE DEPT_CODE 	SALES_AMOUNT 	SALE_RANK 

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

28-OCT-03 East 		99998.63	1 

19-SEP-95 East 		99996.64 	2 

08-OCT-85 East 		99942.49 	3 

06-NOV-98 NE 		49995.44 	1 

15-FEB-06 NE 		49991.96 	2 

28-DEC-98 NE 		49988.92 	3 

09-FEB-05 North 	99983.69 	1 

24-DEC-86 North 	99980.99 	2 

22-OCT-02 North 	99973.49 	3 

16-APR-82 SE 		49998.3 	1 

27-JAN-95 SE 		49996.16 	2

14-SEP-83 SE 		49990.52 	3

27-JUL-96 South 	69996.2 	1

15-JAN-06 South 	69993.7 	2

17-AUG-97 South 	69991.29 	3

10-OCT-96 West 		9999.91 	1

23-AUG-94 West 		9998.76 	2

06-SEP-02 West 		9998.51 	3

Above query pulls top-3 sales figures for all sales region 

using Oracle analytic functions. Deriving this using 

conventional SQL/PLSQL will be a night mare.

SELECT SALE_MONTH, 

MAX(DECODE(sale_rank, 1, dept_code, null)) First,

MAX(DECODE(sale_rank, 2, dept_code, null)) Second,

MAX(DECODE(sale_rank, 3, dept_code, null)) Third

FROM

(SELECT LAST_DAY(SALE_DATE) SALE_MONTH, dept_code, sales_amount,

DENSE_RANK() OVER(PARTITION BY LAST_DAY(SALE_DATE)

ORDER BY SALES_AMOUNT DESC) sale_rank

FROM sales_summary_daily

WHERE SALE_DATE >= '01-JAN-2008'

)

WHERE SALE_RANK <= 3

GROUP BY SALE_MONTH

SALE_MONTH 	FIRST SECOND THIRD 

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

31-JAN-08 	North North  North 

29-FEB-08 	North East   North 

30-APR-08 	East  North  East 

30-JUN-08 	North North  North 

31-JUL-08 	East  North  North 

31-AUG-08 	North East   North 

31-MAR-08 	North North  East 

31-MAY-08 	North North  East 

30-SEP-08 	East  North  North 

Again using conventional SQL/PLSQL needs more effort to make it efficient. With Oracle 11g, this is even more simplified using the pivot operator.