CUBE Extension v13
The CUBE extension is similar to the ROLLUP extension. However, unlike ROLLUP, which produces groupings and results in a hierarchy based on a left to right listing of items in the ROLLUP expression list, a CUBE produces groupings and subtotals based on every permutation of all items in the CUBE expression list. Thus, the result set contains more rows than a ROLLUP performed on the same expression list.
The syntax for a single CUBE is as follows:
CUBE ( { <expr_1> | ( <expr_1a> [, <expr_1b> ] ...) }
[, <expr_2> | ( <expr_2a> [, <expr_2b> ] ...) ] ...)Each expr is an expression that determines the grouping of the result set. If enclosed within parenthesis as ( expr_1a, expr_1b, ...) then the combination of values returned by expr_1a and expr_1b defines a single group.
The base level of aggregates returned in the result set is for each unique combination of values returned by the expression list.
In addition, a subtotal is returned for the first item in the list (expr_1 or the combination of ( expr_1a, expr_1b, ...), whichever is specified) for each unique value. A subtotal is returned for the second item in the list (expr_2 or the combination of ( expr_2a, expr_2b, ...), whichever is specified) for each unique value. A subtotal is also returned for each unique combination of the first item and the second item. Similarly, if there is a third item, a subtotal is returned for each unique value of the third item, each unique value of the third item and first item combination, each unique value of the third item and second item combination, and each unique value of the third item, second item, and first item combination. Finally a grand total is returned for the entire result set.
For the subtotal rows, null is returned for the items across which the subtotal is taken.
The CUBE extension specified within the context of the GROUP BY clause is shown by the following:
SELECT <select_list> FROM ... GROUP BY [... ,] CUBE ( <expression_list> ) [, ...]
The items specified in select_list must also appear in the CUBE expression_list; or they must be aggregate functions such as COUNT, SUM, AVG, MIN, or MAX; or they must be constants or functions whose return values are independent of the individual rows in the group (for example, the SYSDATE function).
The GROUP BY clause may specify multiple CUBE extensions as well as multiple occurrences of other GROUP BY extensions and individual expressions.
The ORDER BY clause should be used if you want the output to display in a meaningful structure. There is no guarantee on the order of the result set if no ORDER BY clause is specified.
The number of grouping levels or totals is 2 raised to the power of n where n represents the number of items in the CUBE expression list. A parenthesized list counts as one item.
The following query produces a cube based on permutations of columns loc, dname, and job.
SELECT loc, dname, job, COUNT(*) AS "employees" FROM emp e, dept d WHERE e.deptno = d.deptno GROUP BY CUBE (loc, dname, job) ORDER BY 1, 2, 3;
The following is the result of the query. There is a count of the number of employees for each combination of loc, dname, and job, as well as subtotals for each combination of loc and dname, for each combination of loc and job, for each combination of dname and job, for each unique value of loc, for each unique value of dname, for each unique value of job, and a grand total displayed on the last line.
loc | dname | job | employees
---------+------------+-----------+-----------
BOSTON | OPERATIONS | ANALYST | 1
BOSTON | OPERATIONS | CLERK | 1
BOSTON | OPERATIONS | MANAGER | 1
BOSTON | OPERATIONS | | 3
BOSTON | RESEARCH | ANALYST | 2
BOSTON | RESEARCH | CLERK | 2
BOSTON | RESEARCH | MANAGER | 1
BOSTON | RESEARCH | | 5
BOSTON | | ANALYST | 3
BOSTON | | CLERK | 3
BOSTON | | MANAGER | 2
BOSTON | | | 8
CHICAGO | SALES | CLERK | 1
CHICAGO | SALES | MANAGER | 1
CHICAGO | SALES | SALESMAN | 4
CHICAGO | SALES | | 6
CHICAGO | | CLERK | 1
CHICAGO | | MANAGER | 1
CHICAGO | | SALESMAN | 4
CHICAGO | | | 6
NEW YORK | ACCOUNTING | CLERK | 1
NEW YORK | ACCOUNTING | MANAGER | 1
NEW YORK | ACCOUNTING | PRESIDENT | 1
NEW YORK | ACCOUNTING | | 3
NEW YORK | | CLERK | 1
NEW YORK | | MANAGER | 1
NEW YORK | | PRESIDENT | 1
NEW YORK | | | 3
| ACCOUNTING | CLERK | 1
| ACCOUNTING | MANAGER | 1
| ACCOUNTING | PRESIDENT | 1
| ACCOUNTING | | 3
| OPERATIONS | ANALYST | 1
| OPERATIONS | CLERK | 1
| OPERATIONS | MANAGER | 1
| OPERATIONS | | 3
| RESEARCH | ANALYST | 2
| RESEARCH | CLERK | 2
| RESEARCH | MANAGER | 1
| RESEARCH | | 5
| SALES | CLERK | 1
| SALES | MANAGER | 1
| SALES | SALESMAN | 4
| SALES | | 6
| | ANALYST | 3
| | CLERK | 5
| | MANAGER | 4
| | PRESIDENT | 1
| | SALESMAN | 4
| | | 17
(50 rows)The following query shows the effect of combining items in the CUBE list within parenthesis.
SELECT loc, dname, job, COUNT(*) AS "employees" FROM emp e, dept d WHERE e.deptno = d.deptno GROUP BY CUBE (loc, (dname, job)) ORDER BY 1, 2, 3;
In the output note that there are no subtotals for permutations involving loc and dname combinations, loc and job combinations, or for dname by itself, or for job by itself.
loc | dname | job | employees
---------+------------+-----------+-----------
BOSTON | OPERATIONS | ANALYST | 1
BOSTON | OPERATIONS | CLERK | 1
BOSTON | OPERATIONS | MANAGER | 1
BOSTON | RESEARCH | ANALYST | 2
BOSTON | RESEARCH | CLERK | 2
BOSTON | RESEARCH | MANAGER | 1
BOSTON | | | 8
CHICAGO | SALES | CLERK | 1
CHICAGO | SALES | MANAGER | 1
CHICAGO | SALES | SALESMAN | 4
CHICAGO | | | 6
NEW YORK | ACCOUNTING | CLERK | 1
NEW YORK | ACCOUNTING | MANAGER | 1
NEW YORK | ACCOUNTING | PRESIDENT | 1
NEW YORK | | | 3
| ACCOUNTING | CLERK | 1
| ACCOUNTING | MANAGER | 1
| ACCOUNTING | PRESIDENT | 1
| OPERATIONS | ANALYST | 1
| OPERATIONS | CLERK | 1
| OPERATIONS | MANAGER | 1
| RESEARCH | ANALYST | 2
| RESEARCH | CLERK | 2
| RESEARCH | MANAGER | 1
| SALES | CLERK | 1
| SALES | MANAGER | 1
| SALES | SALESMAN | 4
| | | 17
(28 rows)The following query shows another variation whereby the first expression is specified outside of the CUBE extension.
SELECT loc, dname, job, COUNT(*) AS "employees" FROM emp e, dept d WHERE e.deptno = d.deptno GROUP BY loc, CUBE (dname, job) ORDER BY 1, 2, 3;
In this output, the permutations are performed for dname and job within each grouping of loc.
loc | dname | job | employees ---------+------------+-----------+----------- BOSTON | OPERATIONS | ANALYST | 1 BOSTON | OPERATIONS | CLERK | 1 BOSTON | OPERATIONS | MANAGER | 1 BOSTON | OPERATIONS | | 3 BOSTON | RESEARCH | ANALYST | 2 BOSTON | RESEARCH | CLERK | 2 BOSTON | RESEARCH | MANAGER | 1 BOSTON | RESEARCH | | 5 BOSTON | | ANALYST | 3 BOSTON | | CLERK | 3 BOSTON | | MANAGER | 2 BOSTON | | | 8 CHICAGO | SALES | CLERK | 1 CHICAGO | SALES | MANAGER | 1 CHICAGO | SALES | SALESMAN | 4 CHICAGO | SALES | | 6 CHICAGO | | CLERK | 1 CHICAGO | | MANAGER | 1 CHICAGO | | SALESMAN | 4 CHICAGO | | | 6 NEW YORK | ACCOUNTING | CLERK | 1 NEW YORK | ACCOUNTING | MANAGER | 1 NEW YORK | ACCOUNTING | PRESIDENT | 1 NEW YORK | ACCOUNTING | | 3 NEW YORK | | CLERK | 1 NEW YORK | | MANAGER | 1 NEW YORK | | PRESIDENT | 1 NEW YORK | | | 3 (28 rows)