When to use GROUPING SETS, CUBE and ROLLUP
I have recently learned about GROUPING SETS, CUBE and ROLLUP for defining multiple grouping sets in sql server.
What I am asking is under what circumstances do we use these features ? What are the benefits and advantages of using them?
SELECT shipperid, YEAR(shippeddate) AS shipyear, COUNT(*) AS numorders FROM Sales.Orders GROUP BY GROUPING SETS ( ( shipperid, YEAR(shippeddate) ), ( shipperid ), ( YEAR(shippeddate) ), ( ) ); SELECT shipperid, YEAR(shippeddate) AS shipyear, COUNT(*) AS numorders FROM Sales.Orders GROUP BY CUBE( shipperid, YEAR(shippeddate) ); SELECT shipcountry, shipregion, shipcity, COUNT(*) AS numorders FROM Sales.Orders GROUP BY ROLLUP( shipcountry, shipregion, shipcity );
Firstly, for those who haven't already read up on the subject:
That being said, don't think about these grouping options as ways to get a result set. These are performance tools.
Let's take ROLLUP as a simple example.
I can use the following query to get the count of records for each value of GrpCol.
SELECT GrpCol, count(*) AS cnt FROM dbo.MyTable GROUP BY GrpCol
And I can use the following query to summarily "roll up" the count of ALL records.
SELECT NULL, count(*) AS cnt FROM dbo.MyTable
And I could UNION ALL the above two queries to get the exact same results I might get if I had written the first query with the ROLLUP clause (that's why I put the NULL in there).
It might actually be more convenient for me to execute this as two different queries because then I have the grouped results separate from my totals. Why would I want my final total mixed right in to the rest of those results? The answer is that doing both together using the ROLLUP clause is more efficient. SQL Server will use an execution plan that calculates all of the aggregations together in one pass. Compare that to the UNION ALL example which would provide the exact same results but use a less efficient execution plan (two table scans instead of one).
Imagine an extreme example in which you are working on a data set so large that each scan of the data takes one whole hour. You have to provide totals on basically every possible dimension (way to slice) that data every day. Aha! I bet one of these grouping options is exactly what you need. If you save off the results of that one scan into a special schema layout, you will then be able to run reports for the rest of the day off the saved results.
So I'm basically saying that you're working on a data warehouse project. For the rest of us it mostly falls into the "neat thing to know" category.
The CUBE is the same of GROUPING SETS with all possible combinations.
So this (using CUBE)
GROUP BY CUBE (C1, C2, C3, ..., Cn-2, Cn-1, Cn)
is the same of this (using GROUPING SETS)
GROUP BY GROUPING SETS ( (C1, C2, C3, ..., Cn-2, Cn-1, Cn) -- All dimensions are included. ,( , C2, C3, ..., Cn-2, Cn-1, Cn) -- n-1 dimensions are included. ,(C1, C3, ..., Cn-2, Cn-1, Cn) … ,(C1, C2, C3, ..., Cn-2, Cn-1,) ,(C3, ..., Cn-2, Cn-1, Cn) -- n-2 dimensions included ,(C1 ..., Cn-2, Cn-1, Cn) … ,(C1, C2) -- 2 dimensions are included. ,… ,(C1, Cn) ,… ,(Cn-1, Cn) ,… ,(C1) -- 1 dimension included ,(C2) ,… ,(Cn-1) ,(Cn) ,() ) -- Grand total, 0 dimension is included.
Then, if you don't really need all combinations, you should use GROUPING SETS rather than CUBE
ROLLUP and CUBE operators generate some of the same result sets and perform some of the same calculations as OLAP applications. The CUBE operator generates a result set that can be used for cross tabulation reports. A ROLLUP operation can calculate the equivalent of an OLAP dimension or hierarchy.
I think an example would help here. Suppose you have a table of number of UFOs sightings by country and gender, like bellow:
╔═════════╦═══════╦═════════╗ ║ COUNTRY ║ GENDER║ #SIGHTS ║ ╠═════════╬═══════╬═════════╣ ║ USA ║ F ║ 450 ║ ║ USA ║ M ║ 1500 ║ ║ ITALY ║ F ║ 704 ║ ║ ITALY ║ M ║ 720 ║ ║ SWEDEN ║ F ║ 317 ║ ║ SWEDEN ║ M ║ 310 ║ ║ BRAZIL ║ F ║ 144 ║ ║ BRAZIL ║ M ║ 159 ║ ╚═════════╩═══════╩═════════╝
Then, if you want to know the totals for each country, by gender and grand total only, you should use GROUPING SETS
select Country, Gender, sum(Number_Of_Sights) from Table1 group by GROUPING SETS((Country), (Gender), ()) order by Country, Gender
To get the same result with GROUP BY, you would use UNION ALL as:
select Country, NULL Gender, sum(Number_Of_Sights) from Table1 GROUP BY Country UNION ALL select NULL Country, Gender, sum(Number_Of_Sights) from Table1 GROUP BY GENDER UNION ALL SELECT NULL Country, NULL Gender, sum(Number_Of_Sights) FROM TABLE1 ORDER BY COUNTRY, GENDER
But it is not possible to obtain the same result with CUBE, since it will return all possibilities.
Now, if you want to know all possible combinations, then you should use CUBE
I find they are good when you're producing a report and the result is not something which can be rolled up within the client.
For example, if you're doing something with COUNT(DISTINCT...) then the result across a larger group is not necessarily the same value as the sum of the parts. Eg, across two individual days you might have 1500 visitors and 2000 visitors, but the total could be anywhere between 2000 and 3500, depending on the overlap. It's nice to do this in the client, but because the client can't tell what the overlap is, you can use GROUPING SETS to provide the answer (and then handle that extra row coming through in the report).