Is it possible to use Aggregate function in a Select statment without using Group By clause?

Question

So far I have written Aggregate function followed by Group By clause to find the values based on SUM, AVG and other Aggregate functions. I have a bit confusion in the Group By clause. When we use Aggregate functions what are the columns I need to specify in the Group By clause. Otherwise Is there any way to use Aggregate functions without using Group By clause.

Solution 1

All columns in the SELECT clause that do not have an aggregate need to be in the GROUP BY

Good:

SELECT col1, col2, col3, MAX(col4)
...
GROUP BY col1, col2, col3

Also good:

SELECT col1, col2, col3, MAX(col4)
...
GROUP BY col1, col2, col3, col5, col6

No other columns = no GROUP BY needed

SELECT MAX(col4)
...

Won't work:

SELECT col1, col2, col3, MAX(col4)
...
GROUP BY col1, col2

Pointless:

SELECT col1, col2, col3, MAX(col4)
...
GROUP BY col1, col2, col3, MAX(col4)

Having an aggregate (MAX etc) with other columns without a GROUP BY makes no sense because the query becomes ambiguous.

Solution 2

You can use Select AGG() OVER() in TSQL

SELECT *,
SUM(Value) OVER()
FROM Table

There are other options for Over such as Partition By if you want to group:

SELECT *,
SUM(Value) OVER(PARTITION By ParentId)
FROM Table

http://msdn.microsoft.com/en-us/library/ms189461.aspx

Solution 3

Yes you can use an aggregate without GROUP BY:

SELECT SUM(col) FROM tbl;

This will return one row only - the sum of the column "col" for all rows in tbl (excluding nulls).

Solution 4

The Columns which are not present in the Aggregate function should come on group by clause:

Select 
Min(col1), 
Avg(col2), 
sum(col3) 
from table

then we do not required group by clause, But if there is some column which are not present in the Aggregate function then you must use group by for that column.

Select 
col1, 
col2, 
sum(col3) 
from  table 
group by col1,col2

then we have to use the group by for the column col1 and col2

Solution 5

You must group by columns that do not have aggregate functions on them.

You may avoid a group by clause if all columns selected have aggregate functions applied.

View more solutions

Comments

[thevan]

So In the Group By Clause, we need to specify all the Columns what we have selected and we may of may not include the aggregate Column. This is what you have specified. Am I right?

[ZygD]

Correct, but you shouldn't include the aggregate column. That also makes no sense

[ZygD]

+1 Note for GROUP BY will collapse rows to the grouping, PARTITION BY will keep all rows. Still useful though

[sventechie]

Sometimes it is possible to just group by the row ID for a given table -- discovered with trial and error in PostgreSQL 9.5.

[ZygD]

@sventechie: PG honours the SQL standard more than others, which says (I think!) that GROUP BY <primary key> should be the same as GROUP BY <primary key>, <nonkey column1>, <nonkey column2> . That is, the PK is unique already so nonkey columns can be ignored

[JoeBass]

Feels super hacky, but it worked for my case. Also seemed to not be overly optimized, as my query slowed down significantly.

[WestCoastProjects]

This is the best answer for the spirit of the question.

[bonnal-enzo]

Note: we call this a window function