sql aggregations - sql

Question

any ideas why this doesn't work?
select [column_name_2], max(count(distinct([column_name_1])))
from [table_name]
group by [column_name_2]
but it works if done like this
select [column_name_2], count(distinct([column_name_1])) as [x]
into #temp_table
from [table_name]
group by [column_name_2]
select max(x)
from #temp_table

Answer 1

Well, that's just the way SQL (the language) is defined to work. When you use GROUP BY, the corresponding SELECT list will produce a row for each group in the result. You're trying to take that result and aggregate twice, once with GROUP BY [column_name_2] and a second time with GROUP BY (), as defined by standard SQL. We can't do that in the same query expression.
The good news is you can break this up into more than one query expression:
WITH cte1 AS (
SELECT count(distinct([column_name_1])) AS cnt
FROM [table_name]
GROUP BY [column_name_2]
)
SELECT MAX(cnt) FROM cte1
;
or use a derived table.
You can even order the initial query result by cnt DESC and limit the result to the first row.
In your case, you may not want just the MAX, but also the other column.
With SQL Server, which you may be using. Note: You should add a database specific tag to the question.
SELECT TOP 1 [column_name_2], count(distinct([column_name_1])) AS cnt
FROM [table_name]
GROUP BY [column_name_2]
ORDER BY cnt DESC
;

Answer 2

I don't understand "this doesn't work", what were you expecting and what did you get? Normally you include the GROUP BY value in the result set. So it would be:
select [column_name_2], max(cnt) cnt
from (select [column_name_2], count(distinct [column_name_1]) cnt
from [table_name]
group by [column_name_2]) x
group by [column_name_2]
Ok, after reading your comment I think above is what you are looking for.

Answer 3

SELECT [column_name_two]
, max(x) x
FROM (
SELECT [column_name_two]
, COUNT(DISTINCT [column_name_one]
FROM table_name
GROUP BY [column_name_two]
) AS Tbl
GROUP BY [column_name_two]