Multiple SQL MAX when items are not in order - sql

Question

I have some data as below:
DECLARE #MyTable AS TABLE
(productName varchar(13), test1 int,test2 int)
INSERT INTO #MyTable
(productName, test1,test2)
VALUES
('a', 1,1),
('a', 2,2),
('a', 3,3),
('b', 1,4),
('b', 2,5),
('b', 3,6),
('a', 1,7),
('a', 4,8),
('a', 5,9)
;
SELECT productname,MAX(test1) from #MyTable group BY productname
a MAX query on test1 column gives
a,5
b,3
but I need to have result as
a,3
b,3
a,5
when I have order by test2

Answer 1

You can solve this by using a trick with row_numbers, so that you assign 2 different row numbers, one for the whole data and one that is partitioned by productname. If you compare the difference between these numbers, you can figure out when product name has changed, and use that to determine the max values for each group.
select productname, max(test1) from (
SELECT *,
row_number() over (order by test2 asc) -
row_number() over (partition by productname order by test2 asc) as GRP
from #MyTable
) X
group by productname, GRP
You can test this in SQL Fiddle
If the test2 column is always a row number without gaps, you can use that too instead of the first row number column. If you need ordering in the data, you'll have to for example to use the max of test1 to do that.

Answer 2

Please check the following SQL Select statement
DECLARE #MyTable AS TABLE (productName varchar(13), test1 int,test2 int)
INSERT INTO #MyTable
(productName, test1,test2)
VALUES
('a', 1,1),
('a', 2,2),
('a', 3,3),
('b', 1,4),
('b', 2,5),
('b', 3,6),
('a', 1,7),
('a', 4,8),
('a', 5,9)
DECLARE #MyTableNew AS TABLE (id int identity(1,1), productName varchar(13), test1 int,test2 int)
insert into #MyTableNew select * from #MyTable
--select * from #MyTableNew
;with cte as (
SELECT
id, productName, test1, test2,
case when (lag(productName,1,'') over (order by id)) = productName then 0 else 1 end ischange
from #MyTableNew
), cte2 as (
select t.*,(select sum(ischange) from cte where id <= t.id) grp from cte t
)
select distinct grp, productName, max(test1) over (partition by grp) from cte2
This is implemented according to the following SQL Server Lag() function tutorial
The Lag() function is used to identify and order the groups in table data

Answer 3

Please try this query
DECLARE #MyTable AS TABLE
(productName varchar(13), test1 int,test2 int)
INSERT INTO #MyTable
(productName, test1,test2)
VALUES
('a', 1,1),
('a', 2,2),
('a', 3,3),
('b', 1,4),
('b', 2,5),
('b', 3,6),
('a', 1,7),
('a', 4,8),
('a', 5,9)
;
SELECT productname,MAX(test1)
from #MyTable
where test1 = test2
group BY productname
union all
SELECT productname,MAX(test1)
from #MyTable
where test1 != test2
group BY productname