Merge 2 rows into single row with data update?

Question

hello all,
i have a table with below data,
declare @tbl table (uid int, uname varchar(10), start_dt date, end_dt date)
insert into @tbl values (1, 'env1', '4/4/2010', '5/5/2012')
insert into @tbl values (2, 'env2', '5/4/2010', '6/6/2012')
--earlier start data is '4/4/2010' from 'env1'
--latest end data is '6/6/2012' from 'env2'
insert into @tbl values (3, 'env1', '3/3/2010', '4/4/2012')
insert into @tbl values (4, 'env2', '2/2/2010', '5/5/2012')
--earlier start data is '2/2/2010' from 'env2'
--latest end data is '5/5/2012' from 'env2'
insert into @tbl values (5, 'env1', '8/8/2010', '12/12/2012')
insert into @tbl values (6, 'env2', '9/9/2010', '10/10/2012')
--earlier start data is '8/8/2010' from 'env1'
--latest end data is '12/12/2012' from 'env1'insert into @tbl values (6, 'envX', '9/9/2010', '10/10/2012')insert into @tbl values (6, 'envY', '9/9/2010', '10/10/2012')
i need to merge 2 rows for column  "uname" having value "env1" & "env2" to "envZ" and need to capture earlier start date and latest end date from both and update with new.
the desire output should be,
declare @tbl table (uid int, uname varchar(10), start_dt date, end_dt date)
insert into @tbl values (1, 'envZ', '4/4/2010', '6/6/2012')
insert into @tbl values (4, 'envZ', '2/2/2010', '5/5/2012')
insert into @tbl values (5, 'envZ', '8/8/2010', '12/12/2012')
insert into @tbl values (6, 'envX', '9/9/2010', '10/10/2012')
insert into @tbl values (6, 'envY', '9/9/2010', '10/10/2012')
note - i must need to update one row and delete other row as i can't insert new rows (having huge data with other columns also).
please suggest optimize query. thanks!

Answer 1

Which version and edition of SQL Server are you using?
Whenever there is a row with 'env1', there is a corresponding 'env2' and vice versa? 
The row with 'env2' is always after the row with 'env1'?   (uid+1)
If the answers are yes to both questions above, here's a possibility:
-- code #1 v3
;with
ENVZ as (
SELECT uid= case when T1.start_dt <= T2.start_dt then T1.uid else T2.uid end,
start_dt= case when T1.start_dt <= T2.start_dt then T1.start_dt else T2.start_dt end,
end_dt= case when T1.end_dt > T2.end_dt then T1.end_dt else T2.end_dt end
from @tbl as T1
inner join @tbl as T2 on T2.uid=T1.uid+1 and T2.uname='env2'
where T1.uname = 'env1'
MERGE
into @tbl as T3
using ENVZ as T4
on T3.uid = T4.uid
when matched and (T3.uname in ('env1','env2')) then
UPDATE set T3.uname= 'envZ',
T3.start_dt= T4.start_dt,
T3.end_dt= T4.end_dt
when not matched by source and (T3.uname in ('env1','env2')) then
DELETE;
The table @tbl is read three times in the above code. There are probably ways to optimize the above code. Or even other more efficient approach.
José Diz     Belo Horizonte, MG - Brasil