SQL query too complicated for me - sql

Question

I have a table with a note column and there can be value 'Start' or 'End'. And then there are other columns, that can have same value, but only difference is in that 'note' column...
I need to select rows which have the 'note' set to 'Start', but only those, there are no row with the same values and 'note' set to 'End'. Sorry, it's complicated to explain. I'll try to show some example.
Coll1 Coll2 Coll3 note
-----------------------------
a a a Start
a a a End
b b b Start
b b b End
c c c Start <- I need select those rows
-- There is no row with 'c c c End' combination in the table
d d d Start
d d d End
e e e Start <- I need select those rows
-- There is no row with 'e e e End' combination in the table
Can anybody help me please?

Answer 1

Try to use
SELECT *
FROM tbl t1
WHERE t1.note = 'Start' AND NOT EXISTS (SELECT *
FROM tbl t2
WHERE t2.note = 'End'
AND t2.Coll1 = t1.Coll1
AND t2.Coll2 = t1.Coll1
AND t2.Coll3 = t1.Coll3)
Maybe this query not optimal, but this query is easy for understand.

Answer 2

The most simple way should be to aggregate the records and check whether there is an End record for the group:
select col1, col2, col3
from mytable
group by col1, col2, col3
having count(case when note = 'Start' then 1 end) = 1
and count(case when note = 'End' then 1 end) = 0;
Adjust this as you like (e.g. if you are fine with several start records make it >= 1 instead of = 1).

Answer 3

Will this work for you?
SELECT *
FROM mytable t1
LEFT JOIN mytable t2 ON
t1.Coll1 = t2.Coll1 AND
t1.Coll2 = t2.Coll2 AND
t1.Coll3 = t2.Coll3 AND
t2.note = 'End'
WHERE t1.note = 'Start' AND t2.Coll1 IS NULL

Answer 4

You'll get more answers if you include CREATE TABLE and INSERT statements in your question. I'm using PostgreSQL; Oracle is similar.
create table test (
col1 char(1) not null,
col2 char(1) not null,
col3 char(1) not null,
note varchar(10) not null
check (note in ('start', 'end')),
primary key (col1, col2, col3, note)
);
I'm assuming primary key (col1, col2, col3, note). The presence of NULL complicates this approach.
insert into test values
('a', 'a', 'a', 'start'),
('a', 'a', 'a', 'end'),
('b', 'b', 'b', 'start'),
('b', 'b', 'b', 'end'),
('c', 'c', 'c', 'start'),
('d', 'd', 'd', 'start'),
('d', 'd', 'd', 'end'),
('e', 'e', 'e', 'start');
Now we can take a set of starts and a set of ends. A left outer join will preserve all the rows in starts; missing rows in ends will be filled with NULL.
with starts as (
select * from test where note = 'start'
), ends as (
select * from test where note = 'end'
)
select s.* from starts s
left outer join ends e
on e.col1 = s.col1
and e.col2 = s.col2
and e.col3 = s.col3
where e.col1 is null
and e.col2 is null
and e.col3 is null
and e.note is null;

Answer 5

SELECT coll1, coll2, coll3, count(*)
FROM tbl
Where note='start'
GROUP BY coll1, coll2, coll3,
HAVING count(*) < 2

Answer 6

Here's another solution to add to the ones you've already got:
WITH sample_data AS (SELECT 'a' Coll1, 'a' Coll2, 'a' Coll3, 'Start' note FROM dual UNION ALL
SELECT 'a' Coll1, 'a' Coll2, 'a' Coll3, 'End' note FROM dual UNION ALL
SELECT 'b' Coll1, 'b' Coll2, 'b' Coll3, 'Start' note FROM dual UNION ALL
SELECT 'b' Coll1, 'b' Coll2, 'b' Coll3, 'End' note FROM dual UNION ALL
SELECT 'c' Coll1, 'c' Coll2, 'c' Coll3, 'Start' note FROM dual UNION ALL
SELECT 'd' Coll1, 'd' Coll2, 'd' Coll3, 'Start' note FROM dual UNION ALL
SELECT 'd' Coll1, 'd' Coll2, 'd' Coll3, 'End' note FROM dual UNION ALL
SELECT 'e' Coll1, 'e' Coll2, 'e' Coll3, 'Start' note FROM dual)
-- End of mimicking a table called "sample_data" with your data in it
-- for use in the query below:
SELECT coll1,
coll2,
coll3,
MAX(CASE WHEN note = 'Start' THEN note END) note
FROM sample_data
GROUP BY coll1,
coll2,
coll3
HAVING MAX(CASE WHEN note = 'Start' THEN note END) = 'Start'
AND MAX(CASE WHEN note = 'End' THEN note END) IS NULL;
COLL1 COLL2 COLL3 NOTE
----- ----- ----- -----
e e e Start
c c c Start
It only has one access through the table (as opposed to the two accesses in Surename's answer), but you should test which solution works best with your data and table structure - one may be faster than the other.

Answer 7

For completeness sake, for it has been suggested in a comment to preferably think in sets:
select col1, col2, col3 from mytable where note = 'Start'
minus
select col1, col2, col3 from mytable where note = 'End';

Answer 8

select col1,col2,col3,count(note) from tb1
group by col1,col2,col3
having count(note)=1
also u can do it
select * from tb1
where note <> 'End' and note = 'Start'