How do you combine query results from different rows into one? - sql

Question

My original query:
SELECT desc, start_date
FROM foo.bar
WHERE desc LIKE 'Fall%' AND desc NOT LIKE '%Med%'
UNION
SELECT desc, end_date
FROM foo.bar
WHERE desc LIKE 'Spring%' AND desc NOT LIKE '%Med%'
ORDER BY start_date;
With this query, I get (roughly) the data set I am looking for. I now need to take that data and combine the results taking two at a time in order and then produce a result like:
DESC
START_DATE
END_DATE
Fall 1971 - Spring 1972
15-AUG-71
15-MAY-72
Fall 1971 - Spring 1972
15-AUG-72
15-MAY-73
Where DESC is a concatenation of the DESC form row 1 and 2, START_DATE is the date from row 1 and END_DATE is the date from row 2. Following this same pattern for the entire data set.
Any help with a query that will produce the result I need is greatly appreciated. Not sure if I'm heading down the right path or if that originally query is just wrong.
As stated above, I tried the supplied query, which gives me the data I need. However, I've been unsuccessful in finding a way to format it into my desired output. It should also be noted that I am running this on an Oracle database.

Answer 1

Instead of union, use each of those queries as CTEs (with a slight modification - include row number you'll later use in JOIN):
Sample data:
SQL> with test (description, datum) as
2 (select 'Fall 1971' , date '1971-08-15' from dual union all
3 select 'Spring 1972', date '1972-05-15' from dual union all
4 select 'Fall 1972' , date '1972-08-15' from dual union all
5 select 'Spring 1973', date '1973-05-15' from dual union all
6 select 'Fall 1973' , date '1973-08-15' from dual union all
7 select 'Spring 1974', date '1974-05-15' from dual union all
8 select 'Fall 1974' , date '1974-08-15' from dual union all
9 select 'Spring 1975', date '1975-05-15' from dual
10 ),
Query begins here: t_start and t_end represent your current queries
11 t_start as
12 (select description, datum,
13 row_number() Over (order by datum) rn
14 from test
15 where description like 'Fall%' and description not like '%Med%'
16 ),
17 t_end as
18 (select description, datum,
19 row_number() Over (order by datum) rn
20 from test
21 where description like 'Spring%' and description not like '%Med%'
22 )
Finally:
23 select s.description ||' - '|| e.description as description,
24 s.datum start_date,
25 e.datum end_date
26 from t_start s join t_end e on s.rn = e.rn
27 order by s.rn;
DESCRIPTION START_DAT END_DATE
------------------------- --------- ---------
Fall 1971 - Spring 1972 15-AUG-71 15-MAY-72
Fall 1972 - Spring 1973 15-AUG-72 15-MAY-73
Fall 1973 - Spring 1974 15-AUG-73 15-MAY-74
Fall 1974 - Spring 1975 15-AUG-74 15-MAY-75
SQL>

Answer 2

You can also use the MODEL clause to avoid to scan the table twice:
with data(description,datum) as (
select 'Fall 1971' , date '1971-08-15' from dual union all
select 'Spring 1972', date '1972-05-15' from dual union all
select 'Fall 1972' , date '1972-08-15' from dual union all
select 'Spring 1973', date '1973-05-15' from dual union all
select 'Fall 1973' , date '1973-08-15' from dual union all
select 'Spring 1974', date '1974-05-15' from dual union all
select 'Fall 1974' , date '1974-08-15' from dual union all
select 'Spring 1975', date '1975-05-15' from dual
)
select description, start_date, end_date
from (
select rn, desc1 as description, start_date, end_date
from (
select row_number() over(order by datum) as rn, description, datum
from data
where description not like '%Med%'
)
model
dimension by (rn)
measures (
cast(' ' as varchar2(256)) as desc1, description, cast(NULL as DATE) start_date, cast(NULL as DATE) end_date , datum
)
rules (
desc1[mod(rn,2)=1] = description[cv()] || ' - ' || description[cv()+1],
start_date[mod(rn,2)=1] = datum[cv()],
end_date[mod(rn,2)=1] = datum[cv()+1]
)
)
where mod(rn,2)=1
;