Oracle - Identifying Dominant Records - string

Question

I have a table that contains two STRING values (all single words) along with a corresponding COUNT for each occurrence of the STRING, e.g;
ID STR_1 COUNT_1 STR_2 COUNT_2
1 ORANGES 2 APPLES 10
2 APPLES 10 ORANGES 2
3 ORANGES 2 BANANAS 1
4 BANANAS 1 APPLES 10
5 BANANAS 1 ORANGES 2
N.B. STR_1 is considered the ‘master’ value. Also, the COUNT for each individual STRING value will be consistent between STR_1 And STR_2 and between rows (e.g. ORANGES will always have a COUNT of 2)
What I’m trying to achieve is to remove records whereby an ‘enantiomer’ exists, for example; in the above data, ID 2 would be considered an ‘enantiomer’ of ID 1 (ID 1.STR_1 = ID.2 STR_2 and ID 1.STR_2 = ID.2 STR_1), however, ID 2 would be considered the dominant record with ID 1 being discarded (because the COUNT for APPLES is greater than the COUNT for ORANGES) – therefore the desired output would be;
ID STR_1 COUNT_1 STR_2 COUNT_2
2 APPLES 10 ORANGES 2
3 ORANGES 2 BANANAS 1
4 BANANAS 1 APPLES 10
IF a scenario exists whereby the COUNT values between different STRINGS match, the longest STRING would be considered the dominant record and retained e.g.;
ID STR_1 COUNT_1 STR_2 COUNT_2
1 ORANGES 10 APPLES 10
2 APPLES 10 ORANGES 10
3 ORANGES 10 BANANAS 1
4 BANANAS 1 APPLES 10
5 BANANAS 1 ORANGES 10
With the desired output being;
ID STR_1 COUNT_1 STR_2 COUNT_2
1 ORANGES 10 APPLES 10
3 ORANGES 10 BANANAS 1
4 BANANAS 1 APPLES 10
Test Data;
WITH
TEST_DATA AS
(
SELECT 1 ID, 'ORANGES' STR_1, 2 COUNT_1, 'APPLES' STR_2, 10 COUNT_2 FROM DUAL
UNION
SELECT 2 ID, 'APPLES' STR_1, 10 COUNT_1, 'ORANGES' STR_2, 2 COUNT_2 FROM DUAL
UNION
SELECT 3 ID, 'ORANGES' STR_1, 2 COUNT_1, 'BANANAS' STR_2, 1 COUNT_2 FROM DUAL
UNION
SELECT 4 ID, 'BANANAS' STR_1, 1 COUNT_1, 'APPLES' STR_2, 10 COUNT_2 FROM DUAL
UNION
SELECT 5 ID, 'BANANAS' STR_1, 1 COUNT_1, 'ORANGES' STR_2, 2 COUNT_2 FROM DUAL
)
Any help finding a solution to the above would be much appreciated.
Many thanks in advance.

Answer 1

Use anti join (not exists operator):
select *
from test_data t
where not exists (
select 1 from test_data t1
where t.str_1 = t1.str_2
and t.str_2 = t1.str_1
and (
t.count_1 < t1.count_1
or
t.count_1 = t1.count_1
and
length( t.str_1 ) < length( t1.str_1 )
)
)
order by id
In a case when for a given pair of rows both counts and lengths are equal, then the query picks both rows.