Compare groups have same binding keys - sql

Question

I have samples as below in 2 different tables. I would have to write a sql to compare same source keys are binded together. binding keys in both the tables doesnt match.
In below example, in table 1 there are 3 binding keys 1, 2 & 3. Binding key 1 has 3 members attached to it ABC, XYZ, & QBC. Similarly binding key 2 & 3 each has got 2 source keys attached to them.
In table 2, binding key 99 has same 3 keys attached which are same as table 1(both count and keys are identical) whereas binding key 78 has got the same count as table 1's binding key 2 but they source keys are different. binding key 64 has 1 source key and binding key 65 has 1.
table 1:
==============================
Binding Key|source Key
1|ABC
1|XYZ
1|QBC
2|xxx
2|yyy
3|uuu
3|ddd
Table 2:
==========================
Binding Key|source Key
99|XYZ
99|QBC
99|ABC
78|xxx
78|QQQ
64|uuu
65|ddd
Expected output is to identify groups that doesnt match the count or the source key members.
Expected Output:
===========================
xxx
yyy
uuu
ddd
QQQ
Many Thanks!!

Answer 1

I found a solution. Its with the help of listagg function to concatenate strings within the same group and then compare them. example sql is shown below.
> SELECT * FROM (SELECT grp ,
> ListAgg( elmnt, ',' ) within GROUP ( ORDER BY pos) AS list FROM table1 GROUP BY grp ) table1 WHERE table1.list NOT IN
> (SELECT ListAgg( elmnt, ',' ) within GROUP ( ORDER BY pos) AS list
> FROM table2 GROUP BY grp ) UNION SELECT * FROM (SELECT grp ,
> ListAgg( elmnt, ',' ) within GROUP ( ORDER BY pos) AS list FROM table2 GROUP BY grp ) table1 WHERE table1.list NOT IN
> (SELECT ListAgg( elmnt, ',' ) within GROUP ( ORDER BY pos) AS list
> FROM table1 GROUP BY grp ) ;

Answer 2

This query gives desired output:
with t1 as ( select row_number() over (partition by grp order by elmnt) pos,
grp, elmnt from table1 ),
t2 as ( select row_number() over (partition by grp order by elmnt) pos,
grp, elmnt from table2 ),
tx1 as (select pos, grp grp1, elmnt,
listagg(elmnt, ',') within group (order by pos)
over (partition by grp) list
from t1),
tx2 as (select pos, grp grp2, elmnt,
listagg(elmnt, ',') within group (order by pos)
over (partition by grp) list
from t2)
select distinct elmnt
from (select * from tx1 full join tx2 using (list, elmnt))
where grp1 is null or grp2 is null;
You could easily change it to show lists, just replace distinct elmnt with distinct list. The difference between your answer and my query is listagg in analytic version and filtered full join instead of union combined with two not in clauses.
First two subqueries (t1 and t2) only adds pos column, which you did not present in original question ;-) Probably this can also be done with minus operator.
Test data and output:
create table table1 (grp number(3), elmnt varchar2(5));
insert into table1 values (1, 'ABC');
insert into table1 values (1, 'XYZ');
insert into table1 values (1, 'QBC');
insert into table1 values (2, 'xxx');
insert into table1 values (2, 'yyy');
insert into table1 values (3, 'uuu');
insert into table1 values (3, 'ddd');
create table table2 (grp number(3), elmnt varchar2(5));
insert into table2 values (99, 'XYZ');
insert into table2 values (99, 'QBC');
insert into table2 values (99, 'ABC');
insert into table2 values (78, 'xxx');
insert into table2 values (78, 'QQQ');
insert into table2 values (64, 'uuu');
insert into table2 values (65, 'ddd');
ELMNT
-----
uuu
QQQ
yyy
ddd
xxx