I have two tables :

TABLE_A

pos_id    res_id     bb_id    bsk_name
1122      10000       1444     type_1 
1122      10000       5678     type_2
1122      10001       1444     type_1
1122      10001       5678     type_2
1122      10002       1467     type_1
1122      10002       5678     type_2
1122      10003       1467     type_1
1122      10003       5678     type_2

table_b

pos_id    row_id    bb_id     bsk_name
1122        1        1444      type_1
1122        1        5678      type_2
1122        2        1467      type_1
1122        2        5678      type_2

I wanted to join the table_a and table_b to get the row_id for each res_id.

the res_id 10000 and 10001 has to have the row_id 1 and res_id 10002 and 10003 has to have the row_id 2. but since there is no unique column to join these two tables i get duplicated values for the bb_id 5678 since they are the same in both row_id's.

So Is there a way like to listagg the bb_id's with the erg_id in table_a and row_id in table_b to join these two tables ?

CodePudding user response：

You can use CAST(COLLECT(...) AS ...) to aggregate the rows into a user-defined collection and then compare the collections.

First, create a nested-table-type collection:

CREATE TYPE int_list AS TABLE OF INT;

Then you can use:

SELECT a.pos_id,
       a.res_id,
       a.bb_id,
       a.bsk_name,
       b.row_id,
       b.bsk_name
FROM   (
  SELECT a.*,
         CAST(
           COLLECT(bb_id) OVER (PARTITION BY pos_id, res_id)
           AS int_list
         ) AS all_bb_ids
  FROM   table_a a
) a
INNER JOIN (
  SELECT b.*,
         CAST(
           COLLECT(bb_id) OVER (PARTITION BY pos_id, row_id)
           AS int_list
         ) AS all_bb_ids
  FROM   table_b b
) b
ON (    a.pos_id = b.pos_id
    AND a.all_bb_ids = b.all_bb_ids
    AND a.bb_id = b.bb_id)

Which, for your sample data:

CREATE TABLE TABLE_A ( pos_id, res_id, bb_id, bsk_name ) AS
SELECT 1122, 10000, 1444, 'type_1' FROM DUAL UNION ALL
SELECT 1122, 10000, 5678, 'type_2' FROM DUAL UNION ALL
SELECT 1122, 10001, 1444, 'type_1' FROM DUAL UNION ALL
SELECT 1122, 10001, 5678, 'type_2' FROM DUAL UNION ALL
SELECT 1122, 10002, 1467, 'type_1' FROM DUAL UNION ALL
SELECT 1122, 10002, 5678, 'type_2' FROM DUAL UNION ALL
SELECT 1122, 10003, 1467, 'type_1' FROM DUAL UNION ALL
SELECT 1122, 10003, 5678, 'type_2' FROM DUAL;

CREATE TABLE table_b (pos_id, row_id, bb_id, bsk_name) AS
SELECT 1122, 1, 1444, 'type_1' FROM DUAL UNION ALL
SELECT 1122, 1, 5678, 'type_2' FROM DUAL UNION ALL
SELECT 1122, 2, 1467, 'type_1' FROM DUAL UNION ALL
SELECT 1122, 2, 5678, 'type_2' FROM DUAL;

Outputs:

POS_ID	RES_ID	BB_ID	BSK_NAME	ROW_ID	BSK_NAME
1122	10000	1444	type_1	1	type_1
1122	10000	5678	type_2	1	type_2
1122	10001	1444	type_1	1	type_1
1122	10001	5678	type_2	1	type_2
1122	10002	1467	type_1	2	type_1
1122	10002	5678	type_2	2	type_2
1122	10003	1467	type_1	2	type_1
1122	10003	5678	type_2	2	type_2

db<>fiddle here

Update

You could also use LISTAGG to aggregate the data; however, if the aggregated string will exceed 4000 bytes then the aggregation will fail (using CAST(COLLECT(...) ...) does not have this limitation but does require you create a collection data type):

SELECT a.pos_id,
       a.res_id,
       a.bb_id,
       a.bsk_name,
       b.row_id,
       b.bsk_name
FROM   (
  SELECT a.*,
         LISTAGG(bb_id, ',') WITHIN GROUP (ORDER BY bb_id)
           OVER (PARTITION BY pos_id, res_id) AS all_bb_ids
  FROM   table_a a
) a
INNER JOIN (
  SELECT b.*,
         LISTAGG(bb_id, ',') WITHIN GROUP (ORDER BY bb_id)
           OVER (PARTITION BY pos_id, row_id) AS all_bb_ids
  FROM   table_b b
) b
ON (    a.pos_id = b.pos_id
    AND a.all_bb_ids = b.all_bb_ids
    AND a.bb_id = b.bb_id)

db<>fiddle here