Oracle 11g:Query to return only 1 to 1 relationship & random selection

Question

Hi
I have a complex query to modify but I have below the sample tables and data with only very few fields(only affected fields).
Query based on 2 tables b_test and s_test.
Pls see below.
create table b_test(building_id number not null,invalid varchar2(1));
create table s_test(sub_building_id number not null,building_id number ,invalid varchar2(1),sequence_no number);
insert into b_test values (1000,'N');
insert into b_test values(2000,'N');
insert into b_test values(3000,'N');
commit;
insert into s_test values(1,1000,'N',90);
insert into s_test values(2,1000,'N',91);
insert into s_test values(3,1000,'N',92);
insert into s_test values(4,1000,'Y',93);
insert into s_test values(5,NULL,'N',NULL);
insert into s_test values(1,2000,'N',94);
insert into s_test values(2,2000,'N',95);
insert into s_test values(3,2000,'N',96);
insert into s_test values(4,2000,'N',97);
insert into s_test values(5,2000,'N',98);
insert into s_test values(6,NULL,'N',NULL);
insert into s_test values(10,3000,'N',99);
insert into s_test values(11,3000,'N',100);
commit;The query below returns all rows required:(also see results:)
select b.building_id,b.invalid,s.sub_building_id,s.sequence_no from b_test b,
(select * from s_test where invalid='N') s
where b.building_id = s.building_id(+)
and b.invalid='N'
Results:
BUILDING_ID INVALID      SUB_BUILDING_ID     SEQUENCE_NO
1000              N     1                     90
1000             N     2                     91
1000             N     3                     92
2000             N     1                     94
2000             N     2                     95
2000             N     3                     96
2000             N     4                     97
2000             N     5                     98
3000             N     10                      99
3000             N     11                     100Now there are 2 requirements:
1)How can the above query be changed so that 1:1 relationship if sub_building_id is returned?i.e For 1 building_id, only show 1 sub_building(This could be a random selection)
(Pls help with query)
The results would be like
BUILDING_ID INVALID     SUB_BUILDING_ID     SEQUENCE_NO
1000               N     1                      90
2000               N     1                      94
3000               N     11                     1002)How can the same SEQUENCE_NO be shown for all sub_buildings for the same building? (Pls help with query)
The results will be:
BUILDING_ID INVALID     SUB_BUILDING_ID     SEQUENCE_NO
1000             N     1                       90
1000             N     2                       90
1000             N     3                       90
2000             N     1                       94
2000              N     2                       94
2000             N     3                       94
2000             N     4                       94
2000             N     5                       94
3000             N     10                       99
3000             N     11                       99Many thanks!
Edited by: Krithi on 08-Nov-2012 08:48
Edited by: Krithi on 08-Nov-2012 08:55

Answer 1

Krithi wrote:
Hi
I have a complex query to modify but I have below the sample tables and data with only very few fields(only affected fields).
Query based on 2 tables b_test and s_test.
Pls see below.
create table b_test(building_id number not null,invalid varchar2(1));
Thanks for posting the CREATE TABLE and INSERT statements, and your existing query; that's very helpful.
The query below returns all rows required:(also see results:)
select b.building_id,b.invalid,s.sub_building_id,s.sequence_no from b_test b,
(select * from s_test where invalid='N') s
where b.building_id = s.building_id(+)
and b.invalid='N'
Results:
BUILDING_ID INVALID      SUB_BUILDING_ID     SEQUENCE_NO
1000              N     1                     90
1000             N     2                     91
1000             N     3                     92
2000             N     1                     94
2000             N     2                     95
2000             N     3                     96
2000             N     4                     97
2000             N     5                     98
3000             N     10                      99
3000             N     11                     100
When I run your query, I get NULL for sequence_no on the last 2 rows, where building_id=3000. The numbers 99 and 100 don't seem to occur in either table. Did you post the worng sample data and/or results?
>
Now there are 2 requirements:
1)How can the above query be changed so that 1:1 relationship if sub_building_id is returned?i.e For 1 building_id, only show 1 sub_building(This could be a random selection)
(Pls help with query) Here's one way:
WITH       got_r_num  AS
     SELECT  sub_building_id
     ,     building_id
     ,     sequence_no
     ,     ROW_NUMBER () OVER ( PARTITION BY  building_id
                               ORDER BY          sequence_no
                       )         AS r_num
     FROM    s_test
     WHERE     invalid     = 'N'
SELECT    b.building_id
,       b.invalid
,       r.sub_building_id
,       r.sequence_no
FROM             b_test     b
LEFT OUTER JOIN      got_r_num  r  ON  r.building_id  = b.building_id
WHERE     NVL ( r.r_num
           , 1
           )          = 1
ORDER BY  b.building_id
;This is called a Top-N Query , because we're picking N items (N = 1 in this case) from the top of an ordered list. What makes one item the "top", and another one "lower"? That's determined by the analytic ORDER BY clause, in this case
ORDER BY      sequence_noThat means the row with the lowest sequence_no (for each building_id) will get r_num=1. If you want a random row from that building_id to be chosen as #1, then you can change the analytic ORDER BY clause to
ORDER BY      dbms_random.valueYou can ORDER BY anything you like, even a constant, but you must have an analytic ORDER BY clause. ROW_NUMBER requires an analytic ORDER BY clause.
The results would be like
BUILDING_ID INVALID     SUB_BUILDING_ID     SEQUENCE_NO
1000               N     1                      90
2000               N     1                      94
3000               N     11                     100
Again, I don't see where the 100 comes from. The results I get are:
BUILDING_ID I SUB_BUILDING_ID SEQUENCE_NO
       1000 N               1          90
       2000 N               1          94
       3000 N              11
2)How can the same SEQUENCE_NO be shown for all sub_buildings for the same building? (Pls help with query)
SELECT    b.building_id
,       b.invalid
,       s.sub_building_id
,       MIN (s.sequence_no) OVER ( PARTITION BY  s.building_id)
                        AS seq_no
FROM             b_test  b
LEFT OUTER JOIN      s_test  s  ON  s.building_id  = b.building_id
                            AND s.invalid      = 'N'
The results will be:
BUILDING_ID INVALID     SUB_BUILDING_ID     SEQUENCE_NO
1000             N     1                       90
1000             N     2                       90
1000             N     3                       90
2000             N     1                       94
2000              N     2                       94
2000             N     3                       94
2000             N     4                       94
2000             N     5                       94
3000             N     10                       99
3000             N     11                       99
Again, I don't see where you get sequence_no = 99. The results I get are:
BUILDING_ID I SUB_BUILDING_ID     SEQ_NO
       1000 N               1         90
       1000 N               2         90
       1000 N               3         90
       2000 N               1         94
       2000 N               2         94
       2000 N               5         94
       2000 N               3         94
       2000 N               4         94
       3000 N              10
       3000 N              11Edited by: Frank Kulash on Nov 8, 2012 12:12 PM
Added explanation and results
Edited by: Frank Kulash on Nov 8, 2012 12:28 PM
It looks like you cahnged your sample data from
insert into s_test values(10,3000,'N',NULL);
insert into s_test values(11,3000,'N',NULL);to
insert into s_test values(10,3000,'N',99);
insert into s_test values(11,3000,'N',100);The queries I posted are niow getting 99, like you requested.