How to avoid duplicated rows using the outer join

Question

Hi everybody,
I have the following query:
select a.usr_login, b.ugp_rolename, b.ugp_display_name from
(select usr.usr_login, usr.usr_key, usg.ugp_key from usr,usg
where usg.usr_key = usr.usr_key
and usr.usr_login IN ('C01015','C01208')) a,
(select ugp.ugp_key, ugp.ugp_display_name from ugp
where ugp.ugp_rolename LIKE 'B-%') b
where a.ugp_key = b.ugp_key (+)
The first query 'a' has the following result:
usr_login <space> usr_key <space> ugp_key
C01015 <space> 49 <space> 565
C01015 <space> 49 <space> 683
C01015 <space> 49 <space> 685
C01015 <space> 49 <space> 3
C01208 <space> 257 <space> 3
The usr_login on table usr is the primary key, and as you can see above, for each usr_login I can find one ore more ugp_key on the table usg.
The query 'b' gives the list of all the usr_login's roles which have the name LIKE 'B-%' (it means '*Business Roles*'), and all the respective role's key (ugp_key)
So, when I join the query 'a' with the query 'b', I expect to find for every usr_login the respective ugp_display_name (the Business Role name).
Because the query 'b' contains ONLY the ugp_keys of the Business Roles, when I execute the complete query, this is the result:
usr_login <space> ugp_rolename <space> ugp_display_name
C01015 <space> BK005 <space> TELLER 1
C01015 <space> BK003 <space> TELLER 2
C01015 <space> null <space> null
C01015 <space> null <space> null
C01208 <space> null <space> null
As you can see, with the outer join I obtain the Business Name (ugp_display_name) for each occurrence (and I have 2 rows duplicated with 'null' for the usr_login C01015); This beacuse the query 'b' doesn't have the two ugp_keys 685 and 3.
Instead I'd like to have the following result:
usr_login <space> ugp_rolename <space> ugp_display_name
C01015 <space> BK005 <space> TELLER 1
C01015 <space> BK003 <space> TELLER 2
C01208 <space> null <space> null
deleting ONLY the duplicated rows with null, when the usr_login has already at least one ugp_display_name not null.
For example:
1) The usr_login 'C01015' has 2 Business Roles (with ugp_key = 565 and 683) and other 2 not-Business Roles (with ugp_key = 685 and 3) --> I want to see ONLY the 2 records related to the Business Roles
2) The usr_login 'C01208' has only one not-Business Roles (with ugp_key = 3) --> I want to see the record related to the not- Business Role
Practically:
1) When a usr_login has one or more Business Roles and other not-Business Roles , I'd like to see ONLY the records about the Business Roles (not the records with 'null','null')
2) When a usr_login doesn't have Business Roles, I'd like to see the records about the same usr_login with 'null','null'
This, because I need to show both usr_logins: with and without Business Roles.
Anybody has any suggestions ? Any help will be appreciated.
Thanks in advance for any help !!
Alex

Answer 1

Hi, Alex,
So you want to display rows from a where either
(1) the row has a match in b, or
(2) no iwith the same usr_login has a match.
Here's one way to do that:
WITH     a     AS
     SELECT  usr.usr_login, usr.usr_key, usg.ugp_key
     FROM      usr
     ,     usg
     WHERE      usg.usr_key     = usr.usr_key
     AND     usr.usr_login     IN ('C01015','C01208')
,     b     AS
     SELECT  ugp.ugp_key, ugp.ugp_display_name
     FROM      ugp
     WHERE     ugp.ugp_rolename     LIKE 'B-%'
,     got_match_cnt     AS
     SELECT     a.usr_login, b.ugp_rolename, b.ugp_display_name
     ,     b.ugp_key
     ,     COUNT (b.ugp_key) OVER (PARTITION BY  a.usr_login)     AS match_cnt
     FROM      a
     ,     b
     WHERE     a.ugp_key     = b.ugp_key (+)
SELECT     usr_login, ugp_rolename, ugp_display_name
FROM     got_match_cnt
WHERE     ugp_key          IS NOT NULL     -- Condition (1)
OR     match_cnt     = 0              -- Condition (2)
;If b.ugp_rolename or b.ugp_display_name can not be NULL, then you could use that just as well as b.ugp_key for testing condition (1).
By the way, you don't need sub-queries for a and b; you can do all the joins and all the filtering (except conditions (1) and (2)) in one query, but the sub-queries aren't hurting anything. If you find the separate sub-queries easier to understand, debug and maintain, then, by all means, keep them.
I hope this answers your question.
If not, post a little sample data (CREATE TABLE and INSERT statements, relevant columns only) for all tables, and also post the results you want from that data.
Explain, using specific examples, how you get those results from that data.
Always say which version of Oracle you're using.