Join without duplicates

Question

hello,
i'm quite new to oracle and dbms in general, so please be patient ;)
so here goes:
i have two tables: people and cities
my task: select all people who live in a city that has a 'z' in its name or where the city_id field is empty.
so i thought i'd join the two tables and see what happens:
select people.* from people, cities
where (people.city_id = cities.city_id and cities.city_name like '%z%')
or (people.city_id is null);
the problem here is that there are duplicate records for the people who have no city_id. how can i change the where-clause in order to avoid this?
i was told not to use distinct, union, intersect.
greets and thx!

Answer 1

Processing ...
with people as (
     select 'a' as p_name,
          1 as city_id
     from dual
     union all
     select 'a' as name,
          1 as city_id
     from dual
), cities as (
     select 'z' as c_name,
          1 as city_id
     from dual
select people.* from people, cities
where (people.city_id = cities.city_id and
cities.c_name like '%z%')
or (people.city_id = cities.city_id (+) and
people.city_id is null)
Query finished, retrieving results...
P_NAME                   CITY_ID               
a                                               1
a                                               1
2 row(s) retrievedBye AlessandroThere are two different things here:
1) duplicate records because source have the same duplicate records i.e. as in your case a, 1 two times
2) duplicate records because join condition is too weak.
Your query Alessandro had duplicates because of (1).
OP original query have duplicates or triplicates or whatever else because he allowed cartesian product between peoples rows which haven't city_id and cities i.e. it was case (2)
It is easier to spot if you select not only people.* but also city columns.
Look what original query gives us:
SQL> with people as (
  2    select 1 p_id, 'Live in z1' name, 1 c_id from dual
  3    union all
  4    select 2 p_id, 'Live in z2' name, 1 c_id from dual
  5    union all
  6    select 3, 'Dont live in z1', 2 from dual
  7    union all
  8    select 4, 'Dont live in z2', 2 from dual
  9    union all
10    select 5, 'Dont live anywhere1', null from dual
11    union all
12    select 6, 'Dont live anywhere2', null from dual
13  ),
14  cities as (
15    select 1 city_id, 'bzz' city_name from dual
16    union all
17    select 2 city_id, 'aaa' city_name from dual
18    union all
19    select 3 city_id, 'bbb' city_name from dual
20  )
21  select * from people, cities
22  where (people.c_id = cities.city_id and cities.city_name like '%z%')
23  or people.c_id is null
24  /
                P_ID NAME                                C_ID              CITY_ID CIT
                   1 Live in z1                             1                    1 bzz
                   2 Live in z2                             1                    1 bzz
                   5 Dont live anywhere1                                         1 bzz
                   6 Dont live anywhere2                                         1 bzz
                   5 Dont live anywhere1                                         2 aaa
                   6 Dont live anywhere2                                         2 aaa
                   5 Dont live anywhere1                                         3 bbb
                   6 Dont live anywhere2                                         3 bbbWe have all combinations between city rows and rows in people, which city_id is null.
If source rows already have duplicates then there isn't any possibility to get rid of them just using joins. Then we need distinct, group by or some set operation.
Gints Plivna
http://www.gplivna.eu