Question

I’ve got a SQL-statement with a – for me not explainable – strange behaviour.
Perhaps you could find what’s wrong:

When I use the statement

select count(*) from department

I got 2755 results

Using the following statement

select 
      building1.street, building1.streetno, building1.plz, building1.city, dept1.buildingid
  from 
      department dept1
    left join 
      supporter sup 
    on 
      dept.supporterid = sup.id
    left join 
      building building1
    on 
      sup.buildingid = building1.ibuildingid
  where 
      dept.usepostaladresssupporter = 1 
  union all
  select 
      building2.street, building2.streetno, building2.plz, building2.city, dept2.buildingid
  from 
      building building2
    right join 
      tueks_department dept2 
    on 
      dept2.buildingid = building2.ibuildingid
  where 
      dept2.usepostaladresssupporter = 0

I got 2755 results too.
But when I want to combine the two statements with a left join:

select count(*) from department
  left join
    (
      select 
          building1.street, building1.streetno, building1.plz, building1.city, dept1.buildingid
      from 
          department dept1
        left join 
          supporter sup 
        on 
          dept.supporterid = sup.id
        left join 
          building building1
        on 
          sup.buildingid = building1.ibuildingid
      where 
          dept.usepostaladresssupporter = 1 
      union all
      select 
          building2.street, building2.streetno, building2.plz, building2.city, dept2.buildingid
      from 
          building building2
        right join 
          tueks_department dept2 
        on 
          dept2.buildingid = building2.ibuildingid
      where 
          dept2.usepostaladresssupporter = 0
    ) postadress
  on
     department.buildingid = postadress.buildingid;

I got 3648513 results.

My expectation was, that I get only 2755 results.
Where’s the mistake?

Thanks for help!

Answer 1

I assume that buildingid is not unique (for my reasoning to hold true, it can’t be unique)

Imagine following simple tables

TableA

create TableA (name VARCHAR(32));
insert into TableA values ('Lieven');
insert into TableA values ('Lieven');      

TableB

create TableB (name VARCHAR(32));
insert into TableB values ('Lieven');
insert into TableB values ('Lieven');
insert into TableB values ('AnyOtherValue');

Select statement

select * from TableA a left outer join TableB b on a.name = b.name

As each record of TableA is matched with each record of TableB where the name is equal, this will result in 4 records (the AnyOtherValue is dissmissed as it doesn’t match)

• The first record of TableA is returned with two of three records of `TableB’
• The second record of TableA is returned with two of three records of `TableB’