Re: sum multiple tables gives wrong answer?

Поиск
Список
Период
Сортировка
От Stephen Frost
Тема Re: sum multiple tables gives wrong answer?
Дата
Msg-id 20100602171359.GJ21875@tamriel.snowman.net
обсуждение исходный текст
Ответ на sum multiple tables gives wrong answer?  ("Michael Diener" <m.diener@gomogi.com>)
Список pgsql-general
* Michael Diener (m.diener@gomogi.com) wrote:
> I have an SQL problem that I thought was easy to do but gives me always the
> wrong answer.

I think it's the right answer- the problem is that you're asking SQL a
different question than what you want the answer to.

> 2 Tables with a column called "flaeche" "double precision", in English
> "area" and I want to sum up the values for flaeche in each table to give me
> the total area for flaeche in each table.
>
> Correct answer comes with this sql
>
> select  sum(flaeche)/10000 as "greens HA"  from green;
>
> Wrong Answer with this query
>
> select
>
>  sum(green.flaeche)/10000 as "greens HA",
>
>  sum (fairway.flaeche)/10000 as "fairway HA"
>
>   from green, fairway;
>
> What is going on ??

These are very different queries.  Take out the sum()'s and see what you
get.  What you'll find is that, because the join is unconstrained,
you're getting a cartesian product.  Basically, each row in green is
being repeated for each row in fairway.

eg:

green has:
column1, column2
a, 1
b, 2
c, 3

fairway has:
column1, column2
x, 1
y, 2
z, 3

select * from green, fairway;

results:

a, 1, x, 1
a, 1, y, 2
a, 1, z, 3
b, 2, x, 1
b, 2, y, 2
b, 2, z, 3
c, 3, x, 1
c, 3, y, 2
c, 3, z, 3

What you really want to do is JOIN these two tables together, such as
in this:

select
  green.column1 as green,
  fairway.column1 as fairway,
  column2
from
  green
  join fairway
    using (column2)
;

results:
green,fairway
a, x, 1
b, y, 2
c, z, 3

    Thanks,

        Stephen

Вложения

В списке pgsql-general по дате отправления:

Предыдущее
От: Merlin Moncure
Дата:
Сообщение: Re: PosttgreSQL on AIX
Следующее
От: Stephen Frost
Дата:
Сообщение: Re: libreadline and Debian 5 - not missing just badly named