Re: slow self-join query

Поиск
Список
Период
Сортировка
От Merlin Moncure
Тема Re: slow self-join query
Дата
Msg-id CAHyXU0w8D5UTe4XNN+s0+=kF=5-_xsyAiJFBJ2OTHkUeJUXkTg@mail.gmail.com
обсуждение исходный текст
Ответ на Re: slow self-join query  (Robert Poor <rdpoor@gmail.com>)
Список pgsql-performance
On Sun, Mar 18, 2012 at 10:57 PM, Robert Poor <rdpoor@gmail.com> wrote:
> On Sun, Mar 18, 2012 at 08:30, Scott Marlowe <scott.marlowe@gmail.com> wrote:
>> Why are you joining twice to the parent table?  If you're trying to
>> recurse without a with clause, then wouldn't you join the last table
>> to the one before it?
>
> I'm FAR from being an SQL expert; there's a significant chance that
> I'm not thinking about this right.  My intention for this query
> (slightly renamed since the original post):
>
>    SELECT F2.leader_id as leader_id, COUNT(U1.id) AS count
>      FROM users AS U1
> INNER JOIN user_associations AS F1 ON F1.follower_id = U1.id
> INNER JOIN user_associations AS F2 ON F2.follower_id = U1.id
>     WHERE F1.leader_id = 321
>  GROUP BY F2.leader_id
>
> is "among users that follow leader 321, who are the most widely
> followed leaders?", or more formally, find all the users that are
> followers of user 321 (via inner join on F1)  Of those users, tally up
> their leaders so we know which leaders are most popular.  Recall that
> the user_associations table is simply a user-to-user association:
>
>  create_table "user_associations", :force => true do |t|
>  t.integer  "follower_id"
>  t.integer  "leader_id"
>  end
>
> Is there a better way to do this?

hm. Something does not seem right with your query.  You're joining in
the same table twice with the same clause:

INNER JOIN user_associations AS F1 ON F1.follower_id = U1.id
INNER JOIN user_associations AS F2 ON F2.follower_id = U1.id

I think you meant to cascade through the follower back to the leader.
(maybe not..it's early monday and the coffee hasn't worked it's way
through the fog yet)...

Also, do you really need to involve the user table?  You're counting
U1.Id which is equivalent to F2.follower_id.

try this and see what pops out (i may not have the F1/F2 join quite right):
SELECT F2.leader_id as leader_id, COUNT(*) AS count
  FROM user_associations AS F1
  INNER JOIN user_associations AS F2 ON F1.follower_id = F2.leader_id
  WHERE F1.leader_id = 321
  GROUP BY 1;

merlin

В списке pgsql-performance по дате отправления:

Предыдущее
От: Robert Poor
Дата:
Сообщение: Re: slow self-join query
Следующее
От: "Kevin Grittner"
Дата:
Сообщение: Re: slow self-join query