Re: [HACKERS] No: implied sort with group by

Поиск
Список
Период
Сортировка
От Bruce Momjian
Тема Re: [HACKERS] No: implied sort with group by
Дата
Msg-id 199801271654.LAA13410@candle.pha.pa.us
обсуждение исходный текст
Ответ на Re: [HACKERS] No: implied sort with group by  ("Thomas G. Lockhart" <lockhart@alumni.caltech.edu>)
Список pgsql-hackers
>
> > > > Does the SQL standard say anything about an implied sort when
> > > > grouping or is it up to the user to include an ORDER BY clause?
>
> Up to the user. SQL is a set-oriented language. The fact that many/most/all
> implementations order results to then do grouping is an implementation
> detail, not a language definition.
>
>
> > This is what I think is missing or broken right now.
> >
> > > > select * from t1;
> > >          a b  c
> > >          1    x
> > >          2    x
> > >          3    z
> > >          2    x
> > >
> > > 4 row(s) retrieved.
> > > > select b,c,sum(a) from t1 group by b,c;
> > > b  c             (sum)
> > >
> > >    x                 5
> > >    z                 3
> > >> 2 row(s) retrieved.
>
> Sorry, I've lost the thread. What is broken? I get this same result, and
> (assuming that column "b" is full of nulls) I think this the correct result.

At one point, it was thought that NULLs shouldn't be grouped, but I
backed out the patch.  There is a problem with GROUP BY on large
datasets, and Vadim knows the cause, and will work on it later.

--
Bruce Momjian
maillist@candle.pha.pa.us

В списке pgsql-hackers по дате отправления:

Предыдущее
От: "Thomas G. Lockhart"
Дата:
Сообщение: Re: [HACKERS] No: implied sort with group by
Следующее
От: Brett McCormick
Дата:
Сообщение: functions with same name, different args