Re: A worst case for qsort

Поиск
Список
Период
Сортировка
От Fabien COELHO
Тема Re: A worst case for qsort
Дата
Msg-id alpine.DEB.2.10.1408062105240.30492@sto
обсуждение исходный текст
Ответ на Re: A worst case for qsort  (Peter Geoghegan <pg@heroku.com>)
Список pgsql-hackers
> Random pivot selection will certainly result in more frequent lopsided 
> partitions without any malicious intent.

Yep. It makes "adversary input" attacks more or less impossible, at the 
price of higher average cost. Maybe a less randomized version would do, 
i.e. select randomly one of the "3" medians found, but would keep some 
nice better performance property. It would need proof, though.

-- 
Fabien.



В списке pgsql-hackers по дате отправления:

Предыдущее
От: Heikki Linnakangas
Дата:
Сообщение: Re: posix_fadvise() and pg_receivexlog
Следующее
От: Fabien COELHO
Дата:
Сообщение: Re: add modulo (%) operator to pgbench