Re: index on user defined type
От | Jeff Davis |
---|---|
Тема | Re: index on user defined type |
Дата | |
Msg-id | 1105582355.2886.396.camel@jeff обсуждение исходный текст |
Ответ на | Re: index on user defined type (Tom Lane <tgl@sss.pgh.pa.us>) |
Ответы |
Re: index on user defined type
|
Список | pgsql-general |
On Wed, 2005-01-12 at 18:12 -0500, Tom Lane wrote: > Stephan Szabo <sszabo@megazone.bigpanda.com> writes: > > I'm wondering if the function under = is an SQL function being inlined. > > Bingo --- that's surely it. After inlining, the expression would no > longer look like it matched the index. > > You don't want to use SQL functions to define indexable operators > anyway. They leak memory, and are slow, and neither of those are > good properties for an index support function. > > regards, tom lane Thanks very much guys. I got confused and tried to simplify it to a simple test case and confused myself more when it still didn't work. Just to be sure I redid it in plpgsql with no problems. I understand that index operations should be written in C for production use, or at least tested in C to see if it helps matters. Comparison operators are of course very simple to write in any language anyway. I attached a proposed documentation patch. I wasn't able to readily see the implications of writing a function in SQL regarding an index, so perhaps this will help someone in the future (not that many people will attempt writing index access methods in SQL, but someone could get confused like I did). Regards, Jeff Davis
Вложения
В списке pgsql-general по дате отправления: