Re: Allowing update of column only from trigger
От | Shawn Harrison |
---|---|
Тема | Re: Allowing update of column only from trigger |
Дата | |
Msg-id | 41FE7A0E.8070900@tbc.net обсуждение исходный текст |
Ответ на | Re: Allowing update of column only from trigger (PFC <lists@boutiquenumerique.com>) |
Список | pgsql-general |
PFC wrote [01/28/05 7:08 PM]: > > First you should use a ltree type for the uri field : Yes, it would be very good if I could use ltree rather than rolling my own, but ltree doesn't meet my use requirements. > - you write it foo.bar instead of /foo/bar That is the problem: The point is to make look-up based on uri or filesystem path very efficient. > - there are operators on ltree types to express "is parent of", "is > children of" > - these operators are indexed I have written similar pure-SQL functions for my parent/uri system, though without the syntax shortcuts of the ltree operators. > [...] > Now about your update problem, when you rename foo.bar into > foo.crum.bar you could, in a single update, replace all foo.bar by > foo.crum.bar in all your table with the ltree operators and special > functions. That's an approach I hadn't considered. Hmmm, I'd have to read the ltree C sources to see how it does this.... > And for your checks, you can add a CHECK on the url field to be sure > it's equal to the url of the parent + the name of the current row. > It'll make one more SELECT request, though. Yes, I'd like to avoid extra selects as much as possible. Thank you very much for your interaction -- very helpful. -- ________________ harrison@tbc.net
В списке pgsql-general по дате отправления: