Two options I've found: 1. c.parent.reload 2. c.clear_association_cache Both do what you would imagine. It feels like changing the underlying id should force this sort of behavior on the next access of the parent, but it doesn't (yet?).
- Michael On 10/12/07, Jordan A. Fowler <[EMAIL PROTECTED]> wrote: > Nathan, > > That's a class method to reset the column information in the event that it > actually changes. Such as when you've made changes to your schema (like in a > migration) and need to have your ActiveRecord model class to reflect the > correct columns. He's talking about caching on a per-object basis. > > I should have a decent response soon... > > --Jordan > > > > On Oct 12, 2007, at 10:33 PM, Nathan Colgate Clark wrote: > > This might help: > > Parent.reset_column_information > > http://api.rubyonrails.org/classes/ActiveRecord/Base.html#M001034 > > -N > > Glenn Little wrote: > It seems like rails is doing some data caching on objects > that is causing inconsistencies. If you do a lookup through > a belongs-to, that result persists even if you change the foreign > key (on the memory object without yet saving): > > parent has-many children > child belongs-to parent > > > > Parent.find(1).surname > => "Smith" > > Parent.find(2).surname > => "Jones" > > > > c = Child.find(1) > c.parent_id > => 1 > > c.parent_id = 2 > c.parent.surname > => "Jones" > > Okay, fine. What I wanted. Change has taken > effect. But now: > > > > c.parent_id = 1 > c.parent.surname > => "Jones" > > Not what I wanted. The earlier result is cached somewhere. > (I did the original id change from 1 to 2 just to show that > the cached value was *not* just whatever the initial database > value was). It's not a matter of "what's in the db versus > what has not been saved yet". It's more arbitrary. Whatever > value c.parent.surname reports is dependent on when you first > did a lookup, and is "stuck" from then on. > > So you get weirdness like this: > > > > c.parent.surname > => "Jones" > > Parent.find(c.parent_id).surname > => "Smith" > > Is this acceptable? If so, how does this not mean I need to ditch the idiom > "c.parent.surname" (nice) in favor of "Parent.find(c.parent_id).surname" > in any of my code where a value *might* have changed and it's possible > I *might* have done accessed through the association since then, and I > haven't saved the child record yet (for good reason, maybe I'm doing this > as part of validation)? > > Is there a way to force rails to not use (or mark as out-of-date) > the cached info? A "c.reload" does not work, because my c object > in memory may deliberately have different info than the database, > and the reload would wipe it. > > Thanks in advance! > > -glenn > > > > _______________________________________________ > Sdruby mailing list > [email protected] > http://lists.sdruby.com/mailman/listinfo/sdruby > > > > _______________________________________________ > Sdruby mailing list > [email protected] > http://lists.sdruby.com/mailman/listinfo/sdruby > > -------------------------------------------------------------- > Jordan A. Fowler > 2621 First Ave Apt 5 > San Diego, CA 92103 > E-mail: [EMAIL PROTECTED] > Website: http://www.jordanfowler.com > Phone: 406.546.8055 > > > _______________________________________________ > Sdruby mailing list > [email protected] > http://lists.sdruby.com/mailman/listinfo/sdruby > > _______________________________________________ Sdruby mailing list [email protected] http://lists.sdruby.com/mailman/listinfo/sdruby
