When I update a child by setting the association method to nil, the
counter of the parent does not get decremented. For a code example see
the first answer to my ticket.
The problem seems to be that the @owner variable in
activerecord/lib/active_record/associations/belongs_to_association.rb
holds the child object, not the parent. That's why it cant have the
counter_cache_column and the test:
if @owner[counter_cache_name] and not @owner.new_record?
fails.
I really would appreciate anyone who gives this a try and reports the
results.
The original bug (and a patch) about this issue has been posted about
2 years ago!
Without an identity map this is very difficult to fix. The correct
fix for this bug would require full bi-directional has_many and
belongs/to associations, and could take a bit of work.
If someone wants to investigate this, I'd be happy to lend a helping
hand. On the other hand, there are some other changes brewing in that
area now, so perhaps it's worth waiting a bit?
Without an identity map this is very difficult to fix. The correct
fix for this bug would require full bi-directional has_many and
belongs/to associations, and could take a bit of work.
If someone wants to investigate this, I'd be happy to lend a helping
hand. On the other hand, there are some other changes brewing in that
area now, so perhaps it's worth waiting a bit?
I just removed @owner[counter_cache_name] from the check. If the
parent does not hold the counter_cache_column then counter_cache_name
would be nil. Since this is tested in the if statement too it should
not be a problem. The AR test suite passes without a failure...
Please give the fix +1
(http://dev.rubyonrails.org/ticket/10804)
_jan