Re: deleting referenced data

Bruno Wolff III wrote:

>On Wed, Aug 27, 2003 at 00:09:34 +0200,
>  Andreas Fromm <Andreas.Fromm@physik.uni-erlangen.de> wrote:
>
>
>>Bruno Wolff III wrote:
>>
>>
>>
>>>On Tue, Aug 26, 2003 at 18:34:04 +0100,
>>>Richard Huxton <dev@archonet.com> wrote:
>>>
>>>
>>>I don't think that is what he wants. He seems to want to go in the other
>>>direction. That is going to require custom triggers and probably adding a
>>>reference count column to data-tables.
>>>
>>>
>>>
>>>
>>This is what I want to do, but my question is how can I easely now if
>>there is someone referencing a certain record. Is it just possible by
>>searching every table that could reference it or is there an other way
>>of doing it without adding an reference count column?
>>
>>
>
>You can use a reference count in the row you are interested in. You
>would need to create triggers on all referncing tables to update the
>reference count as necessary. Also as pointed out previously, you
>want to use serialized isolation (and handle retries) or exclusively lock
>the table (and worry about deadlocks and a concurrency bottleneck)
>if you use this approach.
>
>You could do a scan of all possible referencing tables whenever you update
>any of the referencing tables. However, this approach has pretty much the
>same problems as using a refernce count and will probably be slower.
>
>
As the newbe I am, I don't understand much abaut serialized isolation,
db-locks and so on.
Do I need to care abaut all these considerng that the database will only
be accessed by one user at a time, and the hole db will not have many
entries?
Are there basic dessign considerations that should be taken, or is it
something that can be added in the future?

As I don't have much time to invest in this project, I will have to come
up with some other solution. Maybe something like a 'garbage collection'
routine, that runs over the db from time to time and delets all
unreferenced records.

--
Andreas Fromm