[Dbix-class] Cascade deletes

[Peter Rabbitson]

Bill Moseley wrote:
> 
> 
> On Sat, Mar 6, 2010 at 3:59 AM, Peter Rabbitson <rabbit+dbic at rabbit.us
> <mailto:rabbit%2Bdbic at rabbit.us>> wrote:
> 
>     > I notice that DBIC does a select to find the related rows (i.e. in the
>     > Cd table when looking for rows to cascade delete).   Is there a reason
>     > that's done instead of just deleting directly?
>     >
>     > INSERT INTO artist ( label, name) VALUES ( ?, ? ): '1', 'test aritst'
>     > INSERT INTO cd ( artist) VALUES ( ? ): '25'
>     >
>     > DELETE FROM artist WHERE ( id = ? ): '25'
>     > SELECT me.id <http://me.id> <http://me.id>, me.year, me.name
>     <http://me.name> <http://me.name>,
>     > me.artist FROM cd me WHERE ( me.artist = ? ): '25'
>     > DELETE FROM cd WHERE ( id = ? ): '54'
>     >
>     > Why not just delete from cd where artist = 25?
> 
>     Because this will not invoke the various triggers that could be hidden
>     in lower classes (at the very least if you delete from an rs CDs then
>     dbic will not know to delete from Tracks).
> 
> 
> The delete Isn't it a recursive process?  Or are you saying the select
> is important for invoking the triggers?

If you look at the code in DBIx::Class::Relationship::CascadeActions
it will all make sense (the SELECT-causeing call is clearly visible)

> Also, I'm curious why the cascade deletes happen last instead of first?
>   That is why not work where $artist->delete first deletes all the
> artist's CDs (which first deletes all the CD's tracks, and so on), then
> finally deletes the artist.  That way there's no time when a row in one
> table (e.g. CD table) is referencing a deleted row (e.g. in Artist).

Because if (like in your case) dbic-side delete cascading is enabled,
we still want to give the database a chance to invoke its own triggers.
This is why first the CDs are deleted, and then if cascade is enabled -
anything else (if anything is left at all).

In the meantime we now correctly wrap delete/update in transactions
if deemed necessary: http://dev.catalystframework.org/svnweb/bast/revision?rev=8921

Cheers