[Dbix-class] RecursiveUpdate and many to many

[Pedro Melo]

Hi,

On Sep 30, 2008, at 1:25 PM, Matt S Trout wrote:
> On Mon, Sep 29, 2008 at 05:39:25PM +0100, Pedro Melo wrote:
>> On Sep 29, 2008, at 4:50 PM, Matt S Trout wrote:
>>
>>> On Sun, Sep 28, 2008 at 05:56:11PM +0100, Oliver Gorwits wrote:
>>>> Ideally, DBIC would instead populate the relationship_info details
>>>> after someone has set up __PACKAGE__->many_to_many(...). That would
>>>> avoid this guesswork.
>>>
>>> That would make no sense at all since many-to-many is NOT A
>>> RELATIONSHIP.
>>
>> I've been in this list for some years now that I've seen this phrase
>> written several times.
>>
>> <pedantic>
>> If I recall correctly my time at CS department where I was forced to
>> read the relational model (not the -for-dummies version that is SQL
>> today. but the true horrid version), many-to-many was in fact a
>> relationship.
>
> In DBIx::Class terms, that's not how we model it.

I know and understand that, but from the outside, can you understand  
that given that many_to_many is discussed in  
DBIx::Class::Relationships its only natural to people arriving at  
DBIx::Class to assume that many_to_many is a relation type like any  
other.

I'm not saying that they are right, I'm just saying that its one of  
the things that I teach to people who ask me about it: even with  
many_to_many defined in ::Relationship, its actually not. Just a bit  
of sugar.


>> Having said that, it is trivial to use your own base class for your
>> sources in a way that you can catch many-to-many declarations and
>> cache the result somewhere private (your own version of
>> relationship_info()).
>
> Please if you do that call it something -other- than  
> relationship_info.
>
>> I say that this is possible because I have code running in production
>> that does exactly that: creates a parallel relationship_info()
>
> Much DBIC code and extension stuff assumes that anything in there is
> a relationship in the DBIC sense.
>
> if you start putting many-to-many stuff in there, your code is  
> broken. It
> just might not have caused you any bugs yet.

I did not start to put many_to_many in relationship_info(), I used a  
different name for my relationship introspection code.
relationship_info() is critical for *_related() to work, and I didn't  
want to mess that up.

In case my point was not clear enough, I agree that  
relationship_info() should only return relations that can be used in  
*_related() methods.

But if we have a documented way (cf. ::Relationships) to declare all  
this types, I think it makes sense to have a method (a different one  
most likely) to return the information about all the relationships you  
declared, including many_to_many. This makes it easier to do a lot of  
things (like the current recursive insert/update stuff, or a  
documentation generation tool).

Best regards,
-- 
Pedro Melo
Blog: http://www.simplicidade.org/notes/
XMPP ID: melo at simplicidade.org
Use XMPP!