Re: join keys redux

Cyrus Harmon wrote:
>  (clsql:update-objects-joins (list company1))
> [...]
> #<CLSQL-TESTS::COMPANY #x65FB5F6> has no slot named 
> (CLSQL-TESTS::COMPANYID CLSQL-TESTS::GROUPID).

Yes, I believe that you are correct. update-objects-joins does not
currently support joining with multiple keys.

> As for the point about the dog owners, yeah, I can see the merits of 
> doing it as you describe, but one of the nice things about the OODDL is 
> that it has the potential to hide most selects and do them under the 
> covers. I think this is a good thing. You can argue that the overhead 
> of the new class doesn't warrant this, but I'd argue that the ability 
> to hide this stuff behind the CLSQL MOP allows one to do some really 
> cool meta-data based stuff with automatic joining and updating. I'd 
> just as soon see as much of the logic pushed down into the 
> def-view-class and be able to leave out the kind of select Kevin 
> proposes, which is admittedly a decent workaround.

I think supporting your selects with def-view-class is a reasonable
idea. I doubt that I'd use that behavior (in fact, I rarely use the
object-oriented interface of CLSQL), but adding that might be a
worthwhile improvement for some.

Kevin