[Dbix-class] Re: as_subselect_rs->delete bug?

[Toby Corkindale]

On 11 July 2012 23:35, Rob Kinyon <rob.kinyon at gmail.com> wrote:
> On Wed, Jul 11, 2012 at 6:11 AM, fREW Schmidt <frioux at gmail.com> wrote:
>> I wrote a test that confirms it, but it actually has no where clause at all
>> when I test it, which is even scarier.
>
> Before we go ahead confirming failures and making fixes, let's step
> back and discuss exactly what
> $rs->search(...)->as_subselect_rs->delete should actually do and if it
> should be something that's allowable. The point behind as_subselect_rs
> was to allow a resultset to be used as the RHS for a column in a
> search clause. If $rs->search() could detect that a $rs is the RHS,
> then ->as_subselect_rs() would never be called by a user. More to the
> point, the resultset returned by as_subselect_rs, when I designed the
> subquery thing, was never meant to be used as a general-purpose $rs,
> but only in specific contexts.
>
> So, it makes perfect sense that calling ->delete() on this
> special-purpose $rs wouldn't do the right thing. I am really curious
> as to the chain of reasoning that led to the use of
> ->as_subselect_rs().
>
>> Toby, as a workaround in the meantime if you remove the as_subselect_rs it
>> seems to do the right thing.
>
> Of course it does the right thing. The as_subselect_rs is an error in this case.

It looks clearly wrong in the example I gave, but inside a working
program you tend to be passing around variables. You can confirm that
they contain a ResultSet object, sure, but you don't know much more
than that about them.
But if you have a resultset, you assume you can ->search, ->count, and
even ->delete it.
What I'm saying it -- you don't always know that someone, somewhere
else in the code, has started calling your methods with resultsets
generated by ->as_subselect_rs.
But if you call ->delete on that $rs variable, weird and unexpected
things happen to the data inside your database.

So.. while it is easy to say that I should just remove as_subselect_rs
from the query, please consider that DBIC should throw an error if
this condition occurs.

thanks,
Toby