[Dbix-class] breaking change ?

[Peter Rabbitson]

Gerrit Wichert wrote:
> hello,
> 
> im using the following code  for quite some time (about a year) to get
> values of two related rows of the same table.
> This worked very well. But since the last upgrade of DBIx:_Class to
> version 0.08107 it failed with the following error:
> 
> DBIx::Class::ResultSet::next(): Column id not loaded or not passed to
> new() prior to insert() on
> GS::Model::TimeControl::DBIx::TimeSheet=HASH(0x9707b0c) trying to
> resolve relationship (maybe you forgot to call ->discard_changes to get
> defaults from the db) at ...
> 
> changing the select from:
>    select => [ qw/ me.login me.host me.type me.mtime end.mtime /],
> to
>    select => [ qw/ me.login me.host me.type me.mtime end.mtime me.id/],
> makes it work again.
> So is seems to fail when not including the primary key into the result.
> 
> Is this really necessary to select the primary key?
> Is it a bug or a breaking change or can i write such requests in a
> better way?
> 
> 
> And btw. did anyone know of a db-portable way to deflate DateTime values
> in where statements?
> 
> thanks,
>   Gerrit Wichert
> 
> 
> 
> 
> 
> package GS::DBIC::Schemata::TimeControl::TimeSheet;
> 
> use strict;
> use warnings;
> 
> use base qw/DBIx::Class/;
> 
> __PACKAGE__->load_components(qw/ InflateColumn::DateTime Core /);
> __PACKAGE__->table('time');
> __PACKAGE__->add_columns( qw/ id login type pid host /);
> __PACKAGE__->add_columns(  mtime =>{data_type =>'TIMESTAMP'},);
> __PACKAGE__->set_primary_key('id');
> 
> __PACKAGE__->belongs_to(
>     'end',
>     __PACKAGE__,
>     {'foreign.pid'=>'self.id'},
>     {'join_type' => 'left'}
>     );
> 1;
> 
> 
> 
> ...
> my $rs = $schema->resultset('TimeSheet')->
>           search( {'me.login' => $employeeId,
>                          'me.pid'   => {'=' => 0},
>                          'me.mtime' => {
>                                                     '>'  => $date->ymd,
>                                                     '<'  =>
> $date->clone->add(days=>1)->ymd
>                                                    }
>                         },
>                         { join   => [qw/ end /],        
>                           select => [ qw/ me.login me.host me.type
> me.mtime end.mtime /],
>                         });
> 
> while (my $entry = $rs->next) {
> ...
> }  
> 

Hmmm... this seems like a generalization of another regression that
currently lives in the unresolvable_prefetch/ branch, but nobody had
to look at it yet. Your example will be added as a test, and all of
this will hopefully be fixed in a week or two. Until then the only
solution is to explicitly select the PK.