[Dbix-class] Branch for supporting column attributes checked in

[Matt S Trout]

John Napiorkowski wrote:
> Hi,
> 
> I checked in a branch off of current under
> branches/DBIx-Class/param_bind which alters how SQL
> statements are executed so that you can specify bind
> parameters in your schema classes.
> 
> This is sorely needed by people trying to get
> Postgresql BYTEA and Oracle BLOB columns to work
> without having to hack around.
> 
> The summary of the problem is that typically when you
> bind a value to a placeholder in your SQL statement,
> that value gets treated like a VARCHAR.  For some
> database and column types this is a problem,
> particularly if the column is for a binary type. 
> Typical issues are that the value will get truncated
> at some length, or the sql statement will just fail.
> 
> When using straight DBI, people with these kinds of
> problems can bind the column type as a hashref:
> 
> $sth->bind_param(1, 'value', { pg_type =>
> DBD::Pg::PG_BYTEA });
> 
> This branch makes some key alterations to support this
> in DBIx::Class.  You can add the required attributes
> in you Schema:
> 
> __PACKAGE__->add_columns(
>   "media", { 
>   
>     data_type => "bytea", 
>     default_value => undef, 
>     is_nullable => 1, 
>     bind_attributes => { 
>       pg_type => DBD::Pg::PG_BYTEA },
>   },
> );
> 
> If you have need for this please check it out and see
> if it works for you.  This currently solves my
> particular itch as well as passed all my other tests,
> but I am hoping to get feedback on the best way to
> make this suitable for wider consumption.
> 
> The biggest change I've introduced is that
> SQL::Abstract is initialised with
> "bindtype=>'columns'", which means that the array of
> binds returned will be an arrayref of the column name
> and the value instead of just an array of values.  So
> if you are using the SQL::Abstract inside of
> ..Storage::DBI then this change will affect you.  I
> think I got all the places but I might have missed
> something.
> 
> In order to get the actual bind attributes into the
> storage object I had to mess with ..ResultSet->update
> and ..Row->update|insert.

That's horrific. When we were discussing this I said I thought that the 
$source object was passed and that if it wasn't, it should be. Turns out it 
isn't, so the solution is that it should be - i.e. in Row->update

   my $rows = $self->result_source->storage->update(
                $self->result_source->from, \%to_update, $ident_cond);

would become

   my $rows = $self->result_source->storage->update(
                $self->result_source, \%to_update, $ident_cond);

Also, I see absolutely no reason to pollute the column_info with DB-specific 
crud - why can't Storage::DBI::Pg detect 
$source->column_info($col)->{data_type} being 'bytea' and DWIM?

-- 
      Matt S Trout       Offering custom development, consultancy and support
   Technical Director    contracts for Catalyst, DBIx::Class and BAST. Contact
Shadowcat Systems Ltd.  mst (at) shadowcatsystems.co.uk for more information

+ Help us build a better perl ORM: http://dbix-class.shadowcatsystems.co.uk/ +