Re: Runtime Database connections- one more time

[prev] [thread] [next] [Date index for 2005/01/23]

I have been using Class::DBI for some time now and it has been kind of 
disturbing not to get a real foundation for how to treat the case when I 
need to use separate databases for my clients though they are using the 
same application.

I am using Class::DBI in a Apache-mod_perl-HTML::Mason 
environment(mod_perl 1). Nothing else.

My current configuration:
a mod_perl authentication module handle user login and puts the relevant 
information in $r->pnotes('ticket')

my HTML::Mason module calls Module::DBI containing:

/*---------------------------------------------------------------
package Module::DBI;
use 5.006;
use strict;
use warnings;
use base 'Class::DBI';
our $VERSION = '0.01';

sub db_Main {
   DBI->connect_cached(shift->get_user_connection);
}

sub get_user_connection {
   my $class=shift;
   my $r = $HTML::Mason::Commands::r;
   if ($r) {
     my $ticket=$r->pnotes('ticket') || $r->prev->pnotes('ticket');
     if ($ticket) {
       my $company    = $ticket->{company};
       my $dbuser     = $ticket->{dbuser} || $company;
       my $pw         = $ticket->{pw}||'';
       my $dbprefix   = $r->dir_config("dbprefix") || '';
       my $database   = $dbprefix . $company;
       my $db         = "DBI:mysql:database=" . $database;
       $db.= ";host=" . (($r->dir_config("dbhost")) ? 
$r->dir_config("dbhost") :
			"localhost");
       $db.= ";port=" . $r->dir_config("dbport") if 
$r->dir_config("dbport");
       return(
	     $db,
	     $dbuser,
	     $pw,
	     {AutoCommit => 1, RaiseError => 1, RootClass => 
"DBIx::ContextualFetch"}
	     );
     } else {
       #
       $class->_croak('Wow, no ticket in request object!');
     }
   } else {
     $class->_croak('Wow, no request object!');
   }
}
---------------------------------------------------------------*/

Is this the optimal way?
Many clients = many sleeping database processes?

/Bosse



Perrin Harkins wrote:
> Daniel Collis Puro wrote:
> 
>> It works from the command line, and feels MUCH cleaner than using
>> "our()" . .
> 
> 
> It's identical to your previous approach, except that you are now 
> re-defining a method on every single request.  That looks much less 
> clean to me.
> 
>> any mod_perl concerns?
> 
> 
> You will probably cause more memory to be unshared by constantly 
> re-defining this method.
> 
> I still don't see why you need any of this.  Why can't you make your 
> Class::DBI subclasses get their connection info from the same place as 
> your other classes?  If the connection parameters are identical, it will 
> reuse an open connection.
> 
> - Perrin

        -- 
        Bo Strinnholm
www.payroll.se
Tel: +46 (0)586 725 724, +46 (0)730 639 573