...or... have _load() return a newly created "Factory" object (we'd
have to create such a class, because I don't think such a class currently
exists within TT2) instead of a string. Then _load could set an attribute
of the factory object.
then factory->new() would do something like..
sub new {
my $self = shift;
my $native_factory = $self->native();
if ($self->is_plugin) {
$native_factory->new($self->context, $self->args);
}
else {
$native_factory->new($self->args);
}
}
hmmm... even better yet.
have _load() create and return a Factory object (which would be either a
Factory::NonPlugin or a Factory::Plugin object (both of which would
inherit from a common base Factory object)
Define Factory::NonPlugin::new() like so...
sub new {
my $self = shift;
my $native_factory = $self->native(); # the string currently returned
# by _load()
$native_factory->new($self->args);
}
and Factory::Plugin::new like so...
sub new {
my $self = shift;
my $native_factory = $self->native(); # the string currently returned
# by _load()
$native_factory->new($self->context, $self->args);
}
then from fetch()
$plugin = $factory->new(@$args);
becomes
$plugin = $factory->new(@$args);
and everything is "self-aware"
Quoting Craig Barratt ([EMAIL PROTECTED]):
> > While running ttree I've run into what appears to be a bug when
> > using the LOAD_PERL option. I'm using v2.04
>
> This does appear to be a bug. Your analysis looks correct. Nice work!
>
> There are several possible fixes, but I'm not sure which is most
> elegant.
>
> One way is to make _load return a flag that says whether it is a
> plain perl module or not. fetch() could then save this flag in a
> new hash (eg: $self->{ FACTORY_IS_PLAIN_MODULE }->{ $name }). Or,
> based on the _load return flag, fetch() could save plain modules
> in $self->{ FACTORY_PLAIN }->{ $name }, and normal modules in
> $self->{ FACTORY }->{ $name }. In either case, shifting @args
> should be removed from _load(), and fetch() only needs to
> unshift $context onto @args for the case of a real plugin.
>
> Craig
