what's the problem with the question?

I linked the relevant docs.

One place where I've seen this technique is a ORM package on CPAN - Class::Tables - which auto-generates classes from table names and makes colum names into their methods using AUTOLOAD:

sub AUTOLOAD {
    my $self = shift;
    (my $func = $Class::Tables::AUTOLOAD) =~ s/.*:://;
    
    croak qq{Can't locate object method "$func" via package "$self"}
        unless ref $self and UNIVERSAL::isa( $self, "Class::Tables" );

    unshift @_, $self, $func;
    goto &field;
}
[download]

The field() sub translates its arguments to the appropriate SQL call against the database. I've used that package very much, and I have seen no adverse side effects. And no, I'm not aware of a cleaner way to do this.

I'm still puzzled by the meaning of "AUTOLOAD is localized to a blocks context" part of your question. Does that mean having different lookup-results (depending on scope) for the last-resort AUTOLOAD sub?

Sorry if I'm just being too dumb or having the wrong kind of humour. ;)

perl -le'print map{pack c,($-++?1:13)+ord}split//,ESEL'

Thanks the example from Class::Tables is interesting. :)

And I now understand goto &$method_ref seems to follow the @ISA chain.

It doesn't, this was my brain fart. =)

It's the ->can() which already returned the correct code_ref after following the inheritance chain

And that's highlighting the difference to Class::Table, $func is a string not a ref.

Additionally goto &field is not a dynamic call, it will always call the same sub field in the same package.

Though ->field could be called as a method form other points, since it expects $self as first argument.

Cheers Rolf
_{(addicted to the Perl Programming Language :)
Wikisyntax for the Monastery}

what's the problem with the question I linked the relevant docs.

No problem with your question, but probably with my answer. Sorry about that.

perl -le'print map{pack c,($-++?1:13)+ord}split//,ESEL'