Re^8: Autoloading tie routines

The eval was inherited (in the older meaning of the term) from the predecessor package, which was written in 1998. I've happily changed the second-last code line to *$AUTOLOAD = sub{$val};. In the next statement, I forgot that the object of using a constant name was to get the constant's value, which is why the goto is needed. I may be getting too old for this.

I was not familiar with "God module/antipattern" terminology. Having been in programming since '63 I value modularity highly, and would never create/architect a module like the one I'm working to replace.

But having inherited (in the classic sense) a bloated package, and knowing how much compatibility means to many people, at the moment I'm keeping compatibility. Not that the older package probably has many users: the problems I know about probably preclude that.

On the other hand, having 8 packages inherit from XYZ::ABC may be OK this once, seeing as the only thing "in" each package is a 1;.

Last question for now: to pursue the compromise course, my instantaneous version of this AUTOLOAD contains:

                my $subname = lc($mmx) . '_' . lc($sah) . '_' . lc($fu
+nction);
                *$AUTOLOAD = sub {
                    my $base_sah_ref = shift;
                    # dereference the base "sah"
                    unshift @_, $$base_sah_ref;
                    goto &$subname;  # or &{*$subname{CODE}} ??
                };
                goto &$AUTOLOAD;
[download]

Is Perl smart enough to capture the current value of $subname in the anonymous sub, rather than use a reference to a local variable that will soon go away? Is this one of those "reference count" things wherein perl is superior to C?

cmac
www.animalhead.com

Comment on Re^8: Autoloading tie routines Select or Download Code

Replies are listed 'Best First'.
Re^9: Autoloading tie routines by tilly (Archbishop) on Feb 03, 2009 at 23:14 UTC
The goto is not needed, you can just return $val. To answer your question about your later code sample, since you are using variables declared with my, Perl will correctly capture the current value in the anonymous subroutine. This is called a closure. To address an issue commented on in your code comment, Re: what is the difference between a and a{GLOB}? explains why &$subname is the same as &{$subname{CODE}}. Going back to closures, a classic book to learn programming techniques using them is MJD's book Higher Order Perl. Or see Why I like functional programming (long) and Re (tilly) 1 (perl): What Happened...(perils of porting from c) (short) for a couple of examples I wrote that you can puzzle through. I'm a fan of techniques that use closures, but only* if I am working with people who understand them. There is a definite learning curve in getting used to them. Closures can be used in most languages which support lexical scope and anonymous subroutines. That list includes most current dialects of Lisp (the one in emacs is the major exception), most scripting languages (including Perl, JavaScript, Ruby and Python though Python makes it harder than it should be), and virtually any language that calls itself a functional language. Closures are now making their way into mainstream languages. (They are in C#, and are a likely future addition to Java.)	[reply]

Replies are listed 'Best First'.

Re^9: Autoloading tie routines
by tilly (Archbishop) on Feb 03, 2009 at 23:14 UTC

goto

To answer your question about your later code sample, since you are using variables declared with my, Perl will correctly capture the current value in the anonymous subroutine. This is called a closure.

To address an issue commented on in your code comment, Re: what is the difference between *a and *a{GLOB}? explains why &$subname is the same as &{*$subname{CODE}}.

Going back to closures, a classic book to learn programming techniques using them is MJD's book Higher Order Perl. Or see Why I like functional programming (long) and Re (tilly) 1 (perl): What Happened...(perils of porting from c) (short) for a couple of examples I wrote that you can puzzle through. I'm a fan of techniques that use closures, but only if I am working with people who understand them. There is a definite learning curve in getting used to them.

Closures can be used in most languages which support lexical scope and anonymous subroutines. That list includes most current dialects of Lisp (the one in emacs is the major exception), most scripting languages (including Perl, JavaScript, Ruby and Python though Python makes it harder than it should be), and virtually any language that calls itself a functional language. Closures are now making their way into mainstream languages. (They are in C#, and are a likely future addition to Java.)

[reply]