Re: Re: (Zigster) Re: Re: Perl and Objects, how do you resolve the two?

There is a FarSide cartoon of a court scene with a lawyer interogating a cow in the witness box: "We want to know why, Brown Cow. We know how now, brown cow.".

It seems odd to me that zigster said "I think AUTOLOAD is bad for munging accessor methods and mutators" and you reply in, so many words...."no you don't understand, this is how it's done". I bet you could write a similar function in Java with clever use of exceptions. you wouldn't want to do it in a Java though, it'd be a hack (If it's a hack in Java then....)

The addition of the exists function kinda highlights zigster's argument (although I fear the repecussions of Scott's opinion being affirmed, it's not something we like to encourage ;). It's like using AUTOLOAD to do overload functions on strict datatypes by using a regex to check the content of parameters. The difference is one is necessary and the other is unnecessary (but admirably lazy ;0)

--

Brother Frankus.

Comment on Re: Re: (Zigster) Re: Re: Perl and Objects, how do you resolve the two?

Replies are listed 'Best First'.
Re (tilly) 4: Re: Perl and Objects, how do you resolve the two? by tilly (Archbishop) on Apr 11, 2001 at 22:02 UTC
I think you missed stephen's point. zigster's argument against using AUTOLOAD was that (by his understanding) it tied the methods to the object's implementation and that exposed internal information that you shouldn't expose. There are two major things wrong with that argument. My response was that if you choose to tie the autogenerated methods to the implementation, when you change the implementation it is not hard to add normal methods that maintain the old interface even though the implementation changed. stephen's response is that zigster's assumption that you had to tie the interface to the implementation was wrong. And he replied with actual code showing how to autogenerate methods based on a data structure that can be used to control the interface independently from the internal data structure. So yes, stephen is munging accessor methods and mutators with AUTOLOAD. However he has done it in a way where changes to the structure of the object do not show up in the interface. As an incidental note, I have used AUTOLOAD for a variety of things, but I have not yet used AUTOLOAD as a way to produce accessor methods and mutators for end-user objects. However I have used it as a replacement for auto-generated SWIG tie implementations (IOW for objects that Perl is using) with lazily implemented run-time accessors. It worked well. I got both loading and runtime to be substantially faster than the original. Maintainability was the same. Based on that I have to say that I find AUTOLOAD to be no more or less maintainable than any other way of auto-generating code. If you think about trying to write self-modifying assembly you would think that auto-generated code is the worst idea in the world. However then you go out and use a compiler that does what? Oh right, it autogenerates code to run later! AUTOLOAD should have a huge sign on it saying, "use with extreme caution", but it doesn't need one saying, "you should never use this". That said I use it in only a tiny fraction of my code. But that fraction tends to be key code which a lot of other code builds off of...	[reply]

Replies are listed 'Best First'.

Re (tilly) 4: Re: Perl and Objects, how do you resolve the two?
by tilly (Archbishop) on Apr 11, 2001 at 22:02 UTC

stephen

zigster's argument against using AUTOLOAD was that (by his understanding) it tied the methods to the object's implementation and that exposed internal information that you shouldn't expose.

There are two major things wrong with that argument.

My response was that if you choose to tie the autogenerated methods to the implementation, when you change the implementation it is not hard to add normal methods that maintain the old interface even though the implementation changed.

stephen's response is that zigster's assumption that you had to tie the interface to the implementation was wrong. And he replied with actual code showing how to autogenerate methods based on a data structure that can be used to control the interface independently from the internal data structure.

So yes, stephen is munging accessor methods and mutators with AUTOLOAD. However he has done it in a way where changes to the structure of the object do not show up in the interface.

As an incidental note, I have used AUTOLOAD for a variety of things, but I have not yet used AUTOLOAD as a way to produce accessor methods and mutators for end-user objects. However I have used it as a replacement for auto-generated SWIG tie implementations (IOW for objects that Perl is using) with lazily implemented run-time accessors. It worked well. I got both loading and runtime to be substantially faster than the original. Maintainability was the same.

Based on that I have to say that I find AUTOLOAD to be no more or less maintainable than any other way of auto-generating code. If you think about trying to write self-modifying assembly you would think that auto-generated code is the worst idea in the world. However then you go out and use a compiler that does what? Oh right, it autogenerates code to run later!

AUTOLOAD should have a huge sign on it saying, "use with extreme caution", but it doesn't need one saying, "you should never use this".

That said I use it in only a tiny fraction of my code. But that fraction tends to be key code which a lot of other code builds off of...

[reply]