All code examples are in Java though. Since Java ships with more concurrency related helper classes than other languages you might find some of the examples difficult to translate.

Actually, Perl thankfully doesn’t suffer too much from this; not nearly as much as Ruby, anyway. It’s probably because the language isn’t OO from top to bottom so you can’t do some of the unspeakable things Rubyistas are doing.

And I do think it’s necessary that a language allow re-opening classes… but good grief, it just shouldn’t be used in regular code! The facility is there to permit metaprogramming and possibly for when you’re on a deadline and have to deal with a brokenly designed library, in which case it gives you a bad coping strategy rather than no coping strategy.

But it boggles my mind that the problems with wanton exploitation of open classes aren’t blindingly, searingly obvious. I mean, Rake stuffs a kitchensink of filesystem methods into Object, which breaks Net::SFTP’s reliance on method_missing to delegate methods to the right backend module, which leads Assaf Arkin to conclude that method_missing should be avoided. What the …?

People need to find ways to get private method semantics out of their favourite language, and use them.

Rest of the comment distilled into another entry.

In fact, I think the typeless nature of PHP is the single worst thing about it. And something I use with glee! </hypocrite>

Oh by the way - nice to see more frequent posting here again.

I didn't know about that (I've only scratched the surface of python). I don't think I'd like to see it in Java though. It is essentially a macro, and although it is limited to a single specific use, the purist in me says "No". Using it would nullify the benefits of the unique package based naming convention for Java which you described.

Also I can see it being abused in horrible ways. Perhaps if it were limited such that it were ONLY allowed in cases where two classes of the same name are imported.

Given the relative rarity of class naming conflicts in Java (in my experience anyway. I almost always give my classes names with at least two words), a little verbosity in differentiating types where they appear (variable declarations and function headers mostly, perhaps the odd cast) it doesn't seem like a big deal. My two cents.

You could declare an inner class which extends the class with the really long name.

To use your example with the AWT List and the List data structure:

package my.domain.my.package;

public class MyClass
{
public class AWTList extends java.awt.List{};
public class UtilList extends java.util.List{};

... rest of class code
}

At least, that's how it is on mine :)

/Jonas