Comments on: the “anchor adapter”

By: Kevin Rutherford

Kevin Rutherford — Mon, 21 Dec 2009 20:28:36 +0000

Not the constants themselves, perhaps. But anything that uses them will be more difficult to test.

By: Ashley Moran

Ashley Moran — Sun, 20 Dec 2009 19:43:15 +0000

Do you think that constants that reference value types (and class-method constructors for them) also add friction?

By: Kevin Rutherford

Kevin Rutherford — Sun, 19 Jul 2009 13:52:51 +0000

I tend to think of the gloabl space as somewhere for the runtime to put things so I can get at them from anywhere; it’s certainly not somewhere I want to put any of my own code if I can possibly help it.

By: Ashley Moran

Ashley Moran — Sun, 19 Jul 2009 12:26:04 +0000

No, I can’t think of any off the top of my head. How about this as an experiment: from now on (well, on some project it might work on) I will create only one constant of my own, to a class with exactly one class method, which will provide an object from which everything else can be generated.

Is that feasible, or is that just plain crazy?

By: Kevin Rutherford

Kevin Rutherford — Sun, 19 Jul 2009 12:12:26 +0000

Well, new would have to be redesigned, but that aside I don't foresee any problems. Do you have a specific case in mind where class methods can't be avoiced?

By: Ashley Moran

Ashley Moran — Sun, 19 Jul 2009 10:16:24 +0000

I’ve gone through the “how would you write this if Ruby didn’t have inheritance?” thought process, and concluded that the answer is always, “better”. Now I don’t use (class) inheritance any more.

So, as an idle question… how would you write code if Ruby didn’t have class methods?