This page is no longer maintained — Please continue to the home page at

Should the linearization order be changed to put the derived class first?

No replies
Joined: 2011-10-26,
User offline. Last seen 51 weeks 12 hours ago.


The way I understand things right now, if I have a hierarchy:

trait A
trait B
class C extends B with A

The linearization order is C->A->B. This makes intuitive sense, but I
often find myself wanting to write code like

trait A {
   def things: List[Int]

trait B {
   abstract override def things = 7 :: super.things

val a = new B {
   def x = List(1, 2, 3)

where a.x is now List(7, 1, 2, 3)

In fact, maybe I'm doing things wrong, but I actually want to do this
even more than I want to use regular mixins. As far as I can tell, the
only reason this is not possible is because the derived class comes
first in the linearization order. So what I do is manually change the
linearization order:

trait OneTwoThree extends A {
   def x = List(1, 2, 3)
val a = new OneTwoThree with B

And this is not so bad, other than that it splits up my code in a way
that is different than how I feel the code should logically be

So what I'd like to know is,
1) Is my analysis of the problem correct?
2) If so, is there any chance that the linearization order could be change?


Copyright © 2012 École Polytechnique Fédérale de Lausanne (EPFL), Lausanne, Switzerland