Re: [9front] New Balanced Search Tree Implementation

[Benjamin Purcell <benjapurcell@gmail.com>]

> Then why even implement it then when its always slower than avl?
> you could get rid of the indirection...

1. In the interest of keeping 9front in the tradition of not only an
OS we use but also an OS for research and learning, I think it's good
in its own right to have an implementation of another classic balanced
tree with interesting invariant properties.
2. RB trees are often touted as having better behavior than AVL trees
and especially in recent years there has been controversy over the
left leaning variant. Having comparable implementations of the two
trees together in one OS gives a living example of why AVL and only
AVL is the one true BST. This is In the same way that keeping acme and
sam around clearly shows that ED is the one true editor.

> another point about using enum constants to set the algorithm mode
> has the disadvantage that you always have to link in all the implementations
> even if you are not using them...
>
> a better way would be to provide separate functions, that way if you
> only use bstcreateavl(), you'll only link in the avl code... but you
> can still have your common interface structure.

That is a great point and I'll make that change soon.

> --
> cinap