Re: [boost] [proposal] Sequence container with O(log n) random access _and_ O(log n) insertion/removal (wherever)

From: Martin Knoblauch Revuelta (mkrevuelta_at_[hidden])
Date: 2006-11-02 08:28:04

On 11/1/06, Jason Hise <0xchaos_at_[hidden]> wrote:
> [..]
> Right now I am seeing a dangerous degree of feature bloat resulting
> directly from your desire to build on the optimizations possible
> because of the tree nature of the implementation.

Obviously, I can't impose any undesired extra feature. It was never my
intention. Sorry if it looked like that. I was just trying to convince
you.

Of course, I will be pleased to prepare a version with only the
features approved by the general opinion.

I have aleady made a lot of changes following your advise. The mess of
pointer conversions has been simplified, and I can change it more if
nobody likes how it is right now. The "static where possible" approach
will soon disappear unless somebody expresses a desire to keep it.
I've realized that the stable_sort can be done by the user (using
sort) with some effort, but with the same algorithmic complexity, so
I'll probably remove m_oldpos and stable_sort(). I'll put the
algorithms out of the class and I'll organize it all in modules, so
that everybody can see how it woult look without algorithms and/or
without NPSV.

Thanks again,

Martin