Boost logo

Boost :

Subject: Re: [boost] [countertree] Formal Review Request
From: Luke Simonson (simoluk_at_[hidden])
Date: 2012-10-04 17:39:23

> The main utility of the vector_tree is when you must insert or delete many
> elements in central positions. It is not worst or best than others data
> structures, it is different. It is only an option more, with different
> features than other data structures. If you consider it useful, use .Only
> that

That's the problem, it shouldn't be up to us to figure out if your
library is useful. We've all been happily coding in C++ for 10+ years
without it. We know how to solve every programming problem we are
presented with without the use of your library. Sometimes it is
obvious, but in this case it isn't obvious to me where I would want to
use your library. You need to provide some kind of motivation for why
you think the library is needed. Something like "when you are doing
xyz and you run into problem A you end up settling for solution S, but
my library is a better solution because of t, u, v."

Your iterators are not random access iterators because O(1) iterator
arithmetic is a requirement of the standard. Calling std::sort or
std::binary_search on a tree seems completely wrong. The tree is a
search data structure, why would we want to call an algorithm that
doesn't know how to take advantage of the tree data structure? You
say you wanted to be able to call these stl algorithms on your
iterator types, but you didn't explain why. I can't think of a single
programming activity xyz where I run into problem A where your library
is a better solution than solution S that I would normally apply, but
I shouldn't have to. You should know why you wrote your library and
be able to tell us.

> About the C++11 stateful allocators, I need a bite of time for to study in
> deep , and provide you something more than a trivial opinion Sorry!

It is pretty important to figure out how what you're doing relates to
the new standard. It takes a while for boost libraries to be accepted
and released. Use of the new standard will be even more prevalent by
the time your library would be getting widespread exposure. Keeping a
small object arena around to recycle elements instead of going always
to the allocator is a good idea. I'm pretty sure we can do it with
the C++11 stl by baking it into the stateful allocator. It would be
much more useful as a general allocator than as a feature of yet
another implementation of a few containers.


Boost list run by bdawes at, gregod at, cpdaniel at, john at