Date: 2020-12-28 18:16:07
> -----Original Message-----
> From: Boost <boost-bounces_at_[hidden]> On Behalf Of Soronel Haetir via Boost
> Sent: 27 December 2020 23:50
> To: boost_at_[hidden]
> Cc: Soronel Haetir <soronel.haetir_at_[hidden]>
> Subject: [boost] Potential order statistic tree?
> Boost developers,
> Would there be interest in adding an order statistic tree to the boost library? That is, a tree that
> instead of being sorted simply keeps nodes in whatever order is specified, along with a (potentially
> multi-index) per-node value and a total of per-node values of each node and its left-descent tree.
> I currently have two such types, a very capable intrusive version (not based on boost::intrusive though
> using some of the same ideas) and a less capable non-intrusive form that uses the mechanics of the
> intrusive form as an implementation detail. The reason for not being based on Boost.Intrusive is that
> the tree algorithms that Boost.Intrusive provides are all for search trees and this very definitely is not,
> plus the need to alter node weights during insert, delete and rotations and the Boost.Intrusive
> algorithm providers not having a good place to add that behavior.
> I have personally found this type very useful while implementing a rope container and also while
> implementing tracking for various parts of a text editor (that is, paragraph, line and run positions) (in
> terms of characters, graphemes and screen height).
> Perhaps the major drawback to what I have is that I have not been careful to isolate c++-version
> dependent code. Although I know it does compile with the latest msvc 2019 as well as the clang 10.0
> that MS ships.
You need not be *too* concerned about previous C++ versions. For a new library, you can simply specify the platform and compiler versions (and minimum C++ standard version) required. It is highly desirable that it works with recent MS, Clang and GCC versions (and you have ticked two those boxes).
(You may be able to get it working with older tools by requiring certain features whose availability is defined by macros).
Of course, we would hope that you would do the necessary maintenance for future C++ versions - (perhaps you should worry more about that ð )
(I am entirely unqualified to speak about the usefulness etc of your library).
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk