Re: [boost] [sort] Timsort review reminder

Subject: Re: [boost] [sort] Timsort review reminder
From: Paul A. Bristow (pbristow_at_[hidden])
Date: 2017-06-07 16:03:02

> -----Original Message-----
> From: Boost [mailto:boost-bounces_at_[hidden]] On Behalf Of Steven Ross via Boost
> Sent: 07 June 2017 11:54
> To: boost_at_[hidden]
> Cc: Steven Ross
> Subject: Re: [boost] [sort] Timsort review reminder
> 
> On Wed, Jun 7, 2017 at 6:33 AM Niall Douglas via Boost <
> boost_at_[hidden]> wrote:
> 
> > On 06/06/2017 16:40, Zach Laine via Boost wrote:
> > > Please provide a link to online documentation, if if one exists.  All
> > I've
> > > been able to find so far is a paper and Doxygen reference docs.  I got
> > that
> > > much by cloning the GitHub repo (putting docs online somewhere will get
> > you
> > > a lot more reviewers :).  Did I miss it somewhere?
> >
> > I see the pull request implements no documentation page matching
> >
> > http://www.boost.org/doc/libs/1_64_0/libs/sort/doc/html/sort/sort_hpp/integer_sort.html
> > nor modifying it. Some doxygen comment documentation can be found at:
> >
> >
> > https://github.com/boostorg/sort/pull/12/files#diff-c13fbe11aeffb2befb942f91c401e8ceR85
> 
> 
> Here is some documentation on Timsort:
> 
>    1. Wiki <https://en.wikipedia.org/wiki/Timsort>
>    2. brief description in python dev list
>    <http://svn.python.org/projects/python/trunk/Objects/listsort.txt>
>    3. Original implementation <https://github.com/gfx/cpp-TimSort>
> 
> Francisco and I are in the middle of rewriting the docs for the Sort
> library.  Integrating Timsort in there once done will be relatively simple.
> 
> > I'll be honest here: I personally would have felt this pull request
> > better reviewed and handled by Boost.Sort's maintainer. It's too small
> > and limited for a full fat review by the entire community unless there
> > is something very controversial about it and the maintainer feels the
> > community needs to invest a week into thinking about this.
> >
> > Niall
> 
> I would prefer a mini-review myself (as I agreed to do when adding new
> algorithms to the collection) but Ronald suggested I do a full review.  My
> main questions are: Does anyone care about Timsort?

I agree that TimSort is well worth having in Boost. So a YES to accept.

But provided we can have some documentation on when it is likely to perform well - covered well by the Wikipedia and original
articles (including implementation notes on details like the 'bug' discovered).

It should be clear about acknowledging other people's work too.

Paul

---
Paul A. Bristow
Prizet Farmhouse
Kendal UK LA8 8AB
[phone]