Re: [boost] GSoC: "A Generic Geometric Library"

From: Huseyin Akcan (huseyinakcan_at_[hidden])
Date: 2007-04-13 12:43:18

Dear Janek,

As you already mentioned, the title of the proposal is misleading, as
our focus here
is not to write a whole geometric library, but a subset of it. I'm not
sure whether the proposal is publicly available somewhere but I would
like to include some part of it about the scope, hoping that would
clear the doubt about the expectations from this project.

" Regarding the specifics of this project, what I propose to do in a
summer is a limited scope Boost library for 2D mesh and polygonal
environments, where I will mainly concentrate on the combinatorial
information and primitive geometric operations.  Thus the proposed
library should more aptly be called Boost.2Dmesh than Boost.Geometry.
Note that there would very likely be comments as to the incompleteness
of such a "geometry" library, since I will not cover application
domains like windows / GUI, computer vision, computer graphics (beyond
the simple 2D subdivision), etc.  By purposely limiting myself to the
scope described above, I would provide a foundation for representing
and manipulating planar subdivisions which is already a huge help for
people writing geometric applications.  It is a very common task and
having both a representation and generic algorithms (which can adapt
to user representation as well) suits the design principle of generic
libraries very well. "

Best,

huseyin

Janek Kozicki <janek_listy <at> wp.pl> writes:

>
> The link from http://code.google.com/soc/boost/about.html states:
>
>  "Programming geometric algorithms, even the most trivial ones, may
>   require detailed knowledge about the problem because of the degenerate
>   cases that frequently occur in practice. Moreover, geometric
>   algorithms are full of case analysis and quite error-prone to program,
>   and can get fairly sophisticated in order to improve the asymptotic
>   performance.
>
>   The main motivation of this proposal is to develop an easy-to-use but
>   also high quality (both in software engineering and in algorithmic
>   content) geometric library, sharing the same design principles as the
>   C++ STL, which is generic and extendible. My main focus will be on
>   design of generic algorithms for navigating and manipulating 2D
>   subdivisions."
>
> I am worried. The lenghty discussions about geometric library spanned
> several years till now on this mailing list. Similarly as with
> Boost.Units library everyone had different expectations.
>
> I strongly encourage the authors to change the name of this library
> (similarly as "math toolkit" is suggested to be renamed into
> "statistics toolkit"), or to change its focus.
>
> "2D subdivisions" is not a thing that I'd be looking for inside a
> "generic geometric library".
>