Subject: Re: [geometry] extensions INCLUDING BREAKING CHANGE
From: Krzysztof Czainski (1czajnik_at_[hidden])
Date: 2012-03-16 18:35:40
2012/3/4 Barend Gehrels <barend_at_[hidden]>
> Hi List, Kris,
> Herewith the report on these suggestions:
> On 7-2-2012 17:51, Krzysztof Czainski wrote:
>> First of all, I think I read in boost guidelines, that it is discouraged
>> that a class and namespace enclosing it have the same name. And in such
>> cases, the convention is for the namespace to be named in plural. So I
>> suggest to rename the namespace boost::geometry::projection to projections.
>> I did not prepare a patch for that ;-)
> Done. As said this is a breaking change, all using programs should update
> "boost::geometry::projection::**" to "boost::geometry::projections:**:"
>> Secondly, I wanted to remove the dependency on boost/lexical_cast.hpp. I
>> noticed that BOOST_GEOMETRY_NO_LEXICAL_CAST is already used in some parts
>> of Boost.Geometry, so I just added the use of this to other places, and
>> no_lexical_cast.patch contains the changes.
> This patch is applied. Note that Bruno suggested to solve this more
> structurally and I agree with that. But this was effectively a one line
> change in one file so to finish with this action point, it is already done.
> Will be continued.
>> Next, I work with a Texas Instruments compiler, which has this defect,
>> that static local variables of inline functions are not resolved to a
>> single instance, and it warns about it. So I made two minor changes:
>> static_inline.patch. I think they wouldn't hurt in any case, but I could be
>> wrong here.
> This patch is applied. Note that immediately after this I changed PI to
> boost::math::constants::pi<**double>() and similar, I think TI compiler
> accepts that.
>> And lastly, I encountered a macro problem - another defect of my compiler
>> - it has a #define cosl cos, so there are problems with local variables
>> named cosl. cosl.patch.
> This patch is applied and the conversion program is updated. Besides that
> I did run it on the latest proj trunk (no big differences though) so it is
> up-to-date again. The new proj (I think in preparation for 4.8) defines 5
> additional projections which are not yet supported.
> Kris, thanks again for these patches and suggestions. Will now continue
> with your second batch ;-)
Thanks Barend for applying the patches and adding new_projection.hpp ;-)
I Just wanted to report, that I updated geometry to trunk r77308, and
successfully ran my tests on mingw and on the problematic TI compiler. Like
before, I used boost-1.48.0, and geometry from trunk on top of it (on the
compiler's include list added geometry before boost-1.48). I also needed to
use math/constants from trunk.
One minor issue: geometry/extensions/gis/projections/pj_fwd.hpp seems to be
missing an include: projects.hpp.
Geometry list run by mateusz at loskot.net