Re: [Swig-user] GSOC 2012 query

[William S F. <ws...@fu...>]

On 19/03/12 11:41, Vadim Zeitlin wrote:
> On Mon, 19 Mar 2012 10:15:30 +0530 Sajan Kedia<saj...@gm...>  wrote:
>
> SK>  I am looking at new C++11 standards which can be added in SWIG. which is
> SK>  given in the project idea page idea 1.
>
>   Hello,
>
>   Great to hear this, improving SWIG support for C++11 is definitely very
> important and will become only more so in the future.
>
>   Before discussing the details, notice that from practical point of view,
> the first important thing would be to merge gsoc2009-matevz branch with the
> trunk to at least make the already implemented C++11 support available to
> SWIG users. I don't know if it's going to be part of the GSoC or if William
> (or me?) can do it beforehand but you might need to do this yourself.
>
My plan is to merge this branch into trunk after swig-2.0.5 is released 
and any future C++11 code can branch off gsoc2009-matevz or trunk if the 
merge is done in time.

>   I also think that someone (again, possibly you) needs to examine how well
> the features officially supported by the code in this branch work in
> practice. E.g. looking at the first two of them:
>
> - Move constructors: rvalue reference support is there but I don't see
>    anything about move constructors so maybe they're supported and just
>    a unit test for them is missing but maybe they're not and then it would
>    be quite important to add support for them.
>
I've started adding rvalue reference support to gsoc2009-matevz and I 
don't see this as a gsoc project.

> - Const expressions: again, I see that "constexpr" keyword is supported,
>    but it's less clear if its semantics is implemented, e.g. if SWIG can
>    wrap an array of size returned by a constexpr function correctly.
>
> There are probably more such questions about other stuff...
>
>
> SK>  I want to include multithreading support,
>
>   Not sure what do you mean exactly by this, i.e. how is it going to affect
> SWIG.
>
> SK>  generic programming support,
>
>   This is a bit unclear as well, what exactly do you propose to do about
> this?
>
> SK>  uniform initialization
>
>   According to
>
> 	http://swig.svn.sourceforge.net/viewvc/swig/branches/gsoc2009-matevz/Doc/Manual/Cpp0x.html#Cpp0x_Uniform_initialization
>
> this seems to be already supported, i.e. it's parsed correctly (but doesn't
> affect the target language as it's not clear how it could).
>
> SK>  and performance enhancements. For time performance enhancement, extern
> SK>  template can be used.
>
>   Sorry, I don't understand which enhancements do you have in mind here
> neither, could you please be more specific?
>
>   As for "extern template", it would be nice if it could free us from the
> need to write "%template", this would certainly be a welcome improvement.
>
Support for the new keywords would be useful. But I don't think there is 
much to be done there. There is plenty of work in the STL containers though.

I don't know about "extern template". It wouldn't provide a nice 
language name for the template instantiation. However, solving the need 
for a %template is a problem to be solved by itself, but I don't think 
that is doable in SOC.

William