Re: Please implement GLR + C++ + variant or at least warn user that it is unavailable

[Akim Demaille]

> Le 27 févr. 2015 à 18:01, Askar Safin <address@hidden> a écrit :
> 
> Hi.

Hi Askar,

> Please add variant support to GLR C++ parser. Reason is following:
> * Having C++ parser is cool. It allows us to have non-trivial C++ objects as 
> semantic values

Yes, I agree.

> * But then (if we use non-trivial objects as semantic values) we need variant

Yes, I agree.

> * And (as docs say) nearly any modern programming language needs GLR parser 
> and not LARL1

Well, I kinda disagree about this, but I do agree GLR is
a must-have.

> If you don't plan to add this support at near future, then, please, put big 
> fat warning to docs that GLR + C++ + variant is not supported at the 
> beginning of C++ section. If I knew this earlier I probably would not convert 
> my C parser to C++ parser. Please, CC me if you want to answer

I'm sorry you were painted in the corner :(  Valentin has been
working on a proper implementation of GLR for C++ with variants,
but it's unfinished work.  I expect he'll be back on this during
this year.

Meanwhile, you _might_ want to try this modified version
of version of Bison that does provide this (GLR + Variants + C++),
but it's unsafe, badly written.  It happens to work in some cases...
*Do not* use it for production code.

https://www.lrde.epita.fr/~akim/download/bison-3.0.4.19-fbaf.tar.xz