Re: GLR C++ Variants

[Akim Demaille]

Hi,

> Le 11 janv. 2020 à 10:38, Ookami <address@hidden> a écrit :
> 
> Hello! Five years ago was a feature request about C++ variants for GLR. The 
> Bison documentation says that C++ Variants are not supported in GLR, and 
> C++17 variants will not be supported to maintain backward compatibility with 
> older versions of C++ standards.
> 
> I know, that Valentin implementing C++ GLR variants. Is there any progress?

Yes, there are.  However Valentin had to stop for a while, and
he will certainly come back to this in the near future.  I don't
know exactly his plans though.

> I wish to implement at least C++17 variants, and then move forward for 
> backward compatibility.

The question is not C++17 variants vs. Bison variants, that's a detail.
The problem is a complete implementation of GLR in C++, instead of the
thin C++ wrapper around a C GLR parser as we have it today.  That's
way more ambitious.

> Or I thought about another feature. Something like attributes, that can be 
> attached to rules (for example, I can attach a needed type, like it already 
> implemented by variants in Bison — %type <my_type> rule1 ...), and then in 
> C++ code handling the rule I would like to read this attribute, to access the 
> value from C++17 std::any or std::variant. For example:
> 
> %attr rule1 {type}
> ...
> rule2: rule1 {$$ = std::get<$a1>($1);}
> 
> And yes, I want help with implementing C++ GLR variants.

Sorry, I'm confused, and I don't understand what you mean.  Maybe
you meant to say "nonterminal symbol" instead of "rule"?