Re: Is C++ Variant Considered Stable?

[Akim Demaille]

> Le 28 juin 2017 à 23:14, Simon Richter <address@hidden> a écrit :
> 
> Hi,
> 
> On 28.06.2017 21:33, David McCallum wrote:
> 
>> Are there changes pending for variant that might break my parser somewhere
>> down the road? Or is it considered safe to use?
> 
> Presumably, neither, as I haven't seen any feedback on its usefulness.
> 
> One problem is that you cannot use std::unique_ptr<T>, as the types need
> to be copyable, not just movable. This is something that I could see a
> possible breaking change for – but I'm not sure who would implement such
> a change.

To clarify: AFAICT, they work well – I use them intensively in my project.
They do have a ‘limitation’: they don’t support things like $<foo>1, but
you should not do that, except for midrule actions, for which Bison 3.1
will provide a solution.

Move semantics is definitely a priority, for Bison 3.2.  It is reasonable
to expect this in October this year.