traits: ability to define what optional means

[michael-brade]

traits: ability to define what optional means

[michael-brade]

Yes, I agree and the logic is very good. Thanks a lot!

[ColinH]

Am I right to assume that working with std::optional< bool > is not a viable work-around?

[michael-brade]

yes because that is an API change and it wouldn't be true. The object always has the bool (or int or string for that matter), it's just about how to serialize this object to json.

[ColinH]

After checking the source I believe that is_nothing should never have been applied to non-optional members in the first place.

[ColinH]

Now I'm wondering whether we still need for_nothing_value.

[michael-brade]

Great, I agree with that so far.

Yes, I like for_nothing_value because then I can at least somewhat individually decide if I want the 0/false/empty values in a json representation or not.

[ColinH]

After the most recent change binding::object encodes a member when the object uses for_nothing_value::encode and/or the member uses member_kind::required and/or the member's is_nothing_value is false.

Consequently for_nothing_value does not influence the behaviour of required members. That sounds reasonable and I'm prone to believe that it can stay like this. @michael-brade Do you agree, does this work for your use cases?

template< typename A, template< typename... > class Traits, typename C >
[[nodiscard]] static bool require_encode( const C& x )
{
   return ( N == for_nothing_value::encode ) || ( A::kind == member_kind::required ) || ( !A::template is_nothing< Traits >( x ) );
}