Re: syntax suggestion

Alexander Nasonov <[email protected]> writes:

> Daniel Wallin wrote:
>
>> Alexander Nasonov wrote:
>>> I'd like to see inheritance more natural in langbinding. What do you
>>> think if inheritance relationships could be exposed like this:
>>> 
>>> class_<C>("C")
>>>     , public_<A>
>>>     , public_<B>
>>> [
>>>     // ...
>>> ];
>> 
>> The [] syntax still won't work (IMO) for declaring classes. So you'd end
>> up with something like:

Not to mention that the []s will bind more closely to public_<B> than
to the class_ object.  If we end up building an AST for the
class declaration and then walking over it to evaluate it, I fear
we're going to end up overstressing many compilers.

>>      class_<C>("C")
>>          .public_<A>()
>>          .public_<B>()
>>        .def(..)
>>        .def(..)
>>        ;
>
> I think it's possible to implement []-style but I'm not sure how much time
> it would take to compile.

Agreed, it's possible.  The compilation time is IMO a serious concern.

>> Although it looks kind of cool, 'public_' is completely redundant since
>> we are only interested in accessible bases anyway.
>
> Not sure about this. We may need access to protected C++ bases from non-C++
> class (I don't know how to implement it elegantly, though).

Actually I think the issue here is that we don't need to represent
access.  Why would the system care whether a base was private or
protected or public?

>> When in doubt I tend to go for the less verbose and less complicated
>> alternative, so I guess I like the original version is better:
>> 
>>      class_<C, bases<A, B> >("C")
>>        .def(..)
>>        .def(..)
>>        ;
>> 
>> Here are some other possibilities:
>> 
>>      // Cool, but not very portable
>>      class_<C(A, B)>("C")

I like the syntax; I guess we need to decide whether we care about
older compilers or not.

>>        .def(..)
>>        .def(..)
>>        ;
>> 
>>      // Just add more parameters to class_<>, complicates
>>      // the metaprogramming

Definitely problematic, especially if you want to be able to pass
other parameters.

>>      class_<C, A, B)>("C")
>>        .def(..)
>>        .def(..)
>>        ;
>> 
>> I'm not sure if either one has any significant benefits from the
>> original though.
>> 
>> I think we need more opinions, anyone?

The only advantage I can see to using the []s is that it becomes a
bit cleaner to define nested classes.

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com



-------------------------------------------------------
This SF.Net email is sponsored by: IBM Linux Tutorials
Free Linux tutorial presented by Daniel Robbins, President and CEO of
GenToo technologies. Learn everything from fundamentals to system
administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click