Re: [boost] [function_type] Tag types documentation

From: Tobias Schwinger (tschwinger_at_[hidden])
Date: 2005-07-04 17:12:23

Dave Gomboc wrote:
>>The Function Type library uses tag types to represent a type's
>>attributation, such as its calling convention or wether it is
>>variadic.
>>
>>Tags that represent the values of a single property are called
>>property tags.  These tags can be used to determine whether one
>>property of a type has a particular value.
>>
>>    is_function< T, variadic >
>>    is_function< T, default_call >
>>
>>Does it work?
> 
> 
> It's better than before.  I still think you shouldn't mix "uses tag
> types" (plural) with "a type's attributation" (singular), but you're
> the author. ;-)  I'd still go with "to represent attributations of
> types", similar to how I wrote that line previously.  Oh, also, fix
> the spelling of "whether".
> 

I'm not a native English speaker and I know English has tighter restrictions than 
my mother language at this point. So there's a good chance my sentence is indeed 
ill-formed.

   The Function Type library uses tag types to represent one or more properties
   of a type. Each property describes an optional attribuation.

The object of the sentence is plural now.

> 
> 
>>I have. Try these use cases:
>>
>>      // a remove constness (of the pointee) from a member 
>>function pointer type T
>>      function_type<signature<T>,tag<signature<T>,non_const>
>>::type
>>
>>      // shortcut:
>>      function_type<T, non_const >::type
>> 
>>or
>>
>>      // create a (possibly decorated) function type's 
>>variadic version
>>      function_type<signature<T>,tag<signature<T>,variadic>
>>::type
>>
>>      // shortcut:
>>      function_type<T, variadic >::type
> 
> 
> Is manipulating these using MPL containers too much hassle for too
> little benefit?

In case you are referring to the shortcut: It's just syntactic sugar.

In case you are referring to the tags: It would be overkill - tags hold 
expressions of bitwise integral arithmetic.

In case you are referring to the function's component types:
These /are/ specified by MPL sequences. If they are specified by a type that is 
not a sequence, a shortcut is triggered to create a sequence (signature<T>) from 
that type.

Did any answer fit your question? I'ld be curious to know which one, then.

> 
> If <T, nonvariadic, variadic, variadic, nonvariadic, variadic> is
> really going to be the same as <T, variadic, nonvariadic, variadic>,
> then even testing for equivalence between tags becomes somewhat of a
> chore.
> 

You shouldn't test tags for equivalence and the design keeps you from doing it.

<snip>

I cut this part of the discussion because it's based on a misunderstanding and far 
away from bringing us further. I'll try to clarify my point by ansering a tiny 
part of it:

> "If I assign more than one value at the same time, I really mean the
> last one", which is different behaviour than runtime-C++ gives.

Why are you looking at a specialization of 'tag' being an assignment? It's a 
function! We can define functions like 'tag' in runtime C++ as well:

    tag_t tag(tag_t tag1, tag_t tag2)
    {
       // for each property in tag2:
       //   assign that property value of tag2 to tag1
       // return changed tag1
    }

Are you trying to say that a good interface must not contain such a function?

> That's why this asymmetric disambiguation behaviour feels awkward to
> me.
>

Please note that the asymmetric behaviour is _essential_ and not for disambiguation:

    tag<signature<T>,a_property_tag>

describes a transformation of T's properties. 'signature<T>' already contains all 
properties, the property value described by 'a_property_tag' is overridden!

As a conclusion I will use this ^^ example in place of the obviously confusing 
"tag<variadic,non_variadic>  nonsense".

I apologize for adding to this confusion by my "loud thinking" in the previous post!

Thanks,

Tobias