Re: [boost] Announcing CallableTraits' formal review next week (April 3rd)

Subject: Re: [boost] Announcing CallableTraits' formal review next week (April 3rd)
From: Barrett Adair (barrettellisadair_at_[hidden])
Date: 2017-04-09 05:00:16

On Sat, Apr 8, 2017 at 1:09 PM, Zach Laine via Boost
<boost_at_[hidden]> wrote:
> On Mon, Mar 27, 2017 at 11:38 PM, Louis Dionne via Boost <
> boost_at_[hidden]> wrote:
>
>> Dear Boost community,
>>
>> [snip]
>>
>> We encourage your participation in this review. At a minimum, kindly state:
>> - Whether you believe the library should be accepted into Boost
>>
>
> Yes.
>
>
>> - Your name
>>
>
> Zach Laine
>
>
>> - Your knowledge of the problem domain
>>
>
> I am reasonably knowledgable.  I do extensive metaprogramming from time to
> time, though this is not my specialty.
>
>
>> You are strongly encouraged to also provide additional information:
>> - What is your evaluation of the library's:
>>   * Design
>>
>
> This seems good to me.  There's one thing I consider to be a glaring
> omission, though.  I would love to have pre-17 versions of these traits:
>
> template <class Fn, class... ArgTypes>
> struct is_invocable;
> template <class R, class Fn, class... ArgTypes>
> struct is_invocable_r;
> template <class Fn, class... ArgTypes>
> struct is_nothrow_invocable;
> template <class R, class Fn, class... ArgTypes>
> struct is_nothrow_invocable_r;
>

I had implemented something similar, even to the point of passing
libc++'s std::invoke
test suite, but I removed it after learning about these C++17
additions. You aren't
the only one to request this feature. I've opened an issue to add these.

On a related note, at one point I also implemented is_constexpr_invokable
as an experiment. Do you think this would also have merit for inclusion? The
merit of this trait is questionable to me, and the linked implementation does
things it shouldn't do (coercive template instantiation), but here it is anyway:

https://github.com/badair/constexpr_checks/blob/master/constexpr_checks.hpp#L630

>>    * Implementation
>>
>
> I did not look at the implementation.
>
>
>>   * Documentation
>>
>
> From the first page of the docs, this:
>
> "
> The use cases for CallableTraits overlaop significantly with those of
> Boost.FunctionTypes
> <http://www.boost.org/doc/libs/1_61_0/libs/function_types/doc/html/index.html>.
> Here are some reasons why you might prefer CallableTraits over the latter:
> "
>
> is awkward enough that I had to re-read it.  It mentions two things, but in
> the order A B A, then uses "latter", which is off-putting.  Can I suggest
> instead:
>
> "
> The use cases for CallableTraits overlap significantly with those of
> Boost.FunctionTypes
> <http://www.boost.org/doc/libs/1_61_0/libs/function_types/doc/html/index.html>.
> Here are some reasons why you might prefer CallableTraits:
> "
>
> Also, "overlaop" is a typo. :)
>

Thanks, this is being addressed.

> I also find it odd that the FAQ immediately follows the intro page.  I
> think this would be better placed before the Contact section.  Keeping it
> where it is now disturbs the flow of the reader's thinking when getting to
> know the library for the first time (or at least it did so for me).
>

Good point, I'll move this.

> Nearly every section of the reference docs says this:
>
> "
> Compatibility Notes
>
> Full support on GCC 4.7.4+, Clang 3.5+, Visual Studio 2015, and XCode 6.4+.
> "
>
> This is not a problem per se, but it's noise considering that the same
> statement is made in the intro section.  I think the reference entry should
> only contain such a notice *only* when it differs from the general note
> from the intro.  This has the benefit of being clearer -- I won't see the
> note above and wonder if it differs slightly from the general intro note.
>

Agreed, will fix.

>   * Tests
>>
>
> They seem to provide decent coverage, and line up with the documented
> semantics.  The docs in fact seem to use quoted test code to illustrate
> behavior, which is nice.
>
>
>>   * Usefulness
>>
>
> Very useful.  As someone who is not a metaprogramming specialist, I often
> need facilities like this when doing MP, and I often define my versions
> with subtle defects.
>
>
>> - Did you attempt to use the library?
>>
>
> Not really.  I only ran the tests on Clang 3.9.
>
> - How much effort did you put into your evaluation of the review?
>>
>
> About 5 hours.
>
> Zach
>

Zach, thank you for spending so much time on this review.

> _______________________________________________
> Unsubscribe & other changes: http://lists.boost.org/mailman/listinfo.cgi/boost