Re: [boost] Boost.Local Review

Subject: Re: [boost] Boost.Local Review
From: Lorenzo Caminiti (lorcaminiti_at_[hidden])
Date: 2011-11-17 12:22:44

On Thu, Nov 17, 2011 at 12:10 PM, Vicente J. Botet Escriba
<vicente.botet_at_[hidden]> wrote:
> Le 17/11/11 13:15, Lorenzo Caminiti a écrit :
>>
>> On Thu, Nov 17, 2011 at 4:33 AM, Vicente Botet<vicente.botet_at_[hidden]>
>>  wrote:
>>>
>>> lcaminiti wrote:
>>> I have not used yet C++ lambdas, couldn't the following be used to
>>> support
>>> const binding?
>>>
>>> xtype x;
>>> xtype const&  xcr; // const binder
>>> [xcr]() {
>>>    assert( xcr == 0 );
>>> }();
>>
>> I can't check with a C++11 compiler right now, but my understand is
>> "yes". Using decltype, you don't even have to repeat the type (to make
>> maintenance easier). I think the most general way to write that is:
>>
>> xtype x;
>> decltype(x) const&  const_ref_x = x;
>> [xcr]() {
>>     assert( const_ref_x == 0 );
>> }();
>>
>> The down sides are:
>> 1) The lambda cannot refer to the bound variable using its original
>> name x (this will actually look very strange in Boost.Contract if I
>> had to indicate the user to use some magic names for variables in
>> constant assertions).
>> 2) The extra type manipulation.
>>
> Could you add the preceding lambda alternative to the Alternative sections
> once you could verify it?

Yes.

>>
>> I would instead prefer the following if C++11 allowed it:
>>
>> xtype x;
>> [const&x]() {
>>     assert( x == 0 );
>> }();
>>
>> But C++11 only allows to qualify the bind with the reference&x and
>> not with the additional const const&x.
>
> Maybe it is worth to post the idea to the comp.lang.c++ ML.

Other people have raised the same question (I can't find another
reference I read a while back):
http://stackoverflow.com/questions/3772867/lambda-capture-as-const-reference

I guess I could ask comp.lang.c++ why C++11 does not allow const&
bindings (I'm not really sure myself).

Thanks.
--Lorenzo