Re: [Boost-users] Boost-users Digest, Vol 3442, Issue 1

Subject: Re: [Boost-users] Boost-users Digest, Vol 3442, Issue 1
From: Dan Searles (dansearles_at_[hidden])
Date: 2013-05-11 11:52:20

>>> Hello,
>>>
>>> the code here doesn't compile:
>>>
>>> #include <boost/mpl/string.hpp>
>>> #include <iostream>
>>>
>>> namespace mpl = boost::mpl;
>>>
>>> typedef mpl::string<'h','e','l','l','o',' ','w','o','r','l','d'>::type 
>>> my_hello;
>>>
>>> int main()
>>> {
>>>     std::cout << mpl::c_str<my_hello>::value << '\n';
>>>     std::cout << BOOST_MPL_LIMIT_STRING_SIZE << '\n';
>>> }
>>>
>>> As usual, the example from docs works. The error is:
><snip>
>> 
>> I'm surprised multi-character character constants are being used at all in Boost, they
>> are not supposed to be portable:
>> 
>> http://stackoverflow.com/questions/6944730/multiple-characters-in-a-character-constant
>
>Boost has never shied away from non-portable constructs. However, it
>*does* wrap non-portable constructs in config macros to ensure a
>consistent cross-platform experience, which is what mpl::string does
>with multichar literals.

Does the wrapping include handling 'qwer' on platforms where int is not 4 bytes?
It may, I don't have a way to test it.

>
>> If you only want to use single character constants, you could place the following #define before any includes:
>> 
> >#define BOOST_MPL_LIMIT_STRING_SIZE (32*4)
>> 
>> That would allow up to 32, provided your compiler supports it (yet another portability issue).>>
>
>That would increase the compile time and decrease readability for no
>benefit, IMO.

Opinions may differ on the readability of:

    typedef mpl::string<'hell','o wo','rld'> ......

vs:

    typedef mpl::string<'h','e','l','l','o',' ','w','o','r','l','d'>.....

But increasing compile time would be bad. That's why I'm asking if it would be
possible for there to be an mpl::string<> that takes BOOST_MPL_LIMIT_STRING_SIZE 'char's?
[It looks like 'x' started being a char in C++11, so it may be complicated supporting prior versions]

>
>> Since in C++, the type of 'x' is char, it should be possible to create another overload of
>> mpl::string<> that takes BOOST_MPL_LIMIT_STRING_SIZE 'char's, which would work
>> the way most people would expect..
>
>How do you think most people expect it to work?

My opinion is that most programmers would expect BOOST_MPL_LIMIT_STRING_SIZE
single character constants to work. I know that multi-character constants are not always
portable, so I try not to use them, even if a particular instance may be portable.

>
>-- 
>Eric Niebler
>Boost.org
>http://www.boost.org

Dan Searles