Re: [boost] Ongoing review of printf-like formatting

From: Samuel Krempp (krempp_at_[hidden])
Date: 2002-01-21 05:32:17

On Mon, 2002-01-21 at 04:47, Jeremy Siek wrote:
> index.htm
>   Should link to the other example files and the test files. It would be
>     great to have short summary of each file.

good idea !

>   "glue" is a bad name for this function. How about something like
>      "modifier" or even "format" instead (or some other better name).

glue really was meant to be renamed, and today I chose to rename it to
'manip'. it's shorter than modifier, and just as much explicit.

I'm open to popositions..

>   Under "Differences of behaviour vs printf"
[ .. ]

ah yes, the HTML documentation is far from perfection..

> Portability issue: Boost.Format uses sstream (which bad for older g++'s)

I hope requiring g++ >= 3.0  is not a show stopper..

The MSVC portability requires that I move some definitions, once done it
seems that format can be compiled with
at least
MSVC
g++ >=3.0
comeau

>   trying to compile sample_formats.cc with KCC, got the following
>     warnings and errors. I've only listed each kind of error once.
>     They also occur on other lines. I would pay attention to these
>     errors. The EDG frontend (used by KCC) it highly standards
>     conformant, so these errors are likely problems in the code,
>     not in the compiler.
> 
>   "sample_formats.cc", line 28: warning: temporary used for initial value of
> 	    reference to non-const (anachronism)
> 	oss << format("writing %3,  x=%1 : %2-th step \n") [40.23] [50] ["toto"];

ah.. does someone understands this ?
I'm guessing one of the arguments is passed via a temporary (my guess is
"toto").
The compiler can choose between 
operator[](const T&)
and
operator[](T&)

So he should use the const one for a temporary, shouldn't he ?

>   "../../../boost/format.hpp", line 209: error: type
>           "boost::format::basic_format<charT, Traits>::string_t [with
>           charT=char, Traits=std::char_traits<char>]" is inaccessible
>       typedef typename basic_format<charT, Tr>::string_t   string_t;

Yes, the same error was found with MSVC. I fixed this.
(g++ did not see it, and neither did I..)

>                                                 ^
>   "../../../boost/format/format.ihh", line 448: error: more than one instance of
>           overloaded function "std::basic_string<charT, traits,
>           Allocator>::insert [with charT=char, traits=std::char_traits<char>,
>           Allocator=std::allocator<char>]" matches the argument list:
>             function "std::basic_string<charT, traits,
>                       Allocator>::insert(std::basic_string<charT, traits,
>                       Allocator>::size_type, std::basic_string<charT, traits,
>                       Allocator>::size_type, charT) [with charT=char,
>                       traits=std::char_traits<char>,
>                       Allocator=std::allocator<char>]"
>             function "std::basic_string<charT, traits,
>                       Allocator>::insert(std::basic_string<charT, traits,
>                       Allocator>::iterator, std::basic_string<charT, traits,
>                       Allocator>::size_type, charT) [with charT=char,
>                       traits=std::char_traits<char>,
>                       Allocator=std::allocator<char>]"
>             argument types are: (int, int, char)
>           res.insert(0, 1, ' '); // insert 1 space at  pos 0
>               ^

ah.. it means '0' needs as much conversion to become a 'size_type' as it
needs for 'iterator'. 
It's strange.. I'll explicit the type for 0, and it won't be ambiguous
anymore. 

>   In the examples, no need for the asserts. Examples are not tests.

true. 
 

For your other comments, some are due to the fact I upload new versions
to the vault when implementing suggestions (eg operator[], etc..)
the newest modifications are not refelected in the documentation..

I've sent a note, 
http://groups.yahoo.com/group/boost/message/23411,
to explain what were the different files in the vault.

-- 
Sam