Re: boost::format ctors : why copy arguments by value ?

From: krempp_at_[hidden]
Date: 2001-06-26 18:40:57

--- In boost_at_y..., "David Abrahams" <david.abrahams_at_r...> wrote:
> I think this has been pointed out before. Unfortunately, 
development on

that's right, I had not searched enough on the list. 
I found the thread launched by a discussion on utf-8..
It raised several points, and the thread did not reach clear 
settlements on some of those points.
msg 4634 presented a summary from karl nelson.

It seems to me most points raised showed after discussion that 
boost::format had chosen directions at several steps that are no 
longer desired, eg it chose not to conform to printf's syntax, not 
to be used as manipulator, ..

And in fact, after these discussions the 'format' class wanted for 
boost seems to neither be the current boost::format in the vault nor 
karl nelson's (http://www.ece.ucdavis.edu/~kenelson/ofrstream.cc), 
but more like some clever mix of them both..

> boost::format has lapsed. Would you be interested in picking it up 
and
> polishing it for formal review? It would be an incredibly useful 
library to
> have in the official boost suite.

Well, it would surely be a very profitable task for me, but I am not 
sure I have 
enough knowledge and experience of C++ for this.

I took some time to read most of the previous msgs on the subject, 
and some points are completely unknown to me (I have absolutely no 
experience of wide char types, very little knowledge of stream 
manipulators, .. and am not either a familiar of printf syntax - 
which could possibly be to my honor in a C++ context ;-) - )

I believe defining 2^10 variations of 
 template<typename A, typename B>
    basic_format(const charT* ca, const A& a, const B& b)
(with varying number of arguments and presence/absence of const)

is _not_ what you have in mind to make the current boost::format 
ready for formal review...

Do you suggest I should try my luck at redesigning a 
template<class charT> class format, with the enhancements that have 
been implemented in nelson's format,
and upload it when reasonnably ready for comments-based evolution ?

-- 
Sam