Re: Ongoing review of printf-like formatting

From: s_krempp (krempp_at_[hidden])
Date: 2002-01-21 09:04:55

> - The choice of formats is excessive,  simple positional  and Open Group
> printf is sufficient to cover most needs.  One can always revert to streams
> when they are not.
> If the %p....  formats really are needed then ( the documentation of ?) the
> management of the choices needs significant improvement.

I hope the fault here is on the documentation..

Well, I've written a new paragraph about the syntax design in
http://groups.yahoo.com/group/boost/files/format2/choices.html
please have a look and tell me at what stage of the decision process you disagree with it.

> - the code from format.zip 2002/1/20  does not compile with MSVC6SP5.

rogeef sent detailed instructions on how I should modify the library to make it compatible with MSVC, and I will follow those 
with great care.

> - there are other issues which need fixing but would not prevent acceptance
>     e.g. the documentation at 2002/1/20 has a mixture of []   and  % for
> parameters

I uploaded some 'fixed' version with too much precipitation.
I have corrected these incoherences in my local files, and will upload
everything shortly (once I implement the MSVC compatibility fixes)

> Also why  is str() a member function and a free function ?

Many people ask this.
I wanted it to be a member function (that's what it should be),
but it's clumbersome to use in many cases :
( format( fstring) % x1 % x2 ).str()

The free function approach is better here :
str( format( fstring) % x1 % x2 )

I can move the .str() code into the free function body,
and simply use a friend declaration inside the format class.

I did not know it was bad to provide a free-function that just
calls the corresponding member function on the argument.

Thanks for the comments,
-- 
Samuel