Re: [Libreoffice] message construction bloat ...

On Wed, Nov 23, 2011 at 12:59:00PM +0000, Michael Meeks wrote:

>       The attached test contrasts:

>       fprintf (stderr, "P is %s and b is %d\n", p, b);
> vs.
>       std::cout << "P is " << p << " and b is " << b << "\n";

>       Given that the real per-site difference is larger. Add to this
> the issue that there are real translation problems with the 2nd
> approach (that it cannot be appropriately re-ordered), and IMHO the
> argument here is overwhelmingly against 'cute' operator-overloading
> approaches

Translation? Isn't that for debug logging? Do we translate the strings
in OSL_TRACE et al? That seems like wasted time...

Anyway, for a developer not already well acquainted with the code, the
C++ streams approach has the *huge* advantage that (s)he can add
debugging printfs without finding out how to transform *this* object
into a zero-terminated array of characters. If it makes sense to do
it, just use "<<" and it will do the right thing. *Much* easier.

Also, more generally, in the face of platform-dependent typedefs of
integer types, *I*, for one, never really know which length modifier
to use in the format string :-( %d? %ld? Look at how glibc had to
solve this issue with an extension length modifier 'z' just so that
people can print the size of something!

If I were dictator, the "printf-like format string" approach would be
proscribed. I'm not dictator ;-)

-- 
Lionel