Re: [glob2-devel] boost

[Stéphane Magnenat]

> > Speaking about boost::spirit, don't you think it *does* heavily depends
> > on templates ? ;-)
> > boost::spirit is the reason I added boost deps to liban, but to be
> > honest, I've no intention to use boost::spirit in the future, it's just
> > too undebuggable only for syntaxic sugars. You have not coded a full
> > featured bytecode-based calculator with boost::spirit, haev you?
>
> No, spirit isn't for bytecode, its for text, like parsing the config
> files such as data/buildings.txt, which is what I thought it would be
> usefull for.

Well, I know what spirit is for and I consider that using it produces too 
undebuggable code (and if I want to put bytecode in my calculators that's my 
*right*) ;-)) (but ok, the bytecode thing was for the beauty of the sentence 
and contained no apparent information other that the hint I love bytecode).

The text stream parser is very happy as it is and does not wish to mary 
spirit, it is very materialist, poor soul.

> > So, using boost, ok, abusing of boost, lost of time.
>
> Boost isn't entirely template based, although a great deal of it is. I
> would say that increased compile time isn't neccessarilly a problem,
> Boost developers have been carefull to do their code to reduce code
> bloat and compile times, however using boost will certainly show an
> increase in compile times, the development times and debugging times
> will drop as well.

You know, some people code faster that their g++ compiles... ;-)

But I totally agree with a reasonable usage of boost. Just keep in mind boost 
is not always the holy grail to any of your problem.

Have a nice evening,

Steph

-- 
http://nct.ysagoon.com