Re: [boost] Re: new initialization lib, any interest?

From: Thorsten Ottosen (nesotto_at_[hidden])
Date: 2002-10-09 12:18:21

> Thorsten Ottosen wrote:
> >>>I think that the stuff you downloaded need some refractoring regarding
> >>>hearder files. I'm considering
> >>>two alternatives:
> >>>
> >>>/init/vector.hpp
> >>>/init/deque.hpp
> >>>... <one for each header>
> >>>/enum.hpp
> >>>
> >>>or
> >>>
> >>>/imap.hpp // all map stuff
> >>>/iseq.hpp   // vector, deque
> >>>/iset.hpp
> >>>/iadap.hpp // stack queue
> >>>/enum.hpp
> >>>
> >>>In the last approach one would also need some config macros s.t.
> >>
> > libraries
> >
> >>>without hash based containers canuse the same header. A third
> >>>alternative would be to provide fine granularity for those who wants it
> >>
> > and
> >
> >>>a really big chunk as now. Suggestions are welcome.
> >>
> >>Sounds good, minimize dependencies and all that.
> >>
> >>Vin
> >
> >
> > so which is better?
>
> I'd prefer the first one
> but I'm sure a lot of people will object to the names being so similar
> to the standard headers
>
> With the second one I'd go with fuller names
> iseq.hpp -> isequential.hpp
> or even -> init_sequential.hpp
>
> I would mainly go for that because it took me a while to figure out what
> seq stood for !! :-|

I also think the first is best. Regarding the names, wouldn't understand why
they couln't
be as above. The user will include it like this:

#include <init/include>

just like I include everything from boost as

#include <boost/xxxx.hpp>

The namespace kinda reflects the directory structure. Moreover, the new
headers actually replaces the old ones
for those who is using it which means it is almost the standard header.
Comments are welcome :-)

--Thorsten