Re: [boost] [filesystem] relative path from arbitrary file system location?

Subject: Re: [boost] [filesystem] relative path from arbitrary file system location?
From: Scott McMurray (me22.ca+boost_at_[hidden])
Date: 2009-02-06 23:19:03

On Fri, Feb 6, 2009 at 22:32, Kenny Riddile <kfriddile_at_[hidden]> wrote:
>
> Also, may I suggest the name "as_relative_to" for the member function if
> this is added to the library? as_relative_to would be a const member method
> that take the desired base path as its only parameter and returns the
> desired relative path.
>

The problem I see with that is that to handle symlinks elegantly,
"uncomplete" will need to look at the filesystem, not just the path
elements, which, following as I understand the way things are
organized, prevents it from being part of the path class itself.
(This is the same problem that keeps coming up: ./foo/bar/.. doesn't
have to be ./foo when symlinks are involved.)  In any case, it's the
inverse function to complete, so it should be the same kind of beast.
(It seems p.has_root_path() ? complete(relative(p)) :
relative(complete(p)) should be equivalent to p, assuming compatible
root_name()s in the first branch.)

As I mention in the ticket, I really like the idea of calling the
function relative, but I'm worried about conceptual conflict with the
relative_path() decomposition function (which I don't really like,
since if complete uses initial_path, it seems like relative would be
to initial_path, not to the root of the path in question, though the
member/nonmember may be enough to disambiguate).