|
Boost : |
From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-01-20 13:28:05
----- Original Message -----
From: "Alan Bellingham" <alan_at_[hidden]>
To: <boost_at_[hidden]>
Sent: Sunday, January 20, 2002 1:11 PM
Subject: Re: [boost] More ramblings on a sorted vector container
> Dave:
>
> >Not even. Well, that sounds like a nice approach, but I also want a way
to
> >say "just put it here", and its up to me to make sure things are sorted
> >before the next lookup. The standard associative containers cause me lots
of
> >grief by trying too hard to protect me from the low-level details.
>
> So (for example) you could dump a load of elements into the middle,
> knowing they're already sorted and in the right place, and avoid the
> sort at all? Seems reasonable to me.
Precisely. Access to the underlying container might be enough.
Maybe you can see why I favor an adaptor approach: a linear associative
adaptor could provide the associative interface on top of a vector or deque.
> At the moment, it appears that this gives us three policies:
>
> 1) Sort immediately.
> 2) Sort automatically when examined.
> 3) Sort only when user demands it. (The progammer is always right.)
>
> I'll go away for a bit while looking at that previous submisson.
Great!
-Dave
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk