From: David Abrahams (david.abrahams_at_[hidden])
Date: 2002-01-09 01:45:00
----- Original Message -----
From: "corwinjoy" <cjoy_at_[hidden]>
> In the case of vec_multiset the bookkeeping that goes on behind the
> scenes is primarily to make sure we maintain this 'positional
> invariant'. In fact, whether you are holding objects directly or
> simply holding pointers to the underlying objects I don't see how one
> can maintain the position of an iterator without using bookkeeping
> (or an address search to find where you are) if the sorting container
> is a vector (the standard containers don't have this problem as they
> are using linked nodes). Here I would be curious to understand how
> the other proposed containers handle this issue. For example in the
> associative indirect vector code, I don't really see how the indirect
> iterator would account for this?
The associative indirect vector offers stability of pointers and references.
How does your sorted vector account for that?
I'm being facetious, of course. The answer to your question is simple: they
trade iterator stability for other useful properties. However, I'm sure you
already understood that, so I am wondering the purpose of your posting?
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk