Re: Any interest in Tuples?

From: William Kempf (sirwillard_at_[hidden])
Date: 2000-11-30 12:18:26

--- In boost_at_[hidden], Jeremy Siek <jsiek_at_l...> wrote:
> 
> It really would be good to have VC++ working.
> 
> Would it be possible to have a somewhat dumbed-down version that 
works for
> VC++? Would limiting the tuples to 5 elements or so make it 
possible to
> have it working in VC++?

I've not totally given up on VC++ yet.  Some things mentioned the 
last time I posted about my attempts sparked some thoughts that I've 
yet to try out.  There were also a few other people interested in 
taking what I had so far and trying to get further (I didn't forget 
any of you, and will e-mail it to you in the near future... my time 
just got subverted by several things and I need to clean up the code 
before I can send it, since I've moved it to a non-working state 
while tinkering with it).  I also have hopes that VC++ 7 (which I 
have a beta of now, but haven't had time to evaluate) might have 
fixed some of the problems preventing the partial template simulation 
before, not to mention an actual bug in the code produce that I 
discovered that prevented the tie from working.

Limiting the number of elements won't help.  Removing the cons 
concept and storing all elements, even "unused" ones, would probably 
be workable, but would also likely be wasteful of memory.

My conclusions:  At some time we'll have VC++ versions of tuple (even 
if we must wait until VC++ 8 is released).  It may even be tomorrow 
that we have such a version.  However, because of the difficulty 
involved I think we should exclude VC++ compatibility as a 
requirement before submitting tuples to Boost.  I think the concept 
is too important to delay just because of a broken compiler.  As soon 
as Gary Powell and Jaakko Jarvi finish breaking the code out of LL 
and Boostifying the code I think it should be submitted for approval, 
regardless of VC++ compatibility.

Just my $.02.

Bill Kempf