Re: [Boost-users] bind doesn't use tuple

From: Joel de Guzman (joel_at_[hidden])
Date: 2007-03-22 02:53:42

Sohail Somani wrote:
> Hi, I think whether or not there was a reason to not do it in the first 
> place, the idea should be revisited. I recently implemented a bit of 
> code that figured out the minimum number of parameters that would be 
> required for an arbitrary bind expression. After a while, there was a 
> boatload of code that was just to do with getting around the current 
> bind implementation's decisions that I decided not to pursue it any 
> further - the thing I really wanted to do was do a fusion::for_each... 
> But that was not to be for me!
> 
> -----Original Message-----
> From: boost-users-bounces_at_[hidden] on behalf of Gottlob Frege
> 
> why doesn't bind use tuple (internally)?
> 
> I'm tempted to leave the email as just that. No more to say.  OK,
> maybe one thing - I'm not asking because I necessarily think it
> should, I can imagine there are reasons, just wondering what they
> might be.

FWIW, Phoenix2 uses fusion. I wish I had more time to finally do the
Lambda/Phoenix merger. Unfortunately, now, I'm too busy with Spirit2
and the Boost conference is just around the corner!

Interesting to note that Fusion was developed for the Phoenix/
Lambda merger (among other things).

I'm hoping someone else would want to pick up where I left off.
Dan? Tobias? Joao? I'm CC'ing you guys :-)

(Historical note: lambda did not use tuples for performance reasons.
Jaakko noted that fusing the args had some overhead. That was then
(g++ 2.95.3). Now, I have benchmarks to prove that the fusing of
args into tuples has zero overhead on modern compilers such as
VC7.1/8.0/intel. Alas, g++, still has an abstraction penalty.
Let's hope g++ catches up in terms of optimization).

Regards,
-- 
Joel de Guzman
http://www.boost-consulting.com
http://spirit.sf.net