Re: [boost] [review] Fusion Library Accepted

From: Joel de Guzman (joel_at_[hidden])
Date: 2006-06-06 05:43:42

Jaakko Järvi wrote:
> On Jun 5, 2006, at 9:09 PM, David Abrahams wrote:
> 
>> dan marsden <danmarsden_at_[hidden]> writes:
>>
>>> Of course if somebody has a good idea for natural sounding name,  
>>> that would be cool, and may
>>> help new users identify the library as being suitable/unsuitable  
>>> for their needs.
>> tuples (it's supposed to replace boost.tuple eventually, right?)
>>
> Hmm... That is really what fusion to a large extent is.
> (Old) tuples are mainly a tuple data structure, Fusion is a tuple  
> data-structure (possibly many of them),
> the definition of the interfaces of any tuple-like data-structure,  
> and routines that use this interface
> for manipulating tuples (or tuple-like data-structures)
> 
> And yes, fusion should replace boost.tuple.
> 
> If Boost.Tuple is the baseline of tuples, Fusion could be viewed as  
> "tuples on steroids"
> If Fusion is the baseline of tuples, Boost.Tuple could be viewed as  
> "poor man's tuples"
> 
> So tuple could be a descriptive name --- using it requires that the  
> "switchover" gets done of course.

Fusion can't replace Boost.Tuples while Boost still supports legacy
compilers such as Borland and VC6/7. There was once a Fusion port to
VC6/7 (but not Borland) by Peder Holt (Fusion-1), but Fusion-2 is almost
a total rewrite, I sense that I'll find it impossible to ask him to
invest the time for another port.

I'm not sure until when Boost will continue to support these vintage
compilers. Until we decide to move on and leave those compilers behind,
I suggest:

1) We continue to have Boost.Tuples as-is
2) Make Boost.Tuple a conforming Fusion sequence

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