From: Martin Wille (mw8329_at_[hidden])
Date: 2003-12-17 11:33:47
Joel de Guzman wrote:
> Darren Cook wrote:
>> Joel de Guzman wrote:
>>> Robert Ramey wrote:
>>>> Would it be possible that spirit 1.6 also be included in boost 1.31
>>>> perhaps under different directory/namespace ?
>> Hi Joel,
>> I didn't see you directly answer this: could boost/spirit and
>> boost/spirit.1.6 co-exist in boost 1.31?
> Possible? I wouldn't say no. However, there are lots of things that
> should be put into consideration. The first one that comes to my mind
> is namespace and directory structure. I do not think that putting
> 1.6.1 in a different namespace and directory is a good idea. That
> would hurt backward compatibility. v1.6 code should work as before.
I think keeping an 1.6 version of Spirit in Boost would be easy.
And if people want it to be there then we should support it.
Of course, 1.6 would have to live in its own namespace and
mixing stuff from 1.6 with stuff from 1.8 can't be expected to
I don't think the problems caused by putting 1.6 into a namespace
of its own would hurt compatibility too much. After all, all you'd
have to do is to change all occurences of spirit/ to spirit_1_6/
and all occurences of spirit:: to spirit_1_6:: (or to change
using declarations or namespace aliases accordingly).
> I'm sure there are other issues as well. However, I won't close my
> mind on the idea. If boost only had a *smart* configuration based
> download such that when your compiler is VC6, a copy of Spirit 1.6
> is sent instead, it would be ideal.
I don't think that would be helpful. Libraries that intend to
support VC6 et altera depend on 1.6 and can't use 1.8. Shipping
different configurations of Boost would cause problems.
>> On a related question: I've now got a couple of parsers working using
>> spirit 1.6 (once over the not-too-bad learning curve it is a pleasure
>> to use by the way). When I upgrade to 1.31 is anything likely to break?
I disagree. Some things may break due to changes in the directory
structure. However, I think this will be easy to fix in the client
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk