From: Gennaro Prota (gennaro_prota_at_[hidden])
Date: 2002-11-01 12:57:03
On Wed, 30 Oct 2002 14:46:51 -0300, "Fernando Cacciola"
>Look for the threads [...] specially at the participation from Anthony
>Williams who brought up the technique I was referring to.
I gave it a look. The technique is remarkable, and IMHO useful when a
discriminated union is needed/opportune by design.
But in the example I was thinking of (a generic buffer<T>) the union
would be there as an "optimization" only. That stuff + the
double-buffer technique gives quite a raise to code complexity with an
overall "benefit" which is really difficult to measure (besides being
presumably variable among implementations).
And then, I was thinking to an aid for dynamic_bitset where T is a
built-in type so who cares about exceptions? :-) Well, just kidding!
However I think that's not worth the pain. Any other opinions?
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk