Re: [boost] Boost.Bimap - Informal Review Request

From: John Maddock (john_at_[hidden])
Date: 2006-08-10 12:18:23

Matias Capeletto wrote:
> There are actually two relation classes: "mutant_relation", that uses
> this almost-standard
> guarantee when possible and "standard_relation" that is standard
> compliant. You can
> read more about this here:
>
> http://cablemodem.fibertel.com.ar/mcape/boost/libs/bimap/boost_bimap/rationale.html#boost_bimap.rationale.general_design.relation_implementation
>
> The idea is to switch between this two classes at compile time using a
> conservative approach, so the library can get the standard compliant
> label. You can see the switching mechanism here:
>
> http://cablemodem.fibertel.com.ar/mcape/boost/libs/bimap/doxydoc/relation_8hpp-source.html
>
> At this point it is very conservative, and I think it is downgrading
> to the slower relation in cases where this can be avoided (for
> example, if you do not have 1 byte alignment). Less conservative
> checks can be developed in the future.

Looks like you're way ahead of me.  BTW almost no compilers use 1 byte 
alignment.

John.