Re: [boost] Re: Re: Proposal --- A type-safe union

From: Fernando Cacciola (fcacciola_at_[hidden])
Date: 2002-06-13 09:16:01

----- Original Message -----
From: "Andrei Alexandrescu" <andrewalex_at_[hidden]>
Newsgroups: gmane.comp.lib.boost.devel
To: <boost_at_[hidden]>
Sent: Thursday, June 13, 2002 3:18 AM
Subject: [boost] Re: Re: Proposal --- A type-safe union

> [snipped]
>
> What I also believe, in contradiction with some participants, is that
> Variant should use typelists. If not, then Variant will factionally
> develop its own little typelist facility (as some other libraries in
> boost do today) and this doesn't work very nicely.
>
>
I realized that a typelist interface is the only way (ok, the easiest way)
to allow metaprograms to generate a variant of a set of types collected by
the metaprogram.
That is, with variant<types>, I have all ways to generate the type-set at
compile-time (for instance, using a filter to make sure all types are
different). This is a lot more difficult with variant<T0,T1,...,TN>.

BTW, does variant *require* all types to be different? I think this
requirement is reasonable, and it could turn out to be quite useful for
variant based algorithms.

Fernando Cacciola
Sierra s.r.l.
fcacciola_at_[hidden]
www.gosierra.com