[boost] Re: variant questions

From: David Abrahams (dave_at_[hidden])
Date: 2003-08-29 16:35:43

"Eric Friedman" <ebf_at_[hidden]> writes:

> David Abrahams wrote:
> [snip]
>> 2.
>>
>>        "All members of variant satisfy the strong guarantee of
>>        exception-safety."
>>
>>     Seriously?  What if an underlying type's assignment operator gives
>>     only the basic guarantee?  Surely, if you in fact use the
>>     underlying type's assignment operator you can't give the strong
>>     guarantee?
>
> I guess I should address this one a little more since I'm not exactly sure
> if even the updated documentation describes variant correctly in this
> regard.
>
> Given a variant type with bounded types T1,...,Ti,...,Tj...,Tn and variant
> objects vi (with content of type Ti) and vj (with content of type Tj), the
> exception safety guarantee for variant is as follows:
>
> Assignment:
>   vi = Ti(...)  // guarantee equal to Ti::operator=
>   vi = vi // guarantee equal to Ti::operator=
>   vi = vj // strong guarantee
>
> Swap:
>   swap(vi, vi) // guarantee equal to swap( Ti, Ti )
>   swap(vi, vj) // strong guarantee
>
> Please let me know if the BoostBook documentation clearly describes what I
> present here.

I will when I find out where it is.

BTW, after looking at the implementation I was a bit disappointed to
see two copies of the storage.  It seems to nullify one
important reason for using variants (space savings), and it generates
more code than a single-storage version.  I know you had some rationale for
that but I don't remember what it is; I hope it's in the docs along
with some explanation of the expected size of the resulting object.

Regards,
Dave

-- 
Dave Abrahams
Boost Consulting
www.boost-consulting.com