Re: [boost] [review] Review of Flyweight library begins today January 7

From: JOAQUIN LOPEZ MU?Z (joaquin_at_[hidden])
Date: 2008-01-27 14:23:26

Hello Matías, excuse my late answering, I've been
Internet deprived for the best part of the weekend.

----- Mensaje original -----
De: Matias Capeletto <matias.capeletto_at_[hidden]>
Fecha: Viernes, Enero 25, 2008 6:45 pm
Asunto: Re: [boost] [review] Review of Flyweight library begins	today 
January 7
Para: boost_at_[hidden]

> * What is your evaluation of the design?
> 
> Great. I specially like the simplicity of the out of the box version,
> that will surely cover most of the uses cases.
> 
> I like the current design of configuration parameters.
> With respect to Equality semantics, I have no strong opinion about
> this particular item. I think that the gain in using the &-version is
> too good to be discarded, but I do not like the fact of that small
> possibility of a user being surprised by it. A warning box can be put
> in the docs or this gain can be activated by the user overloading the
> operator== in a per class basis as Joaquin proposed in other 
> thread (A section will be needed in the docs about this and maybe a
> macro to make it easy to activate it).

Another option would be to make the &-version the default and
inform the user about the possibility to override operator== so
as to implement the non-&-version... I'm very undecided about this.

[...]

> * What is your evaluation of the implementation?
> 
> I gave it a quick look. The code is very clean and is already bundle
> with the machinery to pass over compiler quirks. I think that is 
great
> how this library rest over the shoulders of other libraries of boost.
> It has a very nice test suit too.

Thank you. If you allow me a diversion , I think synergies between
Boost libs are one of the best aspects of this project, and this is
why (IMHO) we should resist the temptation to split the package
in smaller units, even if some users request that from time to time:
benefits in the long term compensate for the size of the distribution
and nowadays several hundred MB are not that much of a problem, anyway.

> 
> * What is your evaluation of the documentation?
> 
> Good. It will be cool to have more info about when to use a set
> instead of a hashed container for the factory.
> 
> {{
> Small typo in: tutorial/configuration.html
> Tracking policies > no_tracking >
> There is some _in some_ reduction in memory usage due to the absence
> of reference counters.
> }}

Thanks for spotting the typo! I'll fix it right away.

Thank you for your review,

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo