Re: [Boost-users] [flyweight] Review

From: Joaquín M LópezMuñoz (joaquin_at_[hidden])
Date: 2008-01-30 17:24:17

Hello Neil, thanks for your review!

Neil Hunt <boost <at> thehunts.id.au> writes:

> * What is your evaluation of the design?
> I like ability for the drop in replacement of const T (with assignability).
> It makes it easy to test whether it makes a big difference in run time
> scenarios, before committing to its use.

Well, have in mind drop in replacement only works in
environments where implicit convertibility to T is allowed.
For instance, the following won't compile:

  flyweight<std::string> s;
  ...
  s.append(...);

You'll have to write instead:

  flyweight<std::string> s;
  ...
  s.get().append(...);

If only we had smart references in C++ :(

[...]
> Of course, real world mileage will vary, but some performance indicators 
> or at least discussion of the tradeoff of extra hashing vs reduced dynamic 
> allocations would be useful, as well as any other performance impacts 
> (how big should the data set be before using it, what percentage of unique
> values makes it worthwhile).

I agree with you performance is likely very dependent on the
particular usage scnario, but anyway the section you propose
makes a lot of sense, so I'll put it in my todo list. FWIW, there
is a performance measuring example that you can play with at:

http://svn.boost.org/svn/boost/sandbox/flyweight/libs/flyweight/doc/examples.ht
ml#example5

> 
> * What is your evaluation of the potential usefulness of the library?
> Â 
> Useful to have available. Unless there is a no brainer scenario,
> it seems that real world testing on live data will be needed to
> evaluate it's value in any given scenario?

I couldn't agree more. I was hoping some of the reviewers had
some real scenario of her own where to put the lib in practice.

Best regards,

Joaquín M López Muñoz
Telefónica, Investigación y Desarrollo