Re: [Boost-users] Flyweight: wrapping shared_ptr

Subject: Re: [Boost-users] Flyweight: wrapping shared_ptr
From: Akim Demaille (akim_at_[hidden])
Date: 2014-10-22 11:33:38

Le 22 oct. 2014 à 16:24, Joaquin M Lopez Munoz <joaquin_at_[hidden]> a écrit :

> Akim Demaille <akim <at> lrde.epita.fr> writes:
> 
>> 
>> Hi Joaquin!
>> 
>> Le 14 oct. 2014 à 11:34, Joaquin M Lopez Munoz <joaquin <at> tid.es> a
>> écrit :
>>> 
>>> Hashing and equality comparison of flyweight'd expressions is going
>>> to be blazingly fast as compared with recursive non-flyweight'd
>>> versions.
>> 
>> Yes, that's where I expect to get most of my speed up.  However,
>> for deep expressions, ordered containers were quite fast, more
>> than hashes, since often the comparison does not go deep inside.
>> So I'll have to bench whether to stick to ordered containers, or
>> move to unordered ones.
> 
> I'm not sure I'm following you here... Comparison and hashing of flyweights
> is blazingly fast *and* does not depend on the expression depth of
> the flyweight'd values --it does not depend on the values at all, in fact.

Yes, of course.  I was referring (with the "were", I agree it's subtle :)
to the pre-Flyweight implementation.  I was commenting that I was surprised
that in general ordered structures were faster that unordered ones, and
I suspect it's mostly because hashing is always deep, and comparison
is not.

However, I don't like Flyweight address comparison.  Sure, it's fast.
It also not well defined, not repeatable, not portable etc.  At least
when developing, I like having something more stable, so I will probably
stick with a deep comparison.

So I was saying that moving to flyweights also means, in my case,
looking for places where unordered is ok, and move to unordered in
these places.