Re: [Bloom] Some questions

From: Peter Dimov (pdimov_at_[hidden])
Date: 2025-05-19 18:36:38

Joaquin M López Muñoz wrote:
> El 18/05/2025 a las 19:29, Ivan Matek escribió:
> >
> >
> > On Sun, May 18, 2025 at 5:59 PM Joaquin M López Muñoz via Boost
> > <boost_at_[hidden]> wrote:
> >
> >     I don't know, looks to me ugly and potentially confusing to users,
> >     who have no real use for that.
> >
> > To be honest I do not think it is so bad, you could just document it
> > is a common "primitive" used when it comes to index computation, and
> > then users will know why it is where it is.
> > Only thing I suggest is to not put it in core. I have seen
> > core/util/common/whatever_name bloat in many projects and I am not a
> > fan of that.
> >
> >     In fact, there's another piece of functionality that I'd rather have
> >     migrated to ContainerHash, namely the is_avalanching trait. Currently
> >     this lives in Boost.Unordered, which forces Boost.Bloom to depend
> >     on it
> >     for no other reason than to access this almost trivial utility
> >     --not ideal,
> >     obviously.
> >
> > I would be a big fan philosophically of trait being in a small
> > header/library. For related example I dislike that for iterator_tags
> > you must include heavy std header, when it is trivial functionality.
> > IIRC there are even Boost libs that are defining them just to get
> > around the include. :)
> 
> Boost.IsAvalanching? Not sure I'm buying that :-)

I already have to declare hash_is_avalanching in ContainerHash

https://github.com/boostorg/container_hash/blob/b8179488b20eb1373bdbf5c7fcca963f072512df/include/boost/container_hash/hash.hpp#L564

so I have no objections to moving it there.

IIRC we originally kept in in Unordered because Unordered was C++11
and ContainerHash wasn't yet. But now it is.