From: Arkadiy Vertleyb (vertleyb_at_[hidden])
Date: 2004-04-11 18:43:29
"JOAQUIN LOPEZ MU?Z" <joaquin_at_[hidden]> wrote
> I'm really lost here, naming issues are so nasty.
> Looking fwd to knowing your ideas.
I think we have a rather generic situation here, which is the library has
one main class. In which case it would be rather natural to have the
class's name the same as the library name, and the same as namespace name.
Except that the class name would clash with the namespace.
There are a few other libraries in Boost that are like this, such as
lexical_cast, threads, function, etc. Some don't have their own namespaces,
and just live in "boost" (however, the question is what to do with the
supporting classes? There might be quite a few of them, and one doesn't
want to pollute the boost namespace with the library-specific things).
Others add the suffix 's' to form the namespace name (like tuple and
I think at least one good name was found here, and this is "multi_index". I
honestly think this name should be reserved for the class, and this class
should be under "boost" (or promoted to boost). This is what you had
before, when you had "indexed_set" and "indexed_sets". I kind of lost the
track of why this kind of naming was rejected.
I think that in the situation like this, the library name and the namespace
name should be just derived from the class name using one consistent scheme.
So what about accepting the Boost.Tuple scheme, and having "multi_index",
"Boost.MultiIndex", and "multi_indexes" (or multi_indices)? I would still
preffer "indexes", but native English speakers might not agree with me :)
Boost list run by bdawes at acm.org, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk