Boost logo

Geometry :

Subject: [ggl] spacial index
From: Barend Gehrels (barend)
Date: 2011-03-29 07:28:14

Hi Adam, Bruno,

I agree with more detailing the namespaces.

> It would be good to have various indexes inside geometry::index
> namespace to avoid redundant names:
> geometry::index::some_index::some_index<V> i;

This is inconvenient. It is not impossible (it will compile) but we've
had this before, within distance (the algorithm) and namespace distance.

What is possible is:
geometry::index::rtree<V> (for the rtree index)

This is also more intuitive. You mentioned this also below, so this is
probably what you mean.

> Furthermore I'd like to hide implementation deeper in the namespaces
> hierarchy. So maby a good idea is to have spatial index related code
> in namespace geometry::index::details::some_index.

I agree.

> We would have short path to what's most important:
> geometry::index::rtree<...>


> geometry::index::translator::def<...>
> geometry::index::filters::spatially_filtered(...)
> geometry::index::traits::some_traits<...>
> separated implementation hierarchy
> geometry::index::detail::rtree::node<...>
> geometry::index::detail::rtree::visitors::insert<...>
> geometry::index::detail::rtree::visitors::remove<...>
> and spatial index variant implementation details, e.g. rstar related
> code, deeper
> geometry::index::detail::rtree::rstar::split<...>
> geometry::index::detail::rtree::rstar::choose_next_node<...>

This looks all good to me.

Regards, Barend

Geometry list run by mateusz at