Subject: [ggl] space partitioning
From: Adam Wulkiewicz (adam.wulkiewicz)
Date: 2010-08-26 09:00:29
Barend Gehrels wrote:
> hi Adam,
>
>
>>>
>>>> In the attachment there are:
>>>> runtime_point.cpp - geometry::point concept wrapper providing runtime
>>>> get(index) and set(index, value) methods used in one of the algorithms,
>>>> box.cpp - geometry::box algorithms
>>>> kd-tree.cpp - kd-tree implementation, there should be another tag
>>>> used, indicating type of objects (in this case point_tag).
>
> I've a small problem compiling your code.
>
> From the main program:
> spar::left_aabb<0>(b, 0.0f);
> spar::left_aabb(0, rb, 0.0f);
>
> are mutually inconsistent, they (or one of them) are not compiling,
> maybe the test program was based on another version.
>
> Furthermore the code in RuntimePoint is based on constructions as
> "typename Point::coordinate_type". This was OK, for a geometry::point
> only, until recently (Aug 15). But it actually should be "typename
> coordinate_type<Point>::type" to support also point types NOT being a
> geometry::point. This was also the (one of) reason to remove that
> coordinate_type from point itself, so it is not there anymore...
> Therefore I personally cannot compile it (and you will get errors as
> soon as you update).
>
> This is also for two other metafunctions, so to compile it for any point
> type:
> 1) typename Point::coordinate_type -> typename coordinate_type<Point>::type
> 2) typename Point::coordinate_system -> typename
> coordinate_system<Point>::type
> 3) Point::coordinate_count-> dimension<Point>::value
Sorry that I haven't responsed earlier.
Yes, I'm using the old version, thanks. But I don't know if I'll make
corrections now. It seems that dynamic self balancing spacial index is
desirable. What do you think about it?
Regards,
Adam