Subject: [Boost-commit] svn:boost r84862 - trunk/boost/geometry/index/detail/rtree/linear
Date: 2013-06-21 11:53:09
Date: 2013-06-21 11:53:09 EDT (Fri, 21 Jun 2013)
New Revision: 84862
[geometry][index]: Added a comment regarding implementation of various Indexable types in linear find_greatest_normalized_separation.
Text files modified:
trunk/boost/geometry/index/detail/rtree/linear/redistribute_elements.hpp | 14 +++++++++++++-
1 files changed, 13 insertions(+), 1 deletions(-)
--- trunk/boost/geometry/index/detail/rtree/linear/redistribute_elements.hpp Fri Jun 21 11:09:08 2013 (r84861)
+++ trunk/boost/geometry/index/detail/rtree/linear/redistribute_elements.hpp 2013-06-21 11:53:09 EDT (Fri, 21 Jun 2013) (r84862)
@@ -50,7 +50,19 @@
return difference_dispatch<R>(from, to, is_unsigned());
-// TODO: awulkiew - there are loops inside find_greatest_normalized_separation::apply()
+// TODO: awulkiew
+// In general, all aerial Indexables in the tree with box-like nodes will be analyzed as boxes
+// because they must fit into larger box. Therefore the algorithm could be the same for Bounds type.
+// E.g. if Bounds type is sphere, Indexables probably should be analyzed as spheres.
+// 1. View could be provided to 'see' all Indexables as Bounds type.
+// Not ok in the case of big types like Ring, however it's possible that Rings won't be supported,
+// only simple types. Even then if we consider storing Box inside the Sphere we must calculate
+// the bounding sphere 2x for each box because there are 2 loops. For each calculation this means
+// 4-2d or 8-3d expansions or -, / and sqrt().
+// 2. Additional container could be used and reused if the Indexable type is other than the Bounds type.
+// TODO: awulkiew
+// there are loops inside find_greatest_normalized_separation::apply()
// iteration is done for each DimensionIndex.
// Separations and seeds for all DimensionIndex(es) could be calculated at once, stored, then the greatest would be choosen.
Boost-Commit list run by bdawes at acm.org, david.abrahams at rcn.com, gregod at cs.rpi.edu, cpdaniel at pacbell.net, john at johnmaddock.co.uk