Toblerity/rtree

QUESTION: How can rtree.index.Index properties be used to tweak performance ?

manuel-koch opened this issue · 2 comments

manuel-koch commented

Looking at the docstring of Property it is not clear to me what properties can be used to tweak the performance of the index for individual use cases.

Is it possible to enhance the descriptions of the properties to give hints what values can be used to adapt to certain scenarios:

  • when using many differently sized AABBs
  • when using mostly equal sized AABBs
  • when most of the AABBs are clustered at a special region within the whole Index

E.g. I would like to know details ( typical values, range of values, use cases for values ) about properties like :

  • fill_factor
  • leaf_capacity
  • near_minimum_overlap_factor
  • reinsert_factor
  • split_distribution_factor
  • tight_mbr
  • variant, what are the differences between RT_Linear, RT_Quadratic, and RT_Star

Thank you for providing this library !

hobu commented

These of course all point into https://github.com/libspatialindex/libspatialindex/. Some more description about these properties can be found in the source code there, but it certainly isn't going to be the comprehensive documentation you're seeking.

I don't have much to offer in terms of guidance other than to cook up some simulations and permute parameters to see their effect.

manuel-koch commented

overview.rst, RTree.h
gives very limited hints.