Nagi Teramo and Dirk Eddelbuettel — written Dec 27, 2014 — source

The purpose of this post is to show how to use Boost::Geometry library which was introduced recently in Rcpp. Especially, we focus on R-tree data structure for searching objects in space because only one spatial index is implemented - R-tree Currently in this library.

Boost.Geometry which is part of the Boost C++ Libraries gives us algorithms
for solving geometry problems. In this library, the Boost.Geometry.Index
which is one of components is intended to gather data structures called
spatial indexes which are often used to searching for objects in space
quickly. Generally speaking, spatial indexes stores *geometric objects’
representations* and allows searching for objects occupying some space or
close to some point in space.

R-tree is a tree data structure used for spatial searching, i.e., for indexing multi-dimensional information such as geographical coordinates, rectangles or polygons. R-tree was proposed by Antonin Guttman in 1984 as an expansion of B-tree for multi-dimensional data and plays significant role in both theoretical and applied contexts. It is only one spatial index implemented in Boost::Geometry.

As a real application of this, It is often used to store spatial objects such as restaurant locations or the polygons which typical maps are made of: streets, buildings, outlines of lakes, coastlines, etc in order to perform a spatial query like “Find all stations within 1 km of my current location”, “Let me know all road segments in 2 km of my location” or “find the nearest gas station” which we often ask google seach by your voice recenlty. In this way, the R-tree can be used (nearest neighbor) search for some places.

You can find more explanations about R-tree in Wikipedia.

Now, we write a simple C++ wrapper class of rtree class in Boost::Geometry::Index that we can use in R.

The most important feature to mention here is the use of Rcpp module to expose your own class to R. Although almost all classes in Boost library have a lot of functions, , you do not use all in many cases. In that case, you should write your wrapper class for making your code simple.

First, we create a sample data set of spatial data.

One can use the RTreeCpp class as follows:

[1] 0

[1] 1 0

[1] 2 1 0

Note the re-creation of the `RTreeCpp`

object is of course
inefficient, but the Rcpp Gallery imposes some constraints on how we
present code. For actual application a stateful and persistent
object would be created. This could be done via Rcpp Modules as
well a number of different ways. Here, however, we need to
recreate the object for each call as `knitr`

(which is used behind
the scenes) cannot persist objects between code chunks. This is
simply a technical limitation of the Rcpp Gallery—but not of Rcpp
itself.

**tags:**
boost

- RcppMLPACK2 and the MLPACK Machine Learning Library — Dirk Eddelbuettel
- Using Rcpp with C++11, C++14 and C++17 — Dirk Eddelbuettel
- Using Armadillo with SuperLU — Qiang Kou and Dirk Eddelbuettel