CGAL::Convex_hull_d<R>

Definition

An instance C of type Convex_hull_d<R> is the convex hull of a multi-set S of points in d-dimensional space. We call S the underlying point set and d or dim the dimension of the underlying space. We use dcur to denote the affine dimension of S. The data type supports incremental construction of hulls.

The closure of the hull is maintained as a simplicial complex, i.e., as a collection of simplices. The intersection of any two is a face of both¹. In the sequel we reserve the word simplex for the simplices of dimension dcur. For each simplex there is a handle of type Simplex_handle and for each vertex there is a handle of type Vertex_handle. Each simplex has 1 + dcur vertices indexed from 0 to dcur; for a simplex s and an index i, C.vertex(s,i) returns the i-th vertex of s. For any simplex s and any index i of s there may or may not be a simplex t opposite to the i-th vertex of s. The function C.opposite_simplex(s,i) returns t if it exists and returns Simplex_handle() (the undefined handle) otherwise. If t exists then s and t share dcur vertices, namely all but the vertex with index i of s and the vertex with index C.index_of_vertex_in_opposite_simplex(s,i) of t. Assume that t exists and let j = C.index_of_vertex_in_opposite_simplex(s,i). Then s = C.opposite_simplex(t,j) and i = C.index_of_vertex_in_opposite_simplex(t,j).

The boundary of the hull is also a simplicial complex. All simplices in this complex have dimension dcur - 1. For each boundary simplex there is a handle of type Facet_handle. Each facet has dcur vertices indexed from 0 to dcur - 1. If dcur > 1 then for each facet f and each index i, 0 ≤ i < dcur, there is a facet g opposite to the i-th vertex of f. The function C.opposite_facet(f,i) returns g. Two neighboring facets f and g share dcur - 1 vertices, namely all but the vertex with index i of f and the vertex with index C.index_of_vertex_in_opposite_facet(f,i) of g. Let j = C.index_of_vertex_in_opposite_facet(f,i). Then f = C.opposite_facet(g,j) and i = C.index_of_vertex_in_opposite_facet(g,j).

Types

Note that each iterator fits the handle concept, i.e. iterators can be used as handles. Note also that all iterator and handle types come also in a const flavor, e.g., Vertex_const_iterator is the constant version of Vertex_iterator. Const correctness requires to use the const version whenever the convex hull object is referenced as constant. The Hull_vertex_iterator is convertible to Vertex_iterator and Vertex_handle.

Creation

The data type Convex_hull_d offers neither copy constructor nor assignment operator.

Requirements

R is a model of the concept ConvexHullTraits_d .

Operations

All operations below that take a point x as argument have the common precondition that x is a point of ambient space.

Lists and Iterators

Iteration Statements

forall_ch_vertices(v,C) { ``the vertices of C are successively assigned to v'' }

forall_ch_simplices(s,C) { ``the simplices of C are successively assigned to s'' }

forall_ch_facets(f,C) { ``the facets of C are successively assigned to f'' }

Implementation

The implementation of type Convex_hull_d is based on [CMS93] and [BMS94]. The details of the implementation can be found in the implementation document available at the download site of this package.

The time and space requirements are input dependent. Let C₁, C₂, C₃, … be the sequence of hulls constructed and for a point x let k_i be the number of facets of C_i that are visible from x and that are not already facets of C_i-1. Then the time for inserting x is O(dim ∑_i k_i) and the number of new simplices constructed during the insertion of x is the number of facets of the hull which were not already facets of the hull before the insertion.

The data type Convex_hull_d is derived from Regular_complex_d. The space requirement of regular complexes is essentially 12(dim +2) bytes times the number of simplices plus the space for the points. Convex_hull_d needs an additional 8 + (4 + x)dim bytes per simplex where x is the space requirement of the underlying number type and an additional 12 bytes per point. The total is therefore (16 + x)dim + 32 bytes times the number of simplices plus 28 + x ⋅ dim bytes times the number of points.

Low Dimensional Conversion Routine

Low Dimensional Output Routines

Footnotes