# Utilities

## Operations

Base.:+Function
+(p1::VRep, p2::VRep)

Minkowski sum between p1 and p2 using the V-representation. If the V-representation is not computed for p1 or p2, it is computed.

+(p::Rep, el::Union{Line, Ray})
+(el::Union{Line, Ray}, p::Rep)

Same as p + vrep([el]).

source
Base.:*Function
*(p1::Rep, p2::Rep)

Cartesian product between the polyhedra p1 and p2.

source
*(P::Union{AbstractMatrix, UniformScaling}, p::VRep)

Transform the polyhedron represented by $p$ into $P p$ by transforming each element of the V-representation (points, symmetric points, rays and lines) x into $P x$.

source
*(α::Number, p::Rep)

Transform the polyhedron represented by $p$ into $\alpha p$ by transforming each element of the V-representation (points, symmetric points, rays and lines) x into $\alpha x$.

source
Base.:\Function

(P::Union{AbstractMatrix, UniformScaling}, p::HRep)

Transform the polyhedron represented by $p$ into $P^{-1} p$ by transforming each halfspace $\langle a, x \rangle \le \beta$ into $\langle P^\top a, x \rangle \le \beta$ and each hyperplane $\langle a, x \rangle = \beta$ into $\langle P^\top a, x \rangle = \beta$.

source
Base.:/Function
/(p::HRep, P::Union{AbstractMatrix, UniformScaling})

Transform the polyhedron represented by $p$ into $P^{-T} p$ by transforming each halfspace $\langle a, x \rangle \le \beta$ into $\langle P a, x \rangle \le \beta$ and each hyperplane $\langle a, x \rangle = \beta$ into $\langle P a, x \rangle = \beta$.

source
Base.intersectFunction
intersect(P1::HRep, P2::HRep)

Takes the intersection of P1 and P2 $\{\, x : x \in P_1, x \in P_2 \,\}$. It is very efficient between two H-representations or between two polyhedron for which the H-representation has already been computed. However, if P1 (resp. P2) is a polyhedron for which the H-representation has not been computed yet, it will trigger a representation conversion which is costly. See the Polyhedral Computation FAQ for a discussion on this operation.

The type of the result will be chosen closer to the type of P1. For instance, if P1 is a polyhedron (resp. H-representation) and P2 is a H-representation (resp. polyhedron), intersect(P1, P2) will be a polyhedron (resp. H-representation). If P1 and P2 are both polyhedra (resp. H-representation), the resulting polyhedron type (resp. H-representation type) will be computed according to the type of P1. The coefficient type however, will be promoted as required taking both the coefficient type of P1 and P2 into account.

source
intersect(v::VRepresentation{T}, h::HRepElement)

Compute the intersection of v with an halfspace or hyperplane h. The method used by default is to keep the V-representation element of v that are in h and add new ones generated as the intersection between the hyperplane defining h and the segment between two adjacent V-representation elements of v that are in either sides of the hyperplane. See Lemma 3 of [FP96] for more detail on the method.

[FP96] Fukuda, K. and Prodon, A. Double description method revisited Combinatorics and computer science, Springer, 1996, 91-111

source
Polyhedra.convexhullFunction
convexhull(P1::VRep, P2::VRep)

Takes the convex hull of P1 and P2 $\{\, \lambda x + (1-\lambda) y : x \in P_1, y \in P_2 \,\}$. It is very efficient between two V-representations or between two polyhedron for which the V-representation has already been computed. However, if P1 (resp. P2) is a polyhedron for which the V-representation has not been computed yet, it will trigger a representation conversion which is costly.

The type of the result will be chosen closer to the type of P1. For instance, if P1 is a polyhedron (resp. V-representation) and P2 is a V-representation (resp. polyhedron), convexhull(P1, P2) will be a polyhedron (resp. V-representation). If P1 and P2 are both polyhedra (resp. V-representation), the resulting polyhedron type (resp. V-representation type) will be computed according to the type of P1. The coefficient type however, will be promoted as required taking both the coefficient type of P1 and P2 into account.

source
convexhull(p1::HRepresentation, p2::HRepresentation)

Returns the Balas [Theorem 3.3, B85] extended H-representation of the convex hull of p1 and p2.

[B85] Balas, E., 1985. Disjunctive programming and a hierarchy of relaxations for discrete optimization problems. SIAM Journal on Algebraic Discrete Methods, 6(3), pp.466-486.

source
Polyhedra.translateFunction
translate(p::Polyhedra.Rep, v::AbstractVector)

Computes translation of the polyhedron p with the vector v. That is, computes

$$$\{\, x + v \mid x \in p \,\}.$$$

By default, if the H-representation, it simply translates every hyperplanes and halfspace, otherwise, it translates every points of the V-representation. That is, this operation can be achieved both in the H-representation and V-representation hence does not trigger any representation conversion.

source
Polyhedra.polarFunction
polar(rep::Representation)

Return the polar of the polyhedron rep assumed to contain the origin. The polar of a convex set S is defined as the set of y such that ⟨x, y⟩ ≤ 1 for all x in S. Note that the polar of a V-representation is a H-representation and vice versa.

source

## Volume

Polyhedra.volumeFunction
volume(p::Polyhedron{T}) where {T}

Returns the fulldim(p)-dimensional hyper-volume of the polyhedron p. Returns Inf or -one(T) if it is infinite depending on whether the type T has an infinite value.

source
Polyhedra.surfaceFunction
surface(p::Polyhedron{T}) where {T}

Returns the fulldim(p)-1-dimensional hyper-volume of the surface of the polyhedron p. Returns Inf or -one(T) if it is infinite depending on whether the type T has an infinite value.

source
Polyhedra.center_of_massFunction
center_of_mass(p::Polyhedron{T}) where {T}

Returns the center of mass of p, represented as a Vector{T} of length fulldim(p). Throws an error if p is degenerate.

source

## Largest inscribed ball with center

Polyhedra.maximum_radius_with_centerFunction
maximum_radius_with_center(h::HRep, center)

Return the maximum radius r such that the Euclidean ball of center center and radius r is included in the polyhedron h.

source

## Chebyshev center

Polyhedra.chebyshevcenterFunction
chebyshevcenter(p::Rep[, solver])

If p is a H-representation or is a polyhedron for which the H-representation has already been computed, calls hchebyshevcenter, otherwise, call vchebyshevcenter.

source
Polyhedra.hchebyshevcenterFunction
hchebyshevcenter(p::HRep[, solver]; linearity_detected=false, proper=true)

Return a tuple with the center and radius of the largest euclidean ball contained in the polyhedron p. Throws an error if the polyhedron is empty or if the radius is infinite. Linearity is detected first except if linearity_detected.

Note that a polytope may have several Chebyshev center. In general, the set of Chebyshev center of a polytope p is a polytope which has a lower dimension than p if p has a positive dimension. For instance, if p is the rectangle [-2, 2] x [-1, 1], the Chebyshev radius of p is 1 and the set of Chebyshev centers is [-1, 1] x {0}. The proper Chebyshev center is (0, 0), the Chebyshev center of [-1, 1] x {0}. If !proper then any Chebyshev center is returned (the one returned depends on the solver). Otherwise the proper Chebyshev center is computed. The proper Chebyshev center is defined by induction on the dimension of p. If p has dimension 0 then it is a singleton and its proper Chebyshev center is the only element of p. Otherwise, the dimension of the set q of Chebyshev centers of p is smaller than the dimension of p and the proper Chebyshev center of p is the proper Chebyshev center of q.

source
Polyhedra.vchebyshevcenterFunction
vchebyshevcenter(p::VRep[, solver])

Return a tuple with the center and radius of the smallest euclidean ball containing the polyhedron p. Throws an error if the polyhedron is empty or if the radius is infinite (i.e. p is not a polytope, it contains rays).

source

## Defining new representation

The following macros make it easy to define new representations: