TypedPolynomials

TypedPolynomials.jl provides an implementation of the multivariate polynomial interface from MultivariatePolynomials.jl using strongly typed variables. That is, in this package, the identity of a variable is encoded by its type, so variables x and y are of different types. This allows us to use the type system to handle certain operations, like computing the intersection of two monomials' variables, at compile-time.

Features

Handling variables at the type level makes constructing variables, monomials, and terms more efficient than with DynamicPolynomials.jl.
Despite the heavy use of the type system, this package has no @generated functions and should be compatible with static compilation (though this has not yet been tested).

Caveats

There is no distinction in this package between a variable's name and its identity. That is, two variables named x are exactly the same object, regardless of how they were created.
For problems with large numbers of variables, or for which the set of variables is not known at compile-time, you may see better performance overall with DynamicPolynomials.jl, e.g. #32. This may change in the future.

Usage

The easiest way to create variables is the @polyvar macro:

julia> @polyvar x y z  # Declare three `Variable`s named x, y, and z and assign local variables with the same names
(x, y, z)

julia> typeof(x)
TypedPolynomials.Variable{:x}

julia> typeof(y)
TypedPolynomials.Variable{:y}

Multiplying variables creates a Monomial{V} where V is the vector of variables contained in the monomial:

julia> x * y
xy

julia> typeof(x * y)
TypedPolynomials.Monomial{(x, y),2}

julia> typeof(x^2)
TypedPolynomials.Monomial{(x,),1}

Multiplying a monomial (or variable) by anything other than another Variable or Monomial creates a Term:

julia> 3 * x
3x

julia> typeof(3 * x)
TypedPolynomials.Term{Int64,TypedPolynomials.Monomial{(x,),1}}

julia> typeof(3.0 * x^2 * y)
TypedPolynomials.Term{Float64,TypedPolynomials.Monomial{(x, y),2}}

Addition or subtraction of terms, monomials, and/or variables creates a Polynomial:

julia> x + y
x + y

julia> typeof(x + y) <: Polynomial
true

julia> x + 3y^2 + z/2 + x^3
x^3 + 3.0y^2 + x + 0.5z

More Examples

Differentiation and Substitution

using TypedPolynomials
using Test
@polyvar x y # assigns x (resp. y) to a variable of name x (resp. y)
p = 2x + 3.0x*y^2 + y
@test differentiate(p, x) == 3y^2 + 2 # compute the derivative of p with respect to x
@test differentiate.(p, (x, y)) == (3y^2 + 2, 6*x*y + 1) # compute the gradient of p
@test p((x, y)=>(y, x)) == 2y + 3y*x^2 + x  # replace any x by y and y by x
@test p(y=>x^2) == 2x + 3x*(x^4) + x^2 # replace any occurence of y by x^2
@test p(x=>1, y=>2) == 2 * 1 + 3 * 1 * 2^2 + 2 # evaluate p at [1, 2]

Vectors of Variables

The @polyvar macro can also create a tuple of variables automatically:

using TypedPolynomials
A = rand(3, 3)
@polyvar x[1:3] # assign x to a tuple of variables x1, x2, x3
p = sum(x .* x) # x_1^2 + x_2^2 + x_3^2
p(x[1]=>2, x[3]=>3) # x_2^2 + 13
p(x=>A*vec(x)) # corresponds to dot(A*x, A*x), need vec to convert the tuple to a vector

fgerick / typedpolynomials.jl Goto Github PK