Closed MikaelSlevinsky closed 8 years ago
DEQuadrature.jl has been refactored to use SincFun.jl's Domains and ConformalMaps.
Does it make sense to specify domains with only numeric data and not functions? For example, the Finite
domain could look like this instead:
type Finite{T} <: Domain{T}
ab::Vector{T} # endpoints of the interval
algebraic::Vector{T} # exponents of the algebraic singularities of the endpoints
logarithmic::Vector{T} # exponents of the logarithmic singularities of the endpoints
end
Then one could define the outer map, its inverse and derivative (and any other Function
s that may be required) as:
ψ{T}(d::Finite{T},t) = (d.ab[1]+d.ab[2])/2+(d.ab[2]-d.ab[1])/2*tanh(t)
ψinv{T}(d::Finite{T},t) = atanh((2t-d.ab[1]-d.ab[2])/(d.ab[2]-d.ab[1]))
ψp{T}(d::Finite{T},t) = (d.ab[2]-d.ab[1])/2*sech(t).^2
...
There are Domain types in SincFun.jl, DEQuadrature.jl, and DESincEig. These should be unified, preferably in SincFun.jl as it is a general concept and utility, then imported in the other two packages.