Function interface specs

Author: mauro3

src/DiffEqBase.jl

@@ -5,7 +5,8 @@ module DiffEqBase
   using RecipesBase, Parameters, RecursiveArrayTools
   using Ranges # For plot recipes with units
   import Base: length, size, getindex, endof, show, print,
-               next, start, done, eltype, eachindex
+               next, start, done, eltype, eachindex,
+               Callable
 
   "`DEProblem`: Defines differential equation problems via its internal functions"
   abstract DEProblem
@@ -46,6 +47,7 @@ module DiffEqBase
   include("solutions/solution_interface.jl")
   include("tableaus.jl")
   include("problems/ode_problems.jl")
+  include("problems/functions.jl")
 
   function solve end
 

src/problems/functions.jl

@@ -0,0 +1,96 @@
+# Functions for AbstractODEProblem and possibly other DEProblem types
+#
+# Implements passing in Jacobians (and possibly other functions) via
+# function overloading:
+#
+# - f(...) - objective function
+# - f(Val{:jac}, ...) - Jacobian of objective function
+# - the details of what `...` needs to be depends on the
+#   AbstractODEProblem subtype
+
+
+# Method_exists does not work:
+#
+# julia> f(::Val{:jac}, a, b, c) = 5
+# f (generic function with 1 method)
+#
+# julia> method_exists(f, Tuple{Val{:jac}, Vararg})
+# false
+#
+# Thus hand-code it:
+check_first_arg(f,T::Type) = check_first_arg(typeof(f),T)
+function check_first_arg{F}(::Type{F}, T::Type)
+    typ = Tuple{Any, T, Vararg}
+    for m in Base.MethodList(F.name.mt) # F.name.mt gets the method-table
+        m.sig<:typ && return true
+    end
+    return false
+end
+
+has_jac(f) = check_first_arg(f, Val{:jac})
+
+# Using SimpleTraits to dispatch on existence of jac-method
+using SimpleTraits
+@traitdef HasJac{F}
+@generated SimpleTraits.trait{F}(::Type{HasJac{F}}) = has_jac(F) ? :(HasJac{F}) : :(Not{HasJac{F}})
+# now a trait methods can dispatch on this:
+# @traitfn fn(g::::HasJac, ...) = ...
+# @traitfn fn(g::::(!HasJac), ...) = ...
+
+
+######################
+## Tests (TODO move to test/...)
+#####################
+f123(x,y) = 1
+f123(::Val{:jac}, x) = 2
+
+g123(x,y) = 1
+g123{T}(::Val{:jac}, x::T) = 2
+
+h123(x,y) = 1
+h123{T}(x::T) = 2
+
+immutable T123 end
+(::T123)(a) = 1
+(::T123)(::Val{:jac}, a) = 1
+
+immutable G123 end
+(::G123)(a) = 1
+(::G123)(b, a) = 1
+
+@assert has_jac(f123)
+@assert has_jac(g123)
+@assert !has_jac(h123)
+@assert has_jac(T123())
+@assert !has_jac(G123())
+# Arguably below could be ==false as T123 is a constructor
+# function. However, I'm not sure how to implement this.
+@assert has_jac(T123)
+@assert !has_jac(G123)
+
+@traitfn tfn(f::::HasJac) = true
+@traitfn tfn(f::::(!HasJac)) = false
+
+@assert tfn(f123)
+@assert tfn(g123)
+@assert !tfn(h123)
+@assert tfn(T123())
+@assert !tfn(G123())
+@assert !tfn(T123)  # NOTE this is inconsistent as has_jac(T123)==true
+@assert !tfn(G123)
+
+# ParameterizedFunction:
+type  LotkaVolterra <: ParameterizedFunction
+         a::Float64
+         b::Float64
+end
+(p::LotkaVolterra)(t,u,du) = begin
+         du[1] = p.a * u[1] - p.b * u[1]*u[2]
+         du[2] = -3 * u[2] + u[1]*u[2]
+end
+(p::LotkaVolterra)(::Val{:jac}, t, u, J) = 1
+
+lv = LotkaVolterra(0.0,0.0)
+
+@assert has_jac(lv)
+@assert tfn(lv)