KeywordCalls compatibility (#93)

> # Changes
> 
> **Dependencies**
> 
>     * now using KeywordCalls.jl
> 
>     * Early stages of trait interface for `IsPrimitive` and `IsRepresentative`, using SimpleTraits.jl (see `traits.jl`)
> 
>     * New Distributions.jl version (no real changes here from that)
> 
>     * New TransformVariables has breaking changes, see `transforms.jl`
> 
> 
> **Renaming**
> 
>     * `@measure` is too vague, changed to `@parameterized`
> 
>     * `ElementwiseProductMeasure` is a bit much, it's now `PointwiseProductMeasure`
> 
>     * `Likelihood` is now `LogLikelihood`, and it's callable as a function
> 
>     * `src/probability` is now `sr`

Author: cscherrer

Project.toml

@@ -1,7 +1,7 @@
 name = "MeasureTheory"
 uuid = "eadaa1a4-d27c-401d-8699-e962e1bbc33b"
 authors = ["Chad Scherrer <[email protected]> and contributors"]
-version = "0.6.2"
+version = "0.7.0"
 
 [deps]
 Accessors = "7d9f7c33-5ae7-4f3b-8dc6-eff91059b697"
@@ -13,6 +13,7 @@ DynamicIterators = "6c76993d-992e-5bf1-9e63-34920a5a5a38"
 InfiniteArrays = "4858937d-0d70-526a-a4dd-2d5cb5dd786c"
 IntervalSets = "8197267c-284f-5f27-9208-e0e47529a953"
 IterTools = "c8e1da08-722c-5040-9ed9-7db0dc04731e"
+KeywordCalls = "4d827475-d3e4-43d6-abe3-9688362ede9f"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 MLStyle = "d8e11817-5142-5d16-987a-aa16d5891078"
 MacroTools = "1914dd2f-81c6-5fcd-8719-6d5c9610ff09"
@@ -22,6 +23,7 @@ NamedTupleTools = "d9ec5142-1e00-5aa0-9d6a-321866360f50"
 NestedTuples = "a734d2a7-8d68-409b-9419-626914d4061d"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 RandomNumbers = "e6cf234a-135c-5ec9-84dd-332b85af5143"
+SimpleTraits = "699a6c99-e7fa-54fc-8d76-47d257e15c1d"
 SpecialFunctions = "276daf66-3868-5448-9aa4-cd146d93841b"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 StatsFuns = "4c63d2b9-4356-54db-8cca-17b64c39e42c"
@@ -33,23 +35,25 @@ Tullio = "bc48ee85-29a4-5162-ae0b-a64e1601d4bc"
 Accessors = "0.1"
 ArrayInterface = "2,3"
 ConcreteStructs = "0.2"
-Distributions = "0.23, 0.24"
+Distributions = "0.23, 0.24, 0.25"
 DistributionsAD = "0.6"
 DynamicIterators = "0.4.2"
 InfiniteArrays = "0.7, 0.8, 0.9, 0.10"
 IntervalSets = "0.5"
 IterTools = "1"
+KeywordCalls = "0.1.5"
 MLStyle = "0.4"
 MacroTools = "0.5"
 MappedArrays = "0.3, 0.4"
 MonteCarloMeasurements = "0.10"
 NamedTupleTools = "0.13"
 NestedTuples = "0.3"
 RandomNumbers = "1"
+SimpleTraits = "0.9"
 SpecialFunctions = "0.10, 1"
 StaticArrays = "0.12, 1"
 StatsFuns = "0.9"
-TransformVariables = "0.3"
+TransformVariables = "0.4"
 Tricks = "0.1"
 Tullio = "0.2"
 julia = "1.5"

docs/src/adding.md

@@ -6,10 +6,10 @@ This is by far the most common kind of measure, and is especially useful as a wa
 
 ### Declaring a Parameterized Measure
 
-To start, declare a `@measure`. For example, `Normal` is declared as
+To start, declare a `@parameterized`. For example, `Normal` is declared as
 
 ```julia
-@measure Normal(μ,σ) ≪ (1/sqrt2π) * Lebesgue(ℝ)
+@parameterized Normal(μ,σ) ≪ (1/sqrt2π) * Lebesgue(ℝ)
 ```
 
 [`ℝ` is typed as `\bbR <TAB>`]

src/MeasureTheory.jl

@@ -18,6 +18,7 @@ export AbstractMeasure
 using InfiniteArrays
 using ConcreteStructs
 using DynamicIterators
+using KeywordCalls
 
 const ∞ = InfiniteArrays.∞
 
@@ -40,42 +41,49 @@ Methods for computing density relative to other measures will be
 """
 function logdensity end
 
-include("paramorder.jl")
 include("exp.jl")
 include("domains.jl")
 include("utils.jl")
+include("traits.jl")
 include("absolutecontinuity.jl")
-include("basemeasures.jl")
 include("parameterized.jl")
 include("macros.jl")
+
+include("primitives/counting.jl")
+include("primitives/lebesgue.jl")
+include("primitives/dirac.jl")
+include("primitives/trivial.jl")
+
 include("combinators/weighted.jl")
 include("combinators/superpose.jl")
 include("combinators/product.jl")
 include("combinators/for.jl")
 include("combinators/power.jl")
 include("combinators/likelihood.jl")
-include("combinators/elementwise.jl")
+include("combinators/pointwise.jl")
 include("combinators/transforms.jl")
 include("combinators/spikemixture.jl")
 include("combinators/chain.jl")
+
 include("distributions.jl")
 include("rand.jl")
-include("probability/dirac.jl")
-include("probability/normal.jl")
-include("probability/studentt.jl")
-include("probability/cauchy.jl")
-include("probability/laplace.jl")
-include("probability/uniform.jl")
-include("probability/beta.jl")
-include("probability/gumbel.jl")
-include("probability/exponential.jl")
-include("probability/mvnormal.jl")
-include("probability/inverse-gamma.jl")
-include("probability/bernoulli.jl")
-include("probability/poisson.jl")
-include("probability/binomial.jl")
-include("probability/LKJL.jl")
-include("probability/negativebinomial.jl")
+
+include("parameterized/normal.jl")
+include("parameterized/studentt.jl")
+include("parameterized/cauchy.jl")
+include("parameterized/laplace.jl")
+include("parameterized/uniform.jl")
+include("parameterized/beta.jl")
+include("parameterized/gumbel.jl")
+include("parameterized/exponential.jl")
+include("parameterized/mvnormal.jl")
+include("parameterized/inverse-gamma.jl")
+include("parameterized/bernoulli.jl")
+include("parameterized/poisson.jl")
+include("parameterized/binomial.jl")
+include("parameterized/LKJL.jl")
+include("parameterized/negativebinomial.jl")
+
 include("density.jl")
 # include("pushforward.jl")
 include("kernel.jl")

src/absolutecontinuity.jl

@@ -53,26 +53,24 @@ export representative
 
 """
     representative(μ::AbstractMeasure) -> AbstractMeasure
-"""
-function representative(μ)
-    function f(μ)
-        # Check if we're done
-        isprimitive(μ) && return μ
-        
-        ν = basemeasure(μ)
 
-        # TODO: Make sure we don't leave the equivalence class
-        # Make sure not to leave the equivalence class
-        # (ν ≪ μ) || return μ
+We need to be able to compute `μ ≪ ν` for each `μ` and `ν`. To do this directly
+would require a huge number of methods (quadratic in the number of defined
+measures). 
 
-        return ν
-    end
+This function is a way around that. When defining a new measure `μ`, you should
+also find some equivalent measure `ρ` that's "as primitive as possible". 
 
-    fix(f, μ)
-end
+If possible, `ρ` should be a `PrimitiveMeasure`, or a `Product` of these. If
+not, it should be a  transform (`Pushforward` or `Pullback`) of a
+`PrimitiveMeasure` (or `Product` of these). 
+"""
+function representative(μ) end
 
-# TODO: ≪ needs more work
 function ≪(μ, ν)
     μ == ν && return true
     representative(μ) ≪ representative(ν) && return true
+    return false
 end
+
+@traitfn representative(μ::M) where {M; IsRepresentative{M}} = μ

src/basemeasures.jl

@@ -94,59 +94,3 @@ function dyniterate(fr::DynamicFor, state)
       u, state = ϕ
       fr.f(u), state
 end
-
-
-# using Soss
-
-# hmm = @model begin
-#     ε ~ Exponential() #  transition
-#     σ ~ Exponential() # Observation noise
-#     x ~ Chain(Normal()) do xj
-#         Normal(xj, ε)
-#     end
-
-#     y ~ For(x) do xj
-#         Normal(xj, σ)
-#     end
-# end
-
-# using Soss
-
-# mbind = @model μ,κ begin
-#     x ~ μ
-#     y ~ κ(x)
-#     return y
-# end
-
-# ⋅(μ,κ) = mbind(μ,κ)
-
-# d =  Cauchy() ⋅ (x -> Normal(μ=x)) ⋅ (x -> Normal(μ=x)) ⋅ (x -> Normal(μ=x))
-
-# rand(d)
-# t = xform(d)
-# t(randn(4))
-# simulate(d)
-
-# # julia> d =  Cauchy() ⋅ (x -> Normal(μ=x)) ⋅ (x -> Normal(μ=x)) ⋅ (x -> Normal(μ=x))
-# # ConditionalModel given
-# #     arguments    (:μ, :κ)
-# #     observations ()
-# # @model (μ, κ) begin
-# #         x ~ μ
-# #         y ~ κ(x)
-# #         return y
-# #     end
-
-
-
-# # julia> rand(d)
-# # -3.0414465047589037
-
-# # julia> t = xform(d)
-# # TransformVariables.TransformTuple{NamedTuple{(:x, :y), Tuple{TransformVariables.TransformTuple{NamedTuple{(:x, :y), Tuple{TransformVariables.TransformTuple{NamedTuple{(:x, :y), Tuple{TransformVariables.Identity, TransformVariables.Identity}}}, TransformVariables.Identity}}}, TransformVariables.Identity}}}((x = TransformVariables.TransformTuple{NamedTuple{(:x, :y), Tuple{TransformVariables.TransformTuple{NamedTuple{(:x, :y), Tuple{TransformVariables.Identity, TransformVariables.Identity}}}, TransformVariables.Identity}}}((x = TransformVariables.TransformTuple{NamedTuple{(:x, :y), Tuple{TransformVariables.Identity, TransformVariables.Identity}}}((x = asℝ, y = asℝ), 2), y = asℝ), 3), y = asℝ), 4)
-
-# # julia> t(randn(4))
-# # (x = (x = (x = -0.24259286698966315, y = 0.278190893626807), y = -1.361907586870645), y = 0.05914265096096323)
-
-# # julia> simulate(d)
-# # (value = 5.928939554009484, trace = (x = (value = 5.307006358072237, trace = (x = (value = 3.2023770380851797, trace = (x = 3.677550124255551, y = 3.2023770380851797)), y = 5.307006358072237)), y = 5.928939554009484))

src/combinators/chain.jl

@@ -152,7 +152,7 @@ function Base.rand(rng::AbstractRNG, T::Type, d::ForGenerator)
     Base.Generator(r ∘ d.data.f, d.data.iter)
 end
 
-function MeasureTheory.logdensity(d::ForGenerator, x)
+function logdensity(d::ForGenerator, x)
     sum((logdensity(dj, xj) for (dj, xj) in zip(d.data, x)))
 end
 

src/combinators/elementwise.jl

@@ -1,15 +1,15 @@
-export Likelihood
+export LogLikelihood
 
-@concrete terse struct Likelihood{T,X}
+@concrete terse struct LogLikelihood{T,X}
     x::X
 end
 
-Likelihood(T::Type, x::X) where {X} = Likelihood{T,X}(x)
+LogLikelihood(T::Type, x::X) where {X} = LogLikelihood{T,X}(x)
 
-Likelihood(μ::T, x::X) where {X, T<:AbstractMeasure} = Likelihood{T,X}(x)
+LogLikelihood(μ::T, x::X) where {X, T<:AbstractMeasure} = LogLikelihood{T,X}(x)
 
-logdensity(ℓ::Likelihood, p) = ℓ(p)
+logdensity(ℓ::LogLikelihood, p) = ℓ(p)
 
-(ℓ::Likelihood{T,X})(p) where {T,X} = logdensity(T(p), ℓ.x)
+(ℓ::LogLikelihood{T,X})(p) where {T,X} = logdensity(T(p), ℓ.x)
 
-(ℓ::Likelihood)(;kwargs...) = ℓ((;kwargs...))
+(ℓ::LogLikelihood)(;kwargs...) = ℓ((;kwargs...))