Add a macro wrapper to interface the postetior marginals algorithm

docs/src/library/public.md

@@ -50,7 +50,7 @@ LastStage
 ### Functions
 
 ```@docs
-compile_algo
+@posterior_marginals
 prod
 sum
 redu

docs/src/usage.md

@@ -13,10 +13,9 @@ Calculates the posterior marginal of each variable in the input graph. The
 input graph should be defined in the [UAI model file format](@ref).
 
 ```@example main
-algo = compile_algo("problems/asia/asia.uai")
-eval(algo)
+algo = @posterior_marginals("problems/asia/asia.uai")
 obsvars, obsvals = Int64[], Int64[]
-marginals = run_algo(obsvars, obsvals)
+marginals = algo(obsvars, obsvals)
 ```
 
 #### Example 2
@@ -27,12 +26,11 @@ format](@ref). The evidence variables and values should be given in the [UAI
 evidence file format](@ref).
 
 ```@example main
-algo = compile_algo(
+algo = @posterior_marginals(
          "problems/asia/asia.uai",
          uai_evid_filepath = "problems/asia/asia.uai.evid")
-eval(algo)
 obsvars, obsvals = JunctionTrees.read_uai_evid_file("problems/asia/asia.uai.evid")
-marginals = run_algo(obsvars, obsvals)
+marginals = algo(obsvars, obsvals)
 ```
 
 #### Example 3
@@ -42,13 +40,12 @@ junction tree (which is passed as an argument) is used. This junction tree
 should be defined in the [PACE graph format](@ref).
 
 ```@example main
-algo = compile_algo(
+algo = @posterior_marginals(
          "problems/asia/asia.uai",
          uai_evid_filepath = "problems/asia/asia.uai.evid",
          td_filepath = "problems/asia/asia.td")
-eval(algo)
 obsvars, obsvals = JunctionTrees.read_uai_evid_file("problems/asia/asia.uai.evid")
-marginals = run_algo(obsvars, obsvals)
+marginals = algo(obsvars, obsvals)
 ```
 
 #### Example 4
@@ -57,7 +54,7 @@ Returns the expression of the junction tree algorithm up to the backward pass
 stage.
 
 ```@example main
-backward_pass_expr = compile_algo( "problems/asia/asia.uai", last_stage = BackwardPass)
+backward_pass_expr = @posterior_marginals("problems/asia/asia.uai", last_stage = BackwardPass)
 ```
 
 The stages supported are:

src/JunctionTrees.jl

@@ -3,7 +3,7 @@ Main module for `JunctionTrees.jl` -- a Julia implementation of the junction tre
 
 One main function is exported from this module for public use:
 
-- [`compile_algo`](@ref). Compiles and returns an expression that computes the posterior marginals of the model given evidence using the junction tree algorithm.
+- [`@posterior_marginals`](@ref). Compiles and returns an expression that computes the posterior marginals of the model given evidence using the junction tree algorithm.
 
 # Exports
 
@@ -16,7 +16,7 @@ using Combinatorics: combinations
 using MacroTools: @capture, rmlines
 using MLStyle: @match
 
-export compile_algo, Factor, prod, sum, redu, norm, LastStage, ForwardPass,
+export @posterior_marginals, Factor, prod, sum, redu, norm, LastStage, ForwardPass,
   BackwardPass, JointMarginals, UnnormalizedMarginals, Marginals
 
 import Base:

src/graphical_transformation.jl

@@ -55,12 +55,6 @@ Construct a tree decomposition graph based on `td_filepath`.
 
 The `td_filepath` file format is defined in:
 https://pacechallenge.org/2017/treewidth/.
-
-# Example
-```
-td_filepath = "../problems/Promedus_26/Promedus_26.td"
-td = compile_algo(td_filepath)
-```
 """
 function construct_td_graph(td_filepath::AbstractString)
 

src/junction_tree_algorithm.jl

@@ -1,6 +1,6 @@
 """
 Enumerated type used to select up to which stage an expression of the junction
-tree algorithm should be returned after calling [`compile_algo`](@ref).
+tree algorithm should be returned after calling [`@posterior_marginals`](@ref).
 """
 @enum LastStage begin
   ForwardPass
@@ -9,11 +9,18 @@ tree algorithm should be returned after calling [`compile_algo`](@ref).
   UnnormalizedMarginals
   Marginals
 end
-@doc "When assigned to the keyword argument `last_stage` of [`compile_algo`](@ref), an expression up to and including the forward pass is returned." ForwardPass
-@doc "When assigned to the keyword argument `last_stage` of [`compile_algo`](@ref), an expression up to and including the backward pass is returned." BackwardPass
-@doc "When assigned to the keyword argument `last_stage` of [`compile_algo`](@ref), an expression that computes the cluster joint marginals is returned." JointMarginals
-@doc "When assigned to the keyword argument `last_stage` of [`compile_algo`](@ref), an expression that computes the joint marginals is returned." UnnormalizedMarginals
-@doc "When assigned to the keyword argument `last_stage` of [`compile_algo`](@ref), an expression that computes the posterior marginals is returned (default)." Marginals
+@doc "When assigned to the keyword argument `last_stage` of [`@posterior_marginals`](@ref), an expression up to and including the forward pass is returned." ForwardPass
+@doc "When assigned to the keyword argument `last_stage` of [`@posterior_marginals`](@ref), an expression up to and including the backward pass is returned." BackwardPass
+@doc "When assigned to the keyword argument `last_stage` of [`@posterior_marginals`](@ref), an expression that computes the cluster joint marginals is returned." JointMarginals
+@doc "When assigned to the keyword argument `last_stage` of [`@posterior_marginals`](@ref), an expression that computes the joint marginals is returned." UnnormalizedMarginals
+@doc "When assigned to the keyword argument `last_stage` of [`@posterior_marginals`](@ref), an expression that computes the posterior marginals is returned (default)." Marginals
+
+macro posterior_marginals(uai_filepath, kwargs...)
+  posterior_marginals(
+    Base.eval(__module__, uai_filepath);
+    map(kwarg -> Pair(first(kwarg.args), Base.eval(__module__, last(kwarg.args))), kwargs)...
+  )
+end
 
 """
 $(TYPEDSIGNATURES)
@@ -36,10 +43,9 @@ of all the variables in the model.
 ```
 package_root_dir = pathof(JunctionTrees) |> dirname |> dirname
 uai_filepath = joinpath(package_root_dir, "docs", "src", "problems", "paskin", "paskin.uai")
-algo = compile_algo(uai_filepath)
-eval(algo)
+algo = @posterior_marginals(uai_filepath)
 obsvars, obsvals = Int64[], Int64[]
-marginals = run_algo(obsvars, obsvals)
+marginals = algo(obsvars, obsvals)
 
 # output
 
@@ -56,12 +62,11 @@ marginals = run_algo(obsvars, obsvals)
 package_root_dir = pathof(JunctionTrees) |> dirname |> dirname
 uai_filepath = joinpath(package_root_dir, "docs", "src", "problems", "paskin", "paskin.uai")
 uai_evid_filepath = joinpath(package_root_dir, "docs", "src", "problems", "paskin", "paskin.uai.evid")
-algo = compile_algo(
+algo = @posterior_marginals(
          uai_filepath,
          uai_evid_filepath = uai_evid_filepath)
-eval(algo)
 obsvars, obsvals = JunctionTrees.read_uai_evid_file(uai_evid_filepath)
-marginals = run_algo(obsvars, obsvals)
+marginals = algo(obsvars, obsvals)
 
 # output
 
@@ -74,16 +79,17 @@ marginals = run_algo(obsvars, obsvals)
  Factor{Float64, 1}((6,), [0.6118571666785584, 0.3881428333214415])
 ```
 """
-function compile_algo(uai_filepath::AbstractString;
-                      uai_evid_filepath::AbstractString = "",
-                      td_filepath::AbstractString = "",
-                      apply_partial_evaluation::Bool = false,
-                      last_stage::LastStage = Marginals,
-                      smart_root_selection::Bool = true,
-                      factor_eltype::DataType = Float64,
-                      use_omeinsum::Bool = false,
-                      correct_fp_overflows::Bool = false,
-                     )
+function posterior_marginals(
+  uai_filepath::AbstractString;
+  uai_evid_filepath::AbstractString="",
+  td_filepath::AbstractString="",
+  apply_partial_evaluation::Bool=false,
+  last_stage::LastStage=Marginals,
+  smart_root_selection::Bool=true,
+  factor_eltype::DataType=Float64,
+  use_omeinsum::Bool=false,
+  correct_fp_overflows::Bool=false,
+)
 
   # Read PGM
   nvars, cards, _, factors = read_uai_file(uai_filepath, factor_eltype = factor_eltype)

src/omeinsum.jl

@@ -8,10 +8,9 @@ Speed up the sum-product in the algorithm using `OMEinsum`'s contraction routine
 ```
 package_root_dir = pathof(JunctionTrees) |> dirname |> dirname
 uai_filepath = joinpath(package_root_dir, "docs", "src", "problems", "paskin", "paskin.uai")
-algo = compile_algo(uai_filepath, use_omeinsum = true);
-eval(algo)
+algo = @posterior_marginals(uai_filepath, use_omeinsum = true);
 obsvars, obsvals = Int64[], Int64[]
-marginals = run_algo(obsvars, obsvals)
+marginals = algo(obsvars, obsvals)
 ```
 """
 function boost_algo(algo::Expr; optimizer=GreedyMethod())

test/junction_tree_algorithm.jl

@@ -22,13 +22,12 @@ using JunctionTrees
 
     @debug "    Test: Default (Min-fill heuristic)"
     @debug "      Compiling algo..."
-    algo = compile_algo(
-             uai_filepath;
+    algo = @posterior_marginals(
+             uai_filepath,
              uai_evid_filepath = uai_evid_filepath,
            )
-    eval(algo)
     @debug "      Running algo..."
-    marginals = run_algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
+    marginals = algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
     @test isapprox(marginals, reference_marginals, atol=0.03)
 
     # ------------------------------------------------------------------------------
@@ -37,15 +36,14 @@ using JunctionTrees
 
     @debug "    Test: Using an existing junction tree"
     @debug "      Compiling algo..."
-    algo = compile_algo(
-             uai_filepath;
+    algo = @posterior_marginals(
+             uai_filepath,
              uai_evid_filepath = uai_evid_filepath,
              td_filepath = td_filepath,
              correct_fp_overflows = true,
            )
-    eval(algo)
     @debug "      Running algo..."
-    marginals = run_algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
+    marginals = algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
     @test isapprox(marginals, reference_marginals, atol=0.03)
 
     # ------------------------------------------------------------------------------
@@ -54,15 +52,14 @@ using JunctionTrees
 
     @debug "    Test: Float32 factor values"
     @debug "      Compiling algo..."
-    algo = compile_algo(
-             uai_filepath;
+    algo = @posterior_marginals(
+             uai_filepath,
              uai_evid_filepath = uai_evid_filepath,
              td_filepath = td_filepath,
              factor_eltype = Float32,
            )
-    eval(algo)
     @debug "      Running algo..."
-    marginals = run_algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
+    marginals = algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
     @test isapprox(marginals, reference_marginals)
 
     # ------------------------------------------------------------------------------
@@ -71,16 +68,15 @@ using JunctionTrees
 
     @debug "    Test: Partial evaluation"
     @debug "      Compiling algo..."
-    algo = compile_algo(
-             uai_filepath;
+    algo = @posterior_marginals(
+             uai_filepath,
              uai_evid_filepath = uai_evid_filepath,
              td_filepath = td_filepath,
              apply_partial_evaluation = true,
              correct_fp_overflows = true,
            )
-    eval(algo)
     @debug "      Running algo..."
-    marginal_factors = run_algo(obsvars, obsvals)
+    marginal_factors = algo(obsvars, obsvals)
     # Filter the observed variables from the obtained solution
     marginal_factors_filtered = filter(x -> !(x.vars[1] in obsvars) , marginal_factors)
     marginals = map(y -> y.vals, marginal_factors_filtered)
@@ -94,15 +90,14 @@ using JunctionTrees
 
     @debug "    Test: OMEinsum"
     @debug "      Compiling algo..."
-    algo = compile_algo(
-             uai_filepath;
+    algo = @posterior_marginals(
+             uai_filepath,
              uai_evid_filepath = uai_evid_filepath,
              td_filepath = td_filepath,
              use_omeinsum = true,
            )
-    eval(algo)
     @debug "      Running algo..."
-    marginals = run_algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
+    marginals = algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
     @test isapprox(marginals, reference_marginals, atol=0.03)
 
   end

test/uai2014.jl

@@ -50,16 +50,15 @@ for benchmark in benchmarks
 
         println("    Test type: default")
         println("      Compiling algo...")
-        algo = compile_algo(
-                            uai_filepath;
+        algo = @posterior_marginals(
+                            uai_filepath,
                             uai_evid_filepath = uai_evid_filepath, 
                             td_filepath = td_filepath,
                             factor_eltype = Float64,
                             correct_fp_overflows = true,
                           )
-        eval(algo)
         println("      Running algo...")
-        marginals = run_algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
+        marginals = algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
         @test isapprox(marginals, reference_marginals, atol = 0.01)
 
         # ------------------------------------------------------------------------------
@@ -68,16 +67,15 @@ for benchmark in benchmarks
 
         # println("    Test type: OMEinsum backend")
         # println("      Compiling algo...")
-        # algo = compile_algo(
-        #                     uai_filepath;
+        # algo = @posterior_marginals(
+        #                     uai_filepath,
         #                     uai_evid_filepath = uai_evid_filepath, 
         #                     td_filepath = td_filepath,
         #                     factor_eltype = Float64,
         #                     use_omeinsum = true,
         #                   )
-        # eval(algo)
         # println("      Running algo...")
-        # marginals = run_algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
+        # marginals = algo(obsvars, obsvals) |> x -> map(y -> y.vals, x)
         # @test isapprox(marginals, reference_marginals, atol = 0.01)
 
       end