Added native visualizations to algebraicpetri (#23)

* Added native visualizations to algebraicpetri

* Added missing test

* Version bump

Project.toml

@@ -2,7 +2,7 @@ name = "AlgebraicPetri"
 uuid = "4f99eebe-17bf-4e98-b6a1-2c4f205a959b"
 license = "MIT"
 authors = ["Micah Halter <micah@mehalter.com>"]
-version = "0.5.0"
+version = "0.5.1"
 
 [deps]
 AutoHashEquals = "15f4f7f2-30c1-5605-9d31-71845cf9641f"
@@ -10,6 +10,7 @@ Catlab = "134e5e36-593f-5add-ad60-77f754baafbe"
 LabelledArrays = "2ee39098-c373-598a-b85f-a56591580800"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Petri = "4259d249-1051-49fa-8328-3f8ab9391c33"
+StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 
 [compat]
 AutoHashEquals = "^0.2.0"

examples/covid/_covid.jl

@@ -5,7 +5,6 @@
 using AlgebraicPetri
 using AlgebraicPetri.Epidemiology
 
-using Petri: Model, Graph
 using OrdinaryDiffEq
 using Plots
 
@@ -62,7 +61,7 @@ seird_city = to_hom_expr(FreeBiproductCategory, seird_city)
 
 display_wd(seird_city)
 # -
-Graph(Model(decoration(F(seird_city))))
+Graph(decoration(F(seird_city)))
 
 # create a multi-city SEIRD models
 
@@ -73,7 +72,7 @@ pc_seird_3 = F(seird_3)
 p_seird_3 = decoration(pc_seird_3)
 display_wd(seird_3)
 # -
-Graph(Model(p_seird_3))
+Graph(p_seird_3)
 
 # Define time frame and initial parameters
 

examples/covid/chime/chime-cset.jl

@@ -1,5 +1,4 @@
 using AlgebraicPetri
-using Petri: Model, Graph
 using OrdinaryDiffEq
 using Plots
 using Catlab.Meta
@@ -27,7 +26,7 @@ end
 
 sir_cset= LabelledReactionNet{Function, Float64}((:S=>990, :I=>10, :R=>0), (:inf, β)=>((:S, :I)=>(:I,:I)), (:rec, t->γ)=>(:I=>:R))
 
-Graph(Model(sir_cset))
+Graph(sir_cset)
 
 prob = ODEProblem(sir_cset, (17.0, 120.0))
 sol = OrdinaryDiffEq.solve(prob,Tsit5())

examples/covid/chime/chime.jl

@@ -1,6 +1,5 @@
 using AlgebraicPetri
 using AlgebraicPetri.Epidemiology
-using Petri: Model, Graph
 using OrdinaryDiffEq
 using LabelledArrays
 using Plots
@@ -15,7 +14,7 @@ sir = transmission ⋅ recovery
 p_sir = apex(F_epi(sir));
 display_wd(sir)
 #-
-Graph(Model(p_sir))
+Graph(p_sir)
 
 u0 = LVector(S=990, I=10, R=0);
 t_span = (17.0,120.0)

examples/covid/coexist/coexist.jl

@@ -250,7 +250,7 @@ display_wd(threeNCoexist)
 
 threeNCoexist_algpetri = apex(F_tcx(threeNCoexist))
 threeNCoexist_petri = Petri.Model(threeNCoexist_algpetri)
-Graph(threeNCoexist_petri)
+Graph(threeNCoexist_algpetri)
 
 # We can JSON to convert this Petri net into an
 # easily shareable format

examples/covid/epidemiology.jl

@@ -5,7 +5,6 @@
 using AlgebraicPetri
 using AlgebraicPetri.Epidemiology
 
-using Petri: Model, Graph
 using LabelledArrays
 using OrdinaryDiffEq
 using Plots
@@ -33,7 +32,7 @@ display_wd(sir)
 
 # Use Petri.jl to visualize the C-Set
 
-Graph(Model(cset_sir))
+Graph(cset_sir)
 
 # define initial states and transition rates, then
 # create, solve, and visualize ODE problem
@@ -65,7 +64,7 @@ p_seir = apex(F_epi(seir));
 
 display_wd(seir)
 #-
-Graph(Model(p_seir))
+Graph(p_seir)
 
 # define initial states and transition rates, then
 # create, solve, and visualize ODE problem
@@ -94,7 +93,7 @@ p_seird = apex(F_epi(seird));
 
 display_wd(seird)
 #-
-Graph(Model(p_seird))
+Graph(p_seird)
 
 # define initial states and transition rates, then
 # create, solve, and visualize ODE problem

examples/predation/lotka-volterra.jl

@@ -4,7 +4,6 @@
 
 using AlgebraicPetri
 
-using Petri: Model, Graph
 using OrdinaryDiffEq
 using Plots
 
@@ -21,13 +20,13 @@ display_wd(ex) = to_graphviz(ex, orientation=LeftToRight, labels=true);
 
 petriOb = codom(Open([1], PetriNet(1), [1]))
 birth_petri = Open([1], PetriNet(1, (1, (1,1))), [1]);
-Graph(Model(apex(birth_petri)))
+Graph(birth_petri)
 #-
 predation_petri = Open([1,2], PetriNet(2, ((1,2), (2,2))), [2]);
-Graph(Model(apex(predation_petri)))
+Graph(predation_petri)
 #-
 death_petri = Open([1], PetriNet(1, (1, ())), [1]);
-Graph(Model(apex(death_petri)))
+Graph(death_petri)
 
 # #### Step 2: Define a presentation of the free biproduct category
 # that encodes the domain specific information
@@ -52,7 +51,7 @@ lotka_volterra = (birth ⊗ id(wolves)) ⋅ predation ⋅ death
 lotka_petri = apex(F(lotka_volterra))
 display_wd(lotka_volterra)
 #-
-Graph(Model(lotka_petri))
+Graph(lotka_petri)
 
 # Generate appropriate vector fields, define parameters, and visualize solution
 
@@ -102,7 +101,7 @@ end
 display_wd(dual_lv)
 #-
 dual_lv_petri = apex(F(to_hom_expr(FreeBiproductCategory, dual_lv)))
-Graph(Model(dual_lv_petri))
+Graph(dual_lv_petri)
 
 # Generate a new solver, provide parameters, and analyze results
 

src/AlgebraicPetri.jl

@@ -20,7 +20,7 @@ using Catlab.Theories
 using Petri
 using LabelledArrays
 using LinearAlgebra: mul!
-import Petri: Model, vectorfield
+import Petri: Model, Graph, vectorfield 
 
 vectorify(n::Vector) = n
 vectorify(n::Tuple) = length(n) == 1 ? [n] : n
@@ -81,6 +81,9 @@ add_inputs!(p::AbstractPetriNet,n,t,s;kw...) = add_parts!(p,:I,n;it=t,is=s,kw...
 add_output!(p::AbstractPetriNet,t,s;kw...) = add_part!(p,:O;ot=t,os=s,kw...)
 add_outputs!(p::AbstractPetriNet,n,t,s;kw...) = add_parts!(p,:O,n;ot=t,os=s,kw...)
 
+sname(p::AbstractPetriNet,s) = (1:ns(p))[s]
+tname(p::AbstractPetriNet,t) = (1:nt(p))[t]
+
 # Note: although indexing makes this pretty fast, it is often faster to bulk-convert
 # the PetriNet net into a transition matrix, if you are working with all of the transitions
 inputs(p::AbstractPetriNet,t) = subpart(p,incident(p,t,:it),:is)
@@ -288,6 +291,8 @@ rates(p::AbstractLabelledReactionNet) = begin
   LVector(;[(tnames[t]=>rate(p, t)) for t in 1:nt(p)]...)
 end
 
+include("visualization.jl")
 include("Epidemiology.jl")
 
-end
+
+end

src/visualization.jl

@@ -0,0 +1,38 @@
+using Catlab.Graphics.Graphviz
+import Catlab.Graphics.Graphviz: Graph, Edge
+import Base.Iterators: flatten
+using StatsBase
+
+export Graph
+
+function edgify(δ, transition, reverse::Bool)
+  return [Edge(reverse ? ["T_$transition", "S_$k"] :
+                         ["S_$k", "T_$transition"],
+               Attributes(:label=>"$(δ[k])", :labelfontsize=>"6"))
+           for k in collect(keys(δ)) if δ[k] != 0]
+end
+
+function Graph(p::AbstractPetriNet)
+  statenodes = [Node(string("S_$(sname(p, s))"), Attributes(:shape=>"circle", :color=>"#6C9AC3")) for s in 1:ns(p)]
+  transnodes = [Node(string("T_$(tname(p, k))"), Attributes(:shape=>"square", :color=>"#E28F41")) for k in 1:nt(p)]
+
+  graph_attrs = Attributes(:rankdir=>"LR")
+  node_attrs  = Attributes(:shape=>"plain", :style=>"filled", :color=>"white")
+  edge_attrs  = Attributes(:splines=>"splines")
+
+  stmts = vcat(statenodes, transnodes)
+
+  edges = map(1:nt(p)) do k
+    k_name = tname(p, k)
+    vcat(edgify(countmap(map(x->sname(p,x), inputs(p, k))), k_name, false),
+         edgify(countmap(map(x->sname(p, x), outputs(p, k))), k_name, true))
+  end |> flatten |> collect
+
+  stmts = vcat(stmts, edges)
+  g = Graphviz.Digraph("G", stmts; graph_attrs=graph_attrs, node_attrs=node_attrs, edge_attrs=edge_attrs)
+  return g
+end
+
+function Graph(op::Union{OpenPetriNet, OpenLabelledPetriNetUntyped, OpenReactionNet, OpenLabelledReactionNetUntyped})
+    Graph(apex(op))
+end

test/types.jl

@@ -13,6 +13,13 @@ sir_tpetri= PetriNet(TransitionMatrices(sir_petri))
 @test Petri.Model(sir_petri) == Petri.Model(sir_rxn)
 @test Petri.Model(sir_lpetri) == Petri.Model(sir_lrxn)
 
+@test typeof(Graph(sir_petri)) == Graph
+@test typeof(Graph(sir_lpetri)) == Graph
+@test typeof(Graph(sir_rxn)) == Graph
+@test typeof(Graph(open_sir_rxn)) == Graph
+@test typeof(Graph(sir_lrxn)) == Graph
+@test typeof(Graph(open_sir_lrxn)) == Graph
+
 @test inputs(sir_petri, 1) == [1,2]
 @test outputs(sir_petri, 1) == [2,2]
 @test concentration(sir_rxn, 1) == 990
@@ -54,4 +61,4 @@ add_input!(sir_petri, 4, 4)
 add_output!(sir_petri, 4, 3)
 @test ni(sir_petri) == 5
 @test no(sir_petri) == 5
-@test sir_petri == PetriNet(4, ((1, 2), (2, 2)), (2, 3), (1, 4), (4, 3))
+@test sir_petri == PetriNet(4, ((1, 2), (2, 2)), (2, 3), (1, 4), (4, 3))