Merge pull request #317 from ReactiveBayes/dev-update-3.4.0

Fixes for new versions of GraphPPL and ReactiveMP

Project.toml

@@ -1,7 +1,7 @@
 name = "RxInfer"
 uuid = "86711068-29c9-4ff7-b620-ae75d7495b3d"
 authors = ["Bagaev Dmitry <d.v.bagaev@tue.nl> and contributors"]
-version = "3.3.1"
+version = "3.4.0"
 
 [deps]
 BayesBase = "b4ee3484-f114-42fe-b91c-797d54a0c67e"
@@ -28,13 +28,13 @@ Distributions = "0.25"
 DomainSets = "0.5.2, 0.6, 0.7"
 ExponentialFamily = "1.2"
 FastCholesky = "1.3.0"
-GraphPPL = "~4.2.0"
+GraphPPL = "~4.3.0"
 LinearAlgebra = "1.9"
 MacroTools = "0.5.6"
 Optim = "1.0.0"
 ProgressMeter = "1.0.0"
 Random = "1.9"
-ReactiveMP = "~4.1.0"
+ReactiveMP = "~4.2.0"
 Reexport = "1.2.0"
 Rocket = "1.8.0"
 TupleTools = "1.2.0"

codemeta.json

@@ -9,12 +9,12 @@
     "downloadUrl": "https://github.com/reactivebayes/RxInfer.jl/releases",
     "issueTracker": "https://github.com/reactivebayes/RxInfer.jl/issues",
     "name": "RxInfer.jl",
-    "version": "3.3.1",
+    "version": "3.4.0",
     "description": "Julia package for automated, scalable and efficient Bayesian inference on factor graphs with reactive message passing. ",
     "applicationCategory": "Statistics",
     "developmentStatus": "active",
     "readme": "https://reactivebayes.github.io/RxInfer.jl/stable/",
-    "softwareVersion": "3.3.1",
+    "softwareVersion": "3.4.0",
     "keywords": [
         "Bayesian inference",
         "message passing",

src/model/plugins/reactivemp_inference.jl

@@ -159,8 +159,12 @@ function activate_rmp_variable!(plugin::ReactiveMPInferencePlugin, model::Model,
         # By default it is `UnspecifiedFormConstraint` which means that the form of the resulting distribution is not specified in advance
         # and follows from the computation, but users may override it with other form constraints, e.g. `PointMassFormConstraint`, which
         # constraints the resulting distribution to be of a point mass form
-        messages_form_constraint = getextra(nodedata, GraphPPL.VariationalConstraintsMessagesFormConstraintKey, ReactiveMP.UnspecifiedFormConstraint())
-        marginal_form_constraint = getextra(nodedata, GraphPPL.VariationalConstraintsMarginalFormConstraintKey, ReactiveMP.UnspecifiedFormConstraint())
+        messages_form_constraint = ReactiveMP.preprocess_form_constraints(
+            plugin, model, getextra(nodedata, GraphPPL.VariationalConstraintsMessagesFormConstraintKey, ReactiveMP.UnspecifiedFormConstraint())
+        )
+        marginal_form_constraint = ReactiveMP.preprocess_form_constraints(
+            plugin, model, getextra(nodedata, GraphPPL.VariationalConstraintsMarginalFormConstraintKey, ReactiveMP.UnspecifiedFormConstraint())
+        )
         # Fetch "prod-constraint" for messages and marginals. The prod-constraint usually defines the constraints for a single product of messages
         # It can for example preserve a specific parametrization of distribution 
         messages_prod_constraint = getextra(nodedata, :messages_prod_constraint, ReactiveMP.default_prod_constraint(messages_form_constraint))
@@ -301,3 +305,11 @@ ReactiveMP.setmarginals!(collection::AbstractArray{GraphVariableRef}, marginal)
 
 ReactiveMP.setmessage!(ref::GraphVariableRef, marginal) = setmessage!(ref.variable, marginal)
 ReactiveMP.setmessages!(collection::AbstractArray{GraphVariableRef}, marginal) = ReactiveMP.setmessages!(map(ref -> ref.variable, collection), marginal)
+
+# Form constraint preprocessing 
+
+function ReactiveMP.preprocess_form_constraints(backend::ReactiveMPInferencePlugin, model::Model, constraints)
+    # It is a simple pass-through for now, but can be extended in the future to preprocess constraints that 
+    # are defined in other packages, e.g. in `Distributions` and to support constraints, such as `q(x) :: Normal`
+    return ReactiveMP.preprocess_form_constraints(constraints)
+end