Replace instances of @submodel with to_submodel() (#751)

* Replace remaining instances of @SubModel

* Implement tilde_assume!! for Sampleable / pointwise contexts

* Fix typos in test

* Bump patch version

* Re-add some minimal tests for deprecated @SubModel

* Format

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

---------

Co-authored-by: github-actions[bot] <41898282+github-actions[bot]@users.noreply.github.com>

Project.toml

@@ -1,6 +1,6 @@
 name = "DynamicPPL"
 uuid = "366bfd00-2699-11ea-058f-f148b4cae6d8"
-version = "0.32.0"
+version = "0.32.1"
 
 [deps]
 ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"

src/contexts.jl

@@ -244,7 +244,7 @@ adds the `Prefix` to all parameters.
 This context is useful in nested models to ensure that the names of the parameters are
 unique.
 
-See also: [`@submodel`](@ref)
+See also: [`to_submodel`](@ref)
 """
 struct PrefixContext{Prefix,C} <: AbstractContext
     context::C

src/pointwise_logdensities.jl

@@ -146,6 +146,26 @@ function _pointwise_tilde_observe(
     end
 end
 
+# Note on submodels (penelopeysm)
+#
+# We don't need to overload tilde_observe!! for Sampleables (yet), because it
+# is currently not possible to evaluate a model with a Sampleable on the RHS
+# of an observe statement.
+#
+# Note that calling tilde_assume!! on a Sampleable does not necessarily imply
+# that there are no observe statements inside the Sampleable. There could well
+# be likelihood terms in there, which must be included in the returned logp.
+# See e.g. the `demo_dot_assume_observe_submodel` demo model.
+#
+# This is handled by passing the same context to rand_like!!, which figures out
+# which terms to include using the context, and also mutates the context and vi
+# appropriately. Thus, we don't need to check against _include_prior(context)
+# here.
+function tilde_assume!!(context::PointwiseLogdensityContext, right::Sampleable, vn, vi)
+    value, vi = DynamicPPL.rand_like!!(right, context, vi)
+    return value, vi
+end
+
 function tilde_assume!!(context::PointwiseLogdensityContext, right, vn, vi)
     !_include_prior(context) && return (tilde_assume!!(context.context, right, vn, vi))
     value, logp, vi = tilde_assume(context.context, right, vn, vi)

src/test_utils/models.jl

@@ -437,7 +437,8 @@ end
 
 @model function demo_assume_submodel_observe_index_literal()
     # Submodel prior
-    @submodel s, m = _prior_dot_assume()
+    priors ~ to_submodel(_prior_dot_assume(), false)
+    s, m = priors
     1.5 ~ Normal(m[1], sqrt(s[1]))
     2.0 ~ Normal(m[2], sqrt(s[2]))
 
@@ -462,7 +463,7 @@ function varnames(model::Model{typeof(demo_assume_submodel_observe_index_literal
     return [@varname(s[1]), @varname(s[2]), @varname(m[1]), @varname(m[2])]
 end
 
-@model function _likelihood_mltivariate_observe(s, m, x)
+@model function _likelihood_multivariate_observe(s, m, x)
     return x ~ MvNormal(m, Diagonal(s))
 end
 
@@ -475,7 +476,9 @@ end
     m .~ Normal.(0, sqrt.(s))
 
     # Submodel likelihood
-    @submodel _likelihood_mltivariate_observe(s, m, x)
+    # With to_submodel, we have to have a left-hand side variable to
+    # capture the result, so we just use a dummy variable
+    _ignore ~ to_submodel(_likelihood_multivariate_observe(s, m, x))
 
     return (; s=s, m=m, x=x, logp=getlogp(__varinfo__))
 end

test/compiler.jl

@@ -382,34 +382,13 @@ module Issue537 end
         @test demo2()() == 42
     end
 
-    @testset "@submodel is deprecated" begin
-        @model inner() = x ~ Normal()
-        @model outer() = @submodel x = inner()
-        @test_logs(
-            (
-                :warn,
-                "`@submodel model` and `@submodel prefix=... model` are deprecated; see `to_submodel` for the up-to-date syntax.",
-            ),
-            outer()()
-        )
-
-        @model outer_with_prefix() = @submodel prefix = "sub" x = inner()
-        @test_logs(
-            (
-                :warn,
-                "`@submodel model` and `@submodel prefix=... model` are deprecated; see `to_submodel` for the up-to-date syntax.",
-            ),
-            outer_with_prefix()()
-        )
-    end
-
-    @testset "submodel" begin
+    @testset "to_submodel" begin
         # No prefix, 1 level.
         @model function demo1(x)
             return x ~ Normal()
         end
         @model function demo2(x, y)
-            @submodel demo1(x)
+            _ignore ~ to_submodel(demo1(x), false)
             return y ~ Uniform()
         end
         # No observation.
@@ -441,7 +420,7 @@ module Issue537 end
 
         # Check values makes sense.
         @model function demo3(x, y)
-            @submodel demo1(x)
+            _ignore ~ to_submodel(demo1(x), false)
             return y ~ Normal(x)
         end
         m = demo3(1000.0, missing)
@@ -453,12 +432,10 @@ module Issue537 end
             x ~ Normal()
             return x
         end
-
         @model function demo_useval(x, y)
-            @submodel prefix = "sub1" x1 = demo_return(x)
-            @submodel prefix = "sub2" x2 = demo_return(y)
-
-            return z ~ Normal(x1 + x2 + 100, 1.0)
+            sub1 ~ to_submodel(demo_return(x))
+            sub2 ~ to_submodel(demo_return(y))
+            return z ~ Normal(sub1 + sub2 + 100, 1.0)
         end
         m = demo_useval(missing, missing)
         vi = VarInfo(m)
@@ -472,21 +449,18 @@ module Issue537 end
         @model function AR1(num_steps, α, μ, σ, ::Type{TV}=Vector{Float64}) where {TV}
             η ~ MvNormal(zeros(num_steps), I)
             δ = sqrt(1 - α^2)
-
             x = TV(undef, num_steps)
             x[1] = η[1]
             @inbounds for t in 2:num_steps
                 x[t] = @. α * x[t - 1] + δ * η[t]
             end
-
             return @. μ + σ * x
         end
 
         @model function demo(y)
             α ~ Uniform()
             μ ~ Normal()
             σ ~ truncated(Normal(), 0, Inf)
-
             num_steps = length(y[1])
             num_obs = length(y)
             @inbounds for i in 1:num_obs
@@ -613,14 +587,11 @@ module Issue537 end
         @model demo() = x ~ Normal()
         retval, svi = DynamicPPL.evaluate!!(demo(), SimpleVarInfo(), SamplingContext())
 
-        # Return-value when using `@submodel`
+        # Return-value when using `to_submodel`
         @model inner() = x ~ Normal()
-        # Without assignment.
-        @model outer() = @submodel inner()
-        @test outer()() isa Real
-
-        # With assignment.
-        @model outer() = @submodel x = inner()
+        @model function outer()
+            return _ignore ~ to_submodel(inner())
+        end
         @test outer()() isa Real
 
         # Edge-cases.
@@ -720,8 +691,7 @@ module Issue537 end
             return (; x, y)
         end
         @model function demo_tracked_submodel()
-            @submodel (x, y) = demo_tracked()
-            return (; x, y)
+            return vals ~ to_submodel(demo_tracked(), false)
         end
         for model in [demo_tracked(), demo_tracked_submodel()]
             # Make sure it's runnable and `y` is present in the return-value.

test/deprecated.jl

@@ -0,0 +1,57 @@
+@testset "deprecated" begin
+    @testset "@submodel" begin
+        @testset "is deprecated" begin
+            @model inner() = x ~ Normal()
+            @model outer() = @submodel x = inner()
+            @test_logs(
+                (
+                    :warn,
+                    "`@submodel model` and `@submodel prefix=... model` are deprecated; see `to_submodel` for the up-to-date syntax.",
+                ),
+                outer()()
+            )
+
+            @model outer_with_prefix() = @submodel prefix = "sub" x = inner()
+            @test_logs(
+                (
+                    :warn,
+                    "`@submodel model` and `@submodel prefix=... model` are deprecated; see `to_submodel` for the up-to-date syntax.",
+                ),
+                outer_with_prefix()()
+            )
+        end
+
+        @testset "prefixing still works correctly" begin
+            @model inner() = x ~ Normal()
+            @model function outer()
+                a = @submodel inner()
+                b = @submodel prefix = "sub" inner()
+                return a, b
+            end
+            @test outer()() isa Tuple{Float64,Float64}
+            vi = VarInfo(outer())
+            @test @varname(x) in keys(vi)
+            @test @varname(var"sub.x") in keys(vi)
+        end
+
+        @testset "logp is still accumulated properly" begin
+            @model inner_assume() = x ~ Normal()
+            @model inner_observe(x, y) = y ~ Normal(x)
+            @model function outer(b)
+                a = @submodel inner_assume()
+                @submodel inner_observe(a, b)
+            end
+            y_val = 1.0
+            model = outer(y_val)
+            @test model() == y_val
+
+            x_val = 1.5
+            vi = VarInfo(outer(y_val))
+            DynamicPPL.setindex!!(vi, x_val, @varname(x))
+            @test logprior(model, vi) ≈ logpdf(Normal(), x_val)
+            @test loglikelihood(model, vi) ≈ logpdf(Normal(x_val), y_val)
+            @test logjoint(model, vi) ≈
+                logpdf(Normal(), x_val) + logpdf(Normal(x_val), y_val)
+        end
+    end
+end

test/pointwise_logdensities.jl

@@ -13,7 +13,7 @@
         loglikelihood_true = DynamicPPL.TestUtils.loglikelihood_true(
             model, example_values...
         )
-        logp_true = logprior(model, vi)
+        logprior_true = logprior(model, vi)
 
         # Compute the pointwise loglikelihoods.
         lls = pointwise_loglikelihoods(model, vi)
@@ -30,18 +30,18 @@
         lps_prior = pointwise_prior_logdensities(model, vi)
         @test :x ∉ DynamicPPL.getsym.(keys(lps_prior))
         logp = sum(sum, values(lps_prior))
-        @test logp ≈ logp_true
+        @test logp ≈ logprior_true
 
         # Compute both likelihood and logdensity of prior
-        # using the default DefaultContex        
+        # using the default DefaultContext
         lps = pointwise_logdensities(model, vi)
         logp = sum(sum, values(lps))
-        @test logp ≈ (logp_true + loglikelihood_true)
+        @test logp ≈ (logprior_true + loglikelihood_true)
 
         # Test that modifications of Setup are picked up
         lps = pointwise_logdensities(model, vi, mod_ctx2)
         logp = sum(sum, values(lps))
-        @test logp ≈ (logp_true + loglikelihood_true) * 1.2 * 1.4
+        @test logp ≈ (logprior_true + loglikelihood_true) * 1.2 * 1.4
     end
 end
 

test/runtests.jl

@@ -59,6 +59,7 @@ include("test_util.jl")
         include("serialization.jl")
         include("pointwise_logdensities.jl")
         include("lkj.jl")
+        include("deprecated.jl")
     end
 
     if GROUP == "All" || GROUP == "Group2"