feat: warn on ±Inf values in check_model observe statements

src/abstract_varinfo.jl

@@ -219,10 +219,12 @@ function setlogp!!(vi::AbstractVarInfo, logp::NamedTuple{names}) where {names}
 end
 
 function setlogp!!(vi::AbstractVarInfo, logp::Number)
-    return error("""
-                 `setlogp!!(vi::AbstractVarInfo, logp::Number)` is no longer supported. Use
-                 `setloglikelihood!!`, `setlogjac!!`, and/or `setlogprior!!` instead.
-                 """)
+    return error(
+        """
+        `setlogp!!(vi::AbstractVarInfo, logp::Number)` is no longer supported. Use
+        `setloglikelihood!!`, `setlogjac!!`, and/or `setlogprior!!` instead.
+        """
+    )
 end
 
 """

src/debug_utils.jl

@@ -14,8 +14,10 @@ export check_model, has_static_constraints
 
 An accumulator which checks calls at each tilde-statement for potential errors.
 
-Right now this accumulator only checks for `NaN` values on the left-hand side of observe
-statements, and partially `missing` values on the left-hand side of observe statements.
+Right now this accumulator checks for `NaN` and `±Inf` values on the left-hand 
+side of observe statements, and partially `missing` values on the left-hand side
+of observe statements.
+
 
 Other checks in `check_model` are accomplished via different accumulators.
 """
@@ -72,6 +74,15 @@ _has_nans(x::NamedTuple) = any(_has_nans, x)
 _has_nans(x::AbstractArray) = any(_has_nans, x)
 _has_nans(x) = isnan(x)
 _has_nans(::Missing) = false
+"""
+    _has_infs(x)
+
+Check if `x` is `Inf` or `-Inf`, or contains any such values.
+"""
+_has_infs(x::NamedTuple) = any(_has_infs, x)
+_has_infs(x::AbstractArray) = any(_has_infs, x)
+_has_infs(x) = isinf(x)
+_has_infs(::Missing) = false
 
 function DynamicPPL.accumulate_assume!!(
     acc::DebugAccumulator, val, tval, logjac, vn::VarName, right::Distribution, template
@@ -107,6 +118,16 @@ function DynamicPPL.accumulate_observe!!(
         @warn msg
         failed = true
     end
+    # Check for Inf values, but only warn if the logpdf at that value is -Inf
+    # (i.e., Inf is not in the support of the distribution)
+    if _has_infs(val) && isinf(logpdf(right, val))
+        msg =
+            "Encountered an infinite value on the left-hand side of an" *
+            " observe statement; this may indicate that your data" *
+            " contain Inf or -Inf values."
+        @warn msg
+        failed = true
+    end
     return DebugAccumulator(failed)
 end
 
@@ -130,6 +151,7 @@ needed.
 - Repeated usage of the same or overlapping VarNames
 
 - `NaN` on the left-hand side of observe statements
+- `±Inf` on the left-hand side of observe statements (when the value is not in the support of the distribution)
 
 - (if `fail_if_discrete` is set) Usage of discrete distributions
 

test/compiler.jl

@@ -372,12 +372,14 @@ end
             x .~ [Normal(), Normal()]
             return x
         end
-        expected_error = ArgumentError("""
-            As of v0.35, DynamicPPL does not allow arrays of distributions in `.~`. \
-            Please use `product_distribution` instead, or write a loop if necessary. \
-            See https://github.com/TuringLang/DynamicPPL.jl/releases/tag/v0.35.0 for more \
-            details.\
-            """)
+        expected_error = ArgumentError(
+            """
+As of v0.35, DynamicPPL does not allow arrays of distributions in `.~`. \
+Please use `product_distribution` instead, or write a loop if necessary. \
+See https://github.com/TuringLang/DynamicPPL.jl/releases/tag/v0.35.0 for more \
+details.\
+"""
+        )
         @test_throws expected_error (vector_dot_tilde()(); true)
     end
 

test/debug_utils.jl

@@ -122,6 +122,22 @@ end
         @test_throws ErrorException check_model(m; error_on_failure=true)
     end
 
+    @testset "Inf in data" begin
+        @model function demo_inf_in_data(x)
+            a ~ Normal()
+            for i in eachindex(x)
+                x[i] ~ Normal(a)
+            end
+        end
+        m = demo_inf_in_data([1.0, Inf])
+        @test_throws ErrorException check_model(m; error_on_failure=true)
+        m2 = demo_inf_in_data([1.0, -Inf])
+        @test_throws ErrorException check_model(m2; error_on_failure=true)
+        # Finite data should pass
+        m3 = demo_inf_in_data([1.0, 2.0])
+        @test check_model(m3; error_on_failure=true)
+    end
+
     @testset "incorrect use of condition" begin
         @testset "missing in multivariate" begin
             @model function demo_missing_in_multivariate(x)