ModelCompiler.TraceAllMessages activates Tracing.Trace for all variables (#145)

Tracing.Trace uses System.Diagnostics.Trace.
Added DifficultyAbility.fs to TestFSharp.

src/Compiler/Infer/ModelCompiler.cs

@@ -60,8 +60,7 @@ namespace Microsoft.ML.Probabilistic.Compiler
         }
 
         /// <summary>
-        /// If true, all messages after each iteration will be logged to csv files in a folder named with the model name.
-        /// Use MatlabWriter.WriteFromCsvFolder to convert these to a mat file.
+        /// If true, all variables will implicitly have a TraceMessages attribute.
         /// </summary>
         public bool TraceAllMessages { get; set; }
 
@@ -987,7 +986,10 @@ namespace Microsoft.ML.Probabilistic.Compiler
             if (OptimiseInferenceCode)
                 tc.AddTransform(new DeadCode2Transform(this));
             tc.AddTransform(new ParallelScheduleTransform());
-            if (TraceAllMessages)
+            // All messages after each iteration will be logged to csv files in a folder named with the model name.
+            // Use MatlabWriter.WriteFromCsvFolder to convert these to a mat file.
+            bool useTracingTransform = false;
+            if (TraceAllMessages && useTracingTransform)
                 tc.AddTransform(new TracingTransform());
             bool useArraySizeTracing = false;
             if (useArraySizeTracing)

src/Compiler/Infer/Transforms/MessageTransform.cs

@@ -838,8 +838,9 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
             }
 
             // Support for the 'TraceMessages' and 'ListenToMessages' attributes
-            if (mi.channelDecl != null && (context.InputAttributes.Has<TraceMessages>(mi.channelDecl) ||
-                                           context.InputAttributes.Has<ListenToMessages>(mi.channelDecl)))
+            if (compiler.TraceAllMessages || 
+                (mi.channelDecl != null && (context.InputAttributes.Has<TraceMessages>(mi.channelDecl) ||
+                                           context.InputAttributes.Has<ListenToMessages>(mi.channelDecl))))
             {
                 string msgText = msg.ToString();
                 // Look for TraceMessages attribute that matches this message
@@ -851,7 +852,7 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
                 if (listenTo != null && listenTo.Containing != null && !msgText.Contains(listenTo.Containing)) listenTo = null;
 
 
-                if ((listenTo != null) || (trace != null))
+                if ((listenTo != null) || (trace != null) || compiler.TraceAllMessages)
                 {
                     IExpression textExpr = DebuggingSupport.GetExpressionTextExpression(msg);
                     if (listenTo != null)

src/Compiler/Infer/Transforms/TracingTransform.cs

@@ -24,6 +24,7 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
         Dictionary<Set<IVariableDeclaration>, TableInfo> tableOfIndexVars = new Dictionary<Set<IVariableDeclaration>, TableInfo>();
         MethodInfo writeMethod, writeBytesMethod, writeLineMethod, flushMethod, disposeMethodInfo;
         IMethodDeclaration traceWriterMethod, disposeMethod;
+        public static bool UseToString = true;
 
         public override string Name
         {
@@ -66,7 +67,7 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
             var stmts = traceWriterMethod.Body.Statements;
             string folder = td.Name;
             stmts.Add(Builder.ExprStatement(Builder.StaticMethod(new Func<string, DirectoryInfo>(Directory.CreateDirectory), Builder.LiteralExpr(folder))));
-            IExpression pathExpr = Builder.BinaryExpr(BinaryOperator.Add, name, Builder.LiteralExpr(".csv"));
+            IExpression pathExpr = Builder.BinaryExpr(BinaryOperator.Add, name, Builder.LiteralExpr(UseToString ? ".tsv" : ".csv"));
             pathExpr = Builder.BinaryExpr(BinaryOperator.Add, Builder.LiteralExpr(folder + "/"), pathExpr);
             var writerDecl = Builder.VarDecl("writer", typeof(StreamWriter));
             var ctorExpr = Builder.NewObject(typeof(StreamWriter), pathExpr);
@@ -116,12 +117,13 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
             StringBuilder header = new StringBuilder();
             header.Append("iteration");
             output2.Add(GetWriteStatement(writer, Builder.VarRefExpr(iterationVar)));
-            var delimiter = GetWriteStatement(writer, Builder.LiteralExpr(","));
+            string delimiter = UseToString ? "\t" : ",";
+            var writeDelimiter = GetWriteStatement(writer, Builder.LiteralExpr(delimiter));
             foreach (var indexVar in table.indexVars)
             {
-                header.Append(",");
+                header.Append(delimiter);
                 header.Append(indexVar.Name);
-                output2.Add(delimiter);
+                output2.Add(writeDelimiter);
                 output2.Add(GetWriteStatement(writer, Builder.VarRefExpr(indexVar)));
             }
             foreach (var messageBaseExpr in table.messageExprs)
@@ -141,9 +143,9 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
                     conditions.Push(Builder.BinaryExpr(BinaryOperator.IdentityEquality, messageExpr, Builder.LiteralExpr(null)));
                 if (messageExpr.GetExpressionType().IsPrimitive)
                 {
-                    header.Append(",");
+                    header.Append(delimiter);
                     header.Append(varInfo.Name);
-                    output2.Add(delimiter);
+                    output2.Add(writeDelimiter);
                     output2.Add(GetWriteStatement(writer, AddConditions(messageExpr, conditions)));
                 }
                 else
@@ -151,9 +153,9 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
                     Dictionary<string, IExpression> dict = GetProperties(messageExpr);
                     foreach (var entry in dict)
                     {
-                        header.Append(",");
+                        header.Append(delimiter);
                         header.Append(varInfo.Name + entry.Key);
-                        output2.Add(delimiter);
+                        output2.Add(writeDelimiter);
                         output2.Add(GetWriteStatement(writer, AddConditions(entry.Value, conditions)));
                     }
                 }
@@ -170,6 +172,13 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
         {
             Dictionary<string, IExpression> dict = new Dictionary<string, IExpression>();
             Type type = expr.GetExpressionType();
+            if (UseToString)
+            {
+                var toStringMethod = type.GetMethod("ToString", new Type[0]);
+                dict["ToString"] = Builder.Method(expr, toStringMethod);
+            }
+            else
+            {
                 Type[] faces = type.GetInterfaces();
                 bool hasGetMean = false;
                 bool hasGetVariance = false;
@@ -190,6 +199,7 @@ namespace Microsoft.ML.Probabilistic.Compiler.Transforms
                     var varianceMethod = type.GetMethod("GetVariance", new Type[0]);
                     dict["Variance"] = Builder.Method(expr, varianceMethod);
                 }
+            }
             return dict;
         }
 

src/Runtime/Factors/Tracing.cs

@@ -20,7 +20,7 @@ namespace Microsoft.ML.Probabilistic.Factors
         /// <returns><paramref name="input"/></returns>
         public static T Trace<T>([IsReturned] T input, string text)
         {
-            Debug.WriteLine(StringUtil.JoinColumns(text, ": ", input));
+            System.Diagnostics.Trace.WriteLine(StringUtil.JoinColumns(text, ": ", input));
             return input;
         }
 

test/TestFSharp/DifficultyAbility.fs

@@ -0,0 +1,104 @@
+namespace DifficultyAbilityExample
+
+open System.Collections.Generic
+open Microsoft.ML.Probabilistic
+open Microsoft.ML.Probabilistic.FSharp
+open Microsoft.ML.Probabilistic.Models
+open Microsoft.ML.Probabilistic.Utilities
+open Microsoft.ML.Probabilistic.Distributions
+open Microsoft.ML.Probabilistic.Math
+
+module DifficultyAbility =
+    let main() = 
+        Rand.Restart(0);
+        let nQuestions = 100
+        let nSubjects = 40
+        let nChoices = 4
+        let abilityPrior = Gaussian(0.0, 1.0)
+        let difficultyPrior = Gaussian(0.0, 1.0)
+        let discriminationPrior = Gamma.FromMeanAndVariance(1.0, 0.01)
+
+        let Sample(nSubjects:int,nQuestions:int,nChoices:int,abilityPrior:Gaussian,difficultyPrior:Gaussian,discriminationPrior:Gamma)=
+          let ability= Util.ArrayInit( nSubjects, (fun _-> abilityPrior.Sample()))
+          let difficulty = Util.ArrayInit(nQuestions, (fun _ -> difficultyPrior.Sample()))
+          let discrimination = Util.ArrayInit(nQuestions, (fun _ -> discriminationPrior.Sample()))
+          let trueAnswer = Util.ArrayInit(nQuestions, (fun _ -> Rand.Int(nChoices)))
+          let response:int[][] = Array.zeroCreate nSubjects 
+          for s in 0..(nSubjects-1) do
+            response.[s] <- Array.zeroCreate nQuestions
+            for q in 0..(nQuestions-1) do
+              let advantage = ability.[s] - difficulty.[q]
+              let noise = Gaussian.Sample(0.0, discrimination.[q])
+              let correct = (advantage > noise)
+              if (correct) then
+                  response.[s].[q] <- trueAnswer.[q]
+              else
+                  response.[s].[q] <- Rand.Int(nChoices)
+          (response, ability,difficulty,discrimination,trueAnswer)
+
+
+        let data,trueAbility,trueDifficulty,trueDiscrimination,trueTrueAnswer = Sample(nSubjects,nQuestions,nChoices,abilityPrior,difficultyPrior,discriminationPrior)
+
+        let question = Range(nQuestions).Named("question")
+        let subject = Range(nSubjects).Named("subject")
+        let choice = Range(nChoices).Named("choice")
+
+        //let response = Variable.Array(Variable.Array<int>(question), subject).Named("response")
+        let response = Variable.Array<VariableArray<int>, int [][]>(Variable.Array<int>(question), subject).Named("response")
+
+        response.ObservedValue <- data
+
+        let ability = Variable.Array<double>(subject).Named("ability")
+        Variable.ForeachBlock subject ( fun s ->  ability.[s] <- Variable.Random(abilityPrior) )
+        let difficulty = Variable.Array<double>(question).Named("difficulty")
+        Variable.ForeachBlock question ( fun q ->  difficulty.[q] <- Variable.Random(difficultyPrior) )
+        let discrimination = Variable.Array<double>(question).Named("discrimination")
+        Variable.ForeachBlock question ( fun q ->  discrimination.[q] <- Variable.Random(discriminationPrior) )
+        let trueAnswer = Variable.Array<int>(question).Named("trueAnswer")
+        Variable.ForeachBlock question ( fun q ->  trueAnswer.[q] <- Variable.DiscreteUniform(choice) )
+
+        Variable.ForeachBlock subject (fun s -> 
+          Variable.ForeachBlock question (fun q -> 
+            let advantage = (ability.[s] - difficulty.[q]).Named("advantage")
+            let advantageNoisy = Variable.GaussianFromMeanAndPrecision(advantage, discrimination.[q]).Named("advantageNoisy")
+            let correct = (advantageNoisy >> 0.0).Named("correct")
+            Variable.IfBlock correct (fun _->response.[s].[q] <- trueAnswer.[q]) (fun _->response.[s].[q] <- Variable.DiscreteUniform(choice))
+            ()
+          )
+        )
+
+        let engine = InferenceEngine()
+        engine.NumberOfIterations <- 5
+        subject.AddAttribute(Models.Attributes.Sequential())
+        question.AddAttribute(Models.Attributes.Sequential())
+        let doMajorityVoting = false; // set this to 'true' to do majority voting
+        if doMajorityVoting then
+            ability.ObservedValue <- Util.ArrayInit(nSubjects, (fun i -> 0.0))
+            difficulty.ObservedValue <- Util.ArrayInit(nQuestions, (fun i -> 0.0))
+            discrimination.ObservedValue <- Util.ArrayInit(nQuestions, (fun i -> 0.0))
+
+        let trueAnswerPosterior = engine.Infer<IReadOnlyList<Discrete>>(trueAnswer)
+
+        let mutable numCorrect = 0
+        for q in 0..(nQuestions-1) do
+            let bestGuess = trueAnswerPosterior.[q].GetMode()
+            if (bestGuess = trueTrueAnswer.[q]) then
+              numCorrect<-numCorrect+1
+
+        let pctCorrect:float = 100.0 * (float numCorrect) / (float nQuestions)
+        printfn "%f TrueAnswers correct" pctCorrect
+
+        let difficultyPosterior = engine.Infer<IReadOnlyList<Gaussian>>(difficulty)
+
+
+        for q in 0..(System.Math.Min(nQuestions, 4)-1) do
+          printfn "difficulty[%i] = %A (sampled from %f)"  q  difficultyPosterior.[q] trueDifficulty.[q]
+
+        let discriminationPosterior = engine.Infer<IReadOnlyList<Gamma>>(discrimination)
+        for q in 0..(System.Math.Min(nQuestions, 4)-1) do
+          printfn "discrimination[%i] = %A (sampled from %f)"  q  discriminationPosterior.[q] trueDiscrimination.[q]
+
+        let abilityPosterior = engine.Infer<IReadOnlyList<Gaussian>>(ability)
+        for s in 0..(System.Math.Min(nQuestions, 4)-1) do
+          printfn "ability[%i] = %A (sampled from %f)"  s  abilityPosterior.[s] trueAbility.[s]
+

test/TestFSharp/Program.fs

@@ -4,6 +4,10 @@
 
 #light
 open System
+open System.Diagnostics
+
+let coreAssemblyInfo = FileVersionInfo.GetVersionInfo(typeof<Object>.Assembly.Location)
+printfn "%s .NET version %s mscorlib %s" (if Environment.Is64BitProcess then "64-bit" else "32-bit") (Environment.Version.ToString ()) coreAssemblyInfo.ProductVersion
 
 //main Smoke Test .............................................
 
@@ -13,5 +17,6 @@ let _ = GaussianRangesTutorial.ranges.rangesTestFunc()
 let _ = ClinicalTrialTutorial.clinical.clinicalTestFunc()
 let _ = BayesPointTutorial.bayes.bayesTestFunc()
 let _ = MixtureGaussiansTutorial.mixture.mixtureTestFunc()
+let _ = DifficultyAbilityExample.DifficultyAbility.main()
 
 Console.ReadLine() |> ignore

test/TestFSharp/TestFSharp.fsproj

@@ -7,6 +7,9 @@
     <PlatformTarget>AnyCPU</PlatformTarget>
     <Configurations>Debug;Release;DebugFull;DebugCore;ReleaseFull;ReleaseCore</Configurations>
   </PropertyGroup>
+  <PropertyGroup Condition="'$(Configuration)|$(Platform)'=='DebugFull|AnyCPU'">
+    <Prefer32Bit>false</Prefer32Bit>
+  </PropertyGroup>
   <Choose>
     <When Condition="'$(Configuration)'=='DebugFull' OR '$(Configuration)'=='ReleaseFull'">
       <PropertyGroup>
@@ -35,6 +38,7 @@
     <ProjectReference Include="..\..\src\FSharpWrapper\FSharpWrapper.fsproj" />
   </ItemGroup>
   <ItemGroup>
+    <Compile Include="DifficultyAbility.fs" />
     <Compile Include="..\..\src\Shared\SharedAssemblyFileVersion.fs">
       <Link>SharedAssemblyFileVersion.fs</Link>
     </Compile>

test/Tests/Core/MatlabSerializationTests.cs

@@ -0,0 +1,172 @@
+using System;
+using System.Collections.Generic;
+using System.IO;
+using Microsoft.ML.Probabilistic.Math;
+using Microsoft.ML.Probabilistic.Serialization;
+using Xunit;
+
+namespace Microsoft.ML.Probabilistic.Tests.Core
+{
+    using Assert = Xunit.Assert;
+
+    public class MatlabSerializationTests
+    {
+        [Fact]
+        //[DeploymentItem(@"Data\IRT2PL_10_250.mat", "Data")]
+        public void MatlabReaderTest2()
+        {
+            Dictionary<string, object> dict = MatlabReader.Read(Path.Combine(
+#if NETCORE
+                Path.GetDirectoryName(typeof(PsychTests).Assembly.Location), // work dir is not the one with Microsoft.ML.Probabilistic.Tests.dll on netcore and neither is .Location on netfull
+#endif
+                "Data", "IRT2PL_10_250.mat"));
+            Assert.Equal(5, dict.Count);
+            Matrix m = (Matrix)dict["Y"];
+            Assert.True(m.Rows == 250);
+            Assert.True(m.Cols == 10);
+            Assert.True(m[0, 1] == 0.0);
+            Assert.True(m[1, 0] == 1.0);
+            m = (Matrix)dict["difficulty"];
+            Assert.True(m.Rows == 10);
+            Assert.True(m.Cols == 1);
+            Assert.True(MMath.AbsDiff(m[1], 0.7773) < 2e-4);
+        }
+
+        [Fact]
+        ////[DeploymentItem(@"Data\test.mat", "Data")]
+        public void MatlabReaderTest()
+        {
+            MatlabReaderTester(Path.Combine(
+#if NETCORE
+                Path.GetDirectoryName(typeof(PsychTests).Assembly.Location), // work dir is not the one with Microsoft.ML.Probabilistic.Tests.dll on netcore and neither is .Location on netfull
+#endif
+                "Data", "test.mat"));
+        }
+
+        private void MatlabReaderTester(string fileName)
+        {
+            Dictionary<string, object> dict = MatlabReader.Read(fileName);
+            Assert.Equal(12, dict.Count);
+            Matrix aScalar = (Matrix)dict["aScalar"];
+            Assert.Equal(1, aScalar.Rows);
+            Assert.Equal(1, aScalar.Cols);
+            Assert.Equal(5.0, aScalar[0, 0]);
+            Assert.Equal("string", (string)dict["aString"]);
+            MatlabReader.ComplexMatrix aComplexScalar = (MatlabReader.ComplexMatrix)dict["aComplexScalar"];
+            Assert.Equal(5.0, aComplexScalar.Real[0, 0]);
+            Assert.Equal(3.0, aComplexScalar.Imaginary[0, 0]);
+            MatlabReader.ComplexMatrix aComplexVector = (MatlabReader.ComplexMatrix)dict["aComplexVector"];
+            Assert.Equal(1.0, aComplexVector.Real[0, 0]);
+            Assert.Equal(2.0, aComplexVector.Imaginary[0, 0]);
+            Assert.Equal(3.0, aComplexVector.Real[0, 1]);
+            Assert.Equal(4.0, aComplexVector.Imaginary[0, 1]);
+            var aStruct = (Dictionary<string, object>)dict["aStruct"];
+            Assert.Equal(2, aStruct.Count);
+            Assert.Equal(1.0, ((Matrix)aStruct["field1"])[0]);
+            Assert.Equal("two", (string)aStruct["field2"]);
+            object[,] aCell = (object[,])dict["aCell"];
+            Assert.Equal(1.0, ((Matrix)aCell[0, 0])[0]);
+            int[] intArray = (int[])dict["intArray"];
+            Assert.Equal(1, intArray[0]);
+            int[] uintArray = (int[])dict["uintArray"];
+            Assert.Equal(1, uintArray[0]);
+            bool[] aLogical = (bool[])dict["aLogical"];
+            Assert.True(aLogical[0]);
+            Assert.True(aLogical[1]);
+            Assert.False(aLogical[2]);
+            object[,,] aCell3D = (object[,,])dict["aCell3D"];
+            Assert.Null(aCell3D[0, 0, 0]);
+            Assert.Equal(7.0, ((Matrix)aCell3D[0, 0, 1])[0, 0]);
+            Assert.Equal(6.0, ((Matrix)aCell3D[0, 1, 0])[0, 0]);
+            double[,,,] array4D = (double[,,,])dict["array4D"];
+            Assert.Equal(4.0, array4D[0, 0, 1, 0]);
+            Assert.Equal(5.0, array4D[0, 0, 0, 1]);
+            long[] aLong = (long[])dict["aLong"];
+            Assert.Equal(1234567890123456789L, aLong[0]);
+        }
+
+        [Fact]
+        //[DeploymentItem(@"Data\test.mat", "Data")]
+        public void MatlabWriterTest()
+        {
+            Dictionary<string, object> dict = MatlabReader.Read(Path.Combine(
+#if NETCORE
+                Path.GetDirectoryName(typeof(PsychTests).Assembly.Location), // work dir is not the one with Microsoft.ML.Probabilistic.Tests.dll on netcore and neither is .Location on netfull
+#endif
+                "Data", "test.mat"));
+            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriterTest{Environment.CurrentManagedThreadId}.mat";
+            using (MatlabWriter writer = new MatlabWriter(fileName))
+            {
+                foreach (var entry in dict)
+                {
+                    writer.Write(entry.Key, entry.Value);
+                }
+            }
+            MatlabReaderTester(fileName);
+        }
+
+        [Fact]
+        public void MatlabWriteStringDictionaryTest()
+        {
+            Dictionary<string, string> dictString = new Dictionary<string, string>();
+            dictString["a"] = "a";
+            dictString["b"] = "b";
+            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriteStringDictionaryTest{Environment.CurrentManagedThreadId}.mat";
+            using (MatlabWriter writer = new MatlabWriter(fileName))
+            {
+                writer.Write("dictString", dictString);
+            }
+            Dictionary<string, object> vars = MatlabReader.Read(fileName);
+            Dictionary<string, object> dict = (Dictionary<string, object>)vars["dictString"];
+            foreach (var entry in dictString)
+            {
+                Assert.Equal(dictString[entry.Key], dict[entry.Key]);
+            }
+        }
+
+        [Fact]
+        public void MatlabWriteStringListTest()
+        {
+            List<string> strings = new List<string>();
+            strings.Add("a");
+            strings.Add("b");
+            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriteStringListTest{Environment.CurrentManagedThreadId}.mat";
+            using (MatlabWriter writer = new MatlabWriter(fileName))
+            {
+                writer.Write("strings", strings);
+            }
+            Dictionary<string, object> vars = MatlabReader.Read(fileName);
+            string[] array = (string[])vars["strings"];
+            for (int i = 0; i < array.Length; i++)
+            {
+                Assert.Equal(strings[i], array[i]);
+            }
+        }
+
+        [Fact]
+        public void MatlabWriteEmptyArrayTest()
+        {
+            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriteEmptyArrayTest{Environment.CurrentManagedThreadId}.mat";
+            using (MatlabWriter writer = new MatlabWriter(fileName))
+            {
+                writer.Write("ints", new int[0]);
+            }
+            Dictionary<string, object> vars = MatlabReader.Read(fileName);
+            int[] ints = (int[])vars["ints"];
+            Assert.Empty(ints);
+        }
+
+        [Fact]
+        public void MatlabWriteNumericNameTest()
+        {
+            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriteNumericNameTest{Environment.CurrentManagedThreadId}.mat";
+            using (MatlabWriter writer = new MatlabWriter(fileName))
+            {
+                writer.Write("24", new int[0]);
+            }
+            Dictionary<string, object> vars = MatlabReader.Read(fileName);
+            int[] ints = (int[])vars["24"];
+            Assert.Empty(ints);
+        }
+    }
+}

test/Tests/EpTests.cs

@@ -454,13 +454,14 @@ namespace Microsoft.ML.Probabilistic.Tests
             // generate data from the model
             var hPrior = new Gaussian(hMean, hVariance);
             var hSample = Util.ArrayInit(n, i => hPrior.Sample());
+            // When xMultiplier != 1, we have model mismatch so we want the learned xPrecision to decrease.
             double xMultiplier = 5;
             var xData = Util.ArrayInit(n, i => Gaussian.Sample(xMultiplier * hSample[i], xPrecisionTrue));
             var yData = Util.ArrayInit(n, i => Gaussian.Sample(hSample[i], yPrecisionTrue));
             x.ObservedValue = xData;
             y.ObservedValue = yData;
 
-            // N(x; ah, vx) N(h; mh, vh) = N(h; mh + k*(x - a*mh), (1-ka)vh)
+            // N(x; a*h, vx) N(h; mh, vh) = N(h; mh + k*(x - a*mh), (1-ka)vh)
             // where k = vh*a/(a^2*vh + vx)
             // if x = a*x' then k(x - a*mh) = a*k(x' - mh)
             // a*k = vh/(vh + vx/a^2)

test/Tests/InferTests.cs

@@ -209,6 +209,18 @@ namespace Microsoft.ML.Probabilistic.Tests
             //(new ModelTests()).CoinRunLengths();
         }
 
+        [Fact]
+        public void TraceAllMessagesTest()
+        {
+            Variable<double> x = Variable.GaussianFromMeanAndPrecision(0, 1);
+            Variable.ConstrainPositive(x);
+            Variable.ConstrainPositive(x);
+
+            InferenceEngine engine = new InferenceEngine();
+            engine.Compiler.TraceAllMessages = true;
+            engine.Infer(x);
+        }
+
         [Fact]
         public void MarginalWrongDistributionError()
         {

test/Tests/PsychTests.cs

@@ -20,10 +20,6 @@ namespace Microsoft.ML.Probabilistic.Tests
     
     public class PsychTests
     {
-#if SUPPRESS_UNREACHABLE_CODE_WARNINGS
-#pragma warning disable 162
-#endif
-
         internal void LogisticIrtTest()
         {
             Variable<int> numStudents = Variable.New<int>().Named("numStudents");
@@ -40,7 +36,8 @@ namespace Microsoft.ML.Probabilistic.Tests
             response[student, question] = Variable.BernoulliFromLogOdds(((ability[student] - difficulty[question]).Named("minus")*discrimination[question]).Named("product"));
             bool[,] data;
             double[] discriminationTrue = new double[0];
-            if (false)
+            bool useDummyData = false;
+            if (useDummyData)
             {
                 data = new bool[4,2];
                 for (int i = 0; i < data.GetLength(0); i++)
@@ -70,10 +67,6 @@ namespace Microsoft.ML.Probabilistic.Tests
             Console.WriteLine(StringUtil.JoinColumns(engine.Infer(discrimination), " should be ", StringUtil.ToString(discriminationTrue)));
         }
 
-#if SUPPRESS_UNREACHABLE_CODE_WARNINGS
-#pragma warning restore 162
-#endif
-
         public static bool[,] ConvertToBool(double[,] array)
         {
             int rows = array.GetLength(0);
@@ -89,168 +82,6 @@ namespace Microsoft.ML.Probabilistic.Tests
             return result;
         }
 
-        [Fact]
-        //[DeploymentItem(@"Data\IRT2PL_10_250.mat", "Data")]
-        public void MatlabReaderTest2()
-        {
-            Dictionary<string, object> dict = MatlabReader.Read(Path.Combine(
-#if NETCORE
-                Path.GetDirectoryName(typeof(PsychTests).Assembly.Location), // work dir is not the one with Microsoft.ML.Probabilistic.Tests.dll on netcore and neither is .Location on netfull
-#endif
-                "Data", "IRT2PL_10_250.mat"));
-            Assert.Equal(5, dict.Count);
-            Matrix m = (Matrix) dict["Y"];
-            Assert.True(m.Rows == 250);
-            Assert.True(m.Cols == 10);
-            Assert.True(m[0, 1] == 0.0);
-            Assert.True(m[1, 0] == 1.0);
-            m = (Matrix) dict["difficulty"];
-            Assert.True(m.Rows == 10);
-            Assert.True(m.Cols == 1);
-            Assert.True(MMath.AbsDiff(m[1], 0.7773) < 2e-4);
-        }
-
-        [Fact]
-        ////[DeploymentItem(@"Data\test.mat", "Data")]
-        public void MatlabReaderTest()
-        {
-            MatlabReaderTester(Path.Combine(
-#if NETCORE
-                Path.GetDirectoryName(typeof(PsychTests).Assembly.Location), // work dir is not the one with Microsoft.ML.Probabilistic.Tests.dll on netcore and neither is .Location on netfull
-#endif
-                "Data", "test.mat"));
-        }
-
-        private void MatlabReaderTester(string fileName)
-        {
-            Dictionary<string, object> dict = MatlabReader.Read(fileName);
-            Assert.Equal(12, dict.Count);
-            Matrix aScalar = (Matrix) dict["aScalar"];
-            Assert.Equal(1, aScalar.Rows);
-            Assert.Equal(1, aScalar.Cols);
-            Assert.Equal(5.0, aScalar[0, 0]);
-            Assert.Equal("string", (string) dict["aString"]);
-            MatlabReader.ComplexMatrix aComplexScalar = (MatlabReader.ComplexMatrix) dict["aComplexScalar"];
-            Assert.Equal(5.0, aComplexScalar.Real[0, 0]);
-            Assert.Equal(3.0, aComplexScalar.Imaginary[0, 0]);
-            MatlabReader.ComplexMatrix aComplexVector = (MatlabReader.ComplexMatrix) dict["aComplexVector"];
-            Assert.Equal(1.0, aComplexVector.Real[0, 0]);
-            Assert.Equal(2.0, aComplexVector.Imaginary[0, 0]);
-            Assert.Equal(3.0, aComplexVector.Real[0, 1]);
-            Assert.Equal(4.0, aComplexVector.Imaginary[0, 1]);
-            var aStruct = (Dictionary<string, object>) dict["aStruct"];
-            Assert.Equal(2, aStruct.Count);
-            Assert.Equal(1.0, ((Matrix) aStruct["field1"])[0]);
-            Assert.Equal("two", (string) aStruct["field2"]);
-            object[,] aCell = (object[,]) dict["aCell"];
-            Assert.Equal(1.0, ((Matrix) aCell[0, 0])[0]);
-            int[] intArray = (int[]) dict["intArray"];
-            Assert.Equal(1, intArray[0]);
-            int[] uintArray = (int[])dict["uintArray"];
-            Assert.Equal(1, uintArray[0]);
-            bool[] aLogical = (bool[]) dict["aLogical"];
-            Assert.True(aLogical[0]);
-            Assert.True(aLogical[1]);
-            Assert.False(aLogical[2]);
-            object[,,] aCell3D = (object[,,]) dict["aCell3D"];
-            Assert.Null(aCell3D[0, 0, 0]);
-            Assert.Equal(7.0, ((Matrix) aCell3D[0, 0, 1])[0, 0]);
-            Assert.Equal(6.0, ((Matrix) aCell3D[0, 1, 0])[0, 0]);
-            double[,,,] array4D = (double[,,,]) dict["array4D"];
-            Assert.Equal(4.0, array4D[0, 0, 1, 0]);
-            Assert.Equal(5.0, array4D[0, 0, 0, 1]);
-            long[] aLong = (long[]) dict["aLong"];
-            Assert.Equal(1234567890123456789L, aLong[0]);
-        }
-
-        [Fact]
-        //[DeploymentItem(@"Data\test.mat", "Data")]
-        public void MatlabWriterTest()
-        {
-            Dictionary<string, object> dict = MatlabReader.Read(Path.Combine(
-#if NETCORE
-                Path.GetDirectoryName(typeof(PsychTests).Assembly.Location), // work dir is not the one with Microsoft.ML.Probabilistic.Tests.dll on netcore and neither is .Location on netfull
-#endif
-                "Data", "test.mat"));
-            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriterTest{Environment.CurrentManagedThreadId}.mat";
-            using (MatlabWriter writer = new MatlabWriter(fileName))
-            {
-                foreach (var entry in dict)
-                {
-                    writer.Write(entry.Key, entry.Value);
-                }
-            }
-            MatlabReaderTester(fileName);
-        }
-
-        [Fact]
-        public void MatlabWriteStringDictionaryTest()
-        {
-            Dictionary<string, string> dictString = new Dictionary<string, string>();
-            dictString["a"] = "a";
-            dictString["b"] = "b";
-            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriteStringDictionaryTest{Environment.CurrentManagedThreadId}.mat";
-            using (MatlabWriter writer = new MatlabWriter(fileName))
-            {
-                writer.Write("dictString", dictString);
-            }
-            Dictionary<string, object> vars = MatlabReader.Read(fileName);
-            Dictionary<string, object> dict = (Dictionary<string, object>)vars["dictString"];
-            foreach (var entry in dictString)
-            {
-                Assert.Equal(dictString[entry.Key], dict[entry.Key]);
-            }
-        }
-
-        [Fact]
-        public void MatlabWriteStringListTest()
-        {
-            List<string> strings = new List<string>();
-            strings.Add("a");
-            strings.Add("b");
-            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriteStringListTest{Environment.CurrentManagedThreadId}.mat";
-            using (MatlabWriter writer = new MatlabWriter(fileName))
-            {
-                writer.Write("strings", strings);
-            }
-            Dictionary<string, object> vars = MatlabReader.Read(fileName);
-            string[] array = (string[])vars["strings"];
-            for (int i = 0; i < array.Length; i++)
-            {
-                Assert.Equal(strings[i], array[i]);
-            }
-        }
-
-        [Fact]
-        public void MatlabWriteEmptyArrayTest()
-        {
-            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriteEmptyArrayTest{Environment.CurrentManagedThreadId}.mat";
-            using (MatlabWriter writer = new MatlabWriter(fileName))
-            {
-                writer.Write("ints", new int[0]);
-            }
-            Dictionary<string, object> vars = MatlabReader.Read(fileName);
-            int[] ints = (int[])vars["ints"];
-            Assert.Empty(ints);
-        }
-
-        [Fact]
-        public void MatlabWriteNumericNameTest()
-        {
-            string fileName = $"{System.IO.Path.GetTempPath()}MatlabWriteNumericNameTest{Environment.CurrentManagedThreadId}.mat";
-            using (MatlabWriter writer = new MatlabWriter(fileName))
-            {
-                writer.Write("24", new int[0]);
-            }
-            Dictionary<string, object> vars = MatlabReader.Read(fileName);
-            int[] ints = (int[])vars["24"];
-            Assert.Empty(ints);
-        }
-
-#if SUPPRESS_UNREACHABLE_CODE_WARNINGS
-#pragma warning disable 162
-#endif
-
         /// <summary>
         /// Nonconjugate VMP crashes with improper message on the first iteration.
         /// </summary>
@@ -280,7 +111,8 @@ namespace Microsoft.ML.Probabilistic.Tests
             //response.AddAttribute(new MarginalPrototype(new Gaussian())); 
             bool[,] data;
             double[] discriminationTrue = new double[0];
-            if (false)
+            bool useDummyData = false;
+            if (useDummyData)
             {
                 data = new bool[4,2];
                 for (int i = 0; i < data.GetLength(0); i++)
@@ -315,10 +147,6 @@ namespace Microsoft.ML.Probabilistic.Tests
                 Console.WriteLine(marg[i].GetMean() + " \t " + discriminationTrue[i]);
         }
 
-#if SUPPRESS_UNREACHABLE_CODE_WARNINGS
-#pragma warning restore 162
-#endif
-
         internal void LogisticIrtProductExpTest()
         {
             int numStudents = 20;