added test for column property and some review comments

Author: artidoro

src/Microsoft.ML.Data/Transforms/ColumnConcatenatingEstimator.cs

@@ -11,7 +11,10 @@
 
 namespace Microsoft.ML.Transforms
 {
-    public sealed class ColumnConcatenatingEstimator : IEstimator<ITransformer>
+    /// <summary>
+    /// Concatenates columns in an <see cref="IDataView"/> together.
+    /// </summary>
+    public sealed class ColumnConcatenatingEstimator : IEstimator<ColumnConcatenatingTransformer>
     {
         private readonly IHost _host;
         private readonly string _name;
@@ -38,9 +41,9 @@ internal ColumnConcatenatingEstimator(IHostEnvironment env, string outputColumnN
         }
 
         /// <summary>
-        /// Train and return a transformer.
+        /// Trains and returns a <see cref="ColumnConcatenatingTransformer"/>.
         /// </summary>
-        public ITransformer Fit(IDataView input)
+        public ColumnConcatenatingTransformer Fit(IDataView input)
         {
             _host.CheckValue(input, nameof(input));
             return new ColumnConcatenatingTransformer(_host, _name, _source);

src/Microsoft.ML.Data/Transforms/ColumnConcatenatingTransformer.cs

@@ -33,6 +33,9 @@ namespace Microsoft.ML.Data
 {
     using PfaType = PfaUtils.Type;
 
+    /// <summary>
+    /// Concatenates columns in an <see cref="IDataView"/> together.
+    /// </summary>
     public sealed class ColumnConcatenatingTransformer : RowToRowTransformerBase
     {
         internal const string Summary = "Concatenates one or more columns of the same item type.";
@@ -215,7 +218,8 @@ internal ColumnInfo(ModelLoadContext ctx)
 
         private readonly ColumnInfo[] _columns;
 
-        internal IReadOnlyCollection<ColumnInfo> Columns => _columns.AsReadOnly();
+        public IReadOnlyCollection<(string outputColumnName, string[] inputColumnNames)> Columns
+            => _columns.Select(col => (outputColumnName: col.Name, inputColumnNames: col.Sources.Select(source => source.name).ToArray())).ToArray().AsReadOnly();
 
         /// <summary>
         /// Concatename columns in <paramref name="inputColumnNames"/> into one column <paramref name="outputColumnName"/>.

src/Microsoft.ML.Data/Transforms/ExtensionsCatalog.cs

@@ -45,7 +45,7 @@ public static ColumnCopyingEstimator CopyColumns(this TransformsCatalog catalog,
             => new ColumnCopyingEstimator(CatalogUtils.GetEnvironment(catalog), columns);
 
         /// <summary>
-        /// Concatenates two columns together.
+        /// Concatenates columns together.
         /// </summary>
         /// <param name="catalog">The transform's catalog.</param>
         /// <param name="outputColumnName">Name of the column resulting from the transformation of <paramref name="inputColumnNames"/>.</param>

src/Microsoft.ML.Data/Transforms/Normalizer.cs

@@ -239,7 +239,7 @@ internal NormalizingEstimator(IHostEnvironment env, params ColumnBase[] columns)
         }
 
         /// <summary>
-        /// Train and return a transformer.
+        /// Trains and returns a <see cref="NormalizingTransformer"/>.
         /// </summary>
         public NormalizingTransformer Fit(IDataView input)
         {

src/Microsoft.ML.Transforms/MissingValueHandlingTransformer.cs

@@ -21,9 +21,9 @@
 namespace Microsoft.ML.Transforms
 {
     /// <include file='doc.xml' path='doc/members/member[@name="NAHandle"]'/>
-    public static class MissingValueHandlingTransformer
+    internal static class MissingValueHandlingTransformer
     {
-        internal enum ReplacementKind : byte
+        public enum ReplacementKind : byte
         {
             /// <summary>
             /// Replace with the default value of the column based on its type. For example, 'zero' for numeric and 'empty' for string/text columns.
@@ -56,7 +56,7 @@ internal enum ReplacementKind : byte
             Max = Maximum,
         }
 
-        internal sealed class Options : TransformInputBase
+        public sealed class Options : TransformInputBase
         {
             [Argument(ArgumentType.Multiple | ArgumentType.Required, HelpText = "New column definition(s) (optional form: name:rep:src)", Name = "Column", ShortName = "col", SortOrder = 1)]
             public Column[] Columns;
@@ -74,7 +74,7 @@ internal sealed class Options : TransformInputBase
             public bool Concat = true;
         }
 
-        internal sealed class Column : OneToOneColumn
+        public sealed class Column : OneToOneColumn
         {
             [Argument(ArgumentType.AtMostOnce, HelpText = "The replacement method to utilize")]
             public ReplacementKind? Kind;

src/Microsoft.ML.Transforms/MissingValueReplacing.cs

@@ -893,7 +893,8 @@ private bool SaveAsOnnxCore(OnnxContext ctx, int iinfo, ColInfo info, string src
 
     public sealed class MissingValueReplacingEstimator : IEstimator<MissingValueReplacingTransformer>
     {
-        public static class Defaults
+        [BestFriend]
+        internal static class Defaults
         {
             public const ColumnInfo.ReplacementMode ReplacementMode = ColumnInfo.ReplacementMode.DefaultValue;
             public const bool ImputeBySlot = true;
@@ -912,9 +913,23 @@ public enum ReplacementMode : byte
                 Maximum = 3,
             }
 
+            /// <summary>
+            /// Name of the column resulting from the transformation of <see cref="InputColumnName"/>
+            /// </summary>
             public readonly string Name;
+            /// <summary>
+            /// Name of column to transform.
+            /// </summary>
             public readonly string InputColumnName;
+            /// <summary>
+            /// If true, per-slot imputation of replacement is performed.
+            /// Otherwise, replacement value is imputed for the entire vector column. This setting is ignored for scalars and variable vectors,
+            /// where imputation is always for the entire column.
+            /// </summary>
             public readonly bool ImputeBySlot;
+            /// <summary>
+            /// What to replace the missing value with.
+            /// </summary>
             public readonly ReplacementMode Replacement;
 
             /// <summary>
@@ -927,7 +942,7 @@ public enum ReplacementMode : byte
             /// Otherwise, replacement value is imputed for the entire vector column. This setting is ignored for scalars and variable vectors,
             /// where imputation is always for the entire column.</param>
             public ColumnInfo(string name, string inputColumnName = null, ReplacementMode replacementMode = Defaults.ReplacementMode,
-                bool imputeBySlot = MissingValueReplacingEstimator.Defaults.ImputeBySlot)
+                bool imputeBySlot = Defaults.ImputeBySlot)
             {
                 Contracts.CheckNonWhiteSpace(name, nameof(name));
                 Name = name;
@@ -985,7 +1000,7 @@ public SchemaShape GetOutputSchema(SchemaShape inputSchema)
         }
 
         /// <summary>
-        /// Train and return a transformer.
+        /// Trains and returns a <see cref="MissingValueReplacingTransformer"/>.
         /// </summary>
         public MissingValueReplacingTransformer Fit(IDataView input) => new MissingValueReplacingTransformer(_host, input, _columns);
     }

test/Microsoft.ML.Benchmarks/Helpers/EnvironmentFactory.cs

@@ -39,7 +39,7 @@ internal static MLContext CreateRankingEnvironment<TEvaluator, TLoader, TTransfo
             environment.ComponentCatalog.RegisterAssembly(typeof(TTransformer).Assembly);
             environment.ComponentCatalog.RegisterAssembly(typeof(TTrainer).Assembly);
 
-            environment.ComponentCatalog.RegisterAssembly(typeof(MissingValueHandlingTransformer).Assembly);
+            environment.ComponentCatalog.RegisterAssembly(typeof(MissingValueDroppingTransformer).Assembly);
 
             return ctx;
         }

test/Microsoft.ML.Tests/Transformers/ConcatTests.cs

@@ -6,6 +6,7 @@
 using Microsoft.Data.DataView;
 using Microsoft.ML.Data;
 using Microsoft.ML.Data.IO;
+using Microsoft.ML.Internal.Utilities;
 using Microsoft.ML.RunTests;
 using Microsoft.ML.Transforms;
 using Xunit;
@@ -107,6 +108,18 @@ ColumnType GetType(Schema schema, string name)
                 new ColumnConcatenatingTransformer.ColumnInfo("f3", new[] { ("float4", "FLOAT4"), ("float1", "FLOAT1") }));
             data = concater.Transform(data);
 
+            // Test Columns property.
+            var columns = concater.Columns;
+            var colEnumerator = columns.GetEnumerator();
+            colEnumerator.MoveNext();
+            Assert.True(colEnumerator.Current.outputColumnName == "f2" && 
+                colEnumerator.Current.inputColumnNames[0] == "float1" && 
+                colEnumerator.Current.inputColumnNames[1] == "float1");
+            colEnumerator.MoveNext();
+            Assert.True(colEnumerator.Current.outputColumnName == "f3" &&
+                colEnumerator.Current.inputColumnNames[0] == "float4" &&
+                colEnumerator.Current.inputColumnNames[1] == "float1");
+
             ColumnType t;
             t = GetType(data.Schema, "f2");
             Assert.True(t is VectorType vt2 && vt2.ItemType == NumberType.R4 && vt2.Size == 2);