dotnet/machinelearning · error · ArgumentException

The data type '{labelType}' is not handled currently.

Error message

The data type '{labelType}' is not handled currently.

What it means

Utils.GetCSharpType maps a Microsoft.ML.Data.DataKind label/column type to the corresponding C# System.Type via an exhaustive switch. Hitting the default branch means the DataKind has no C# mapping implemented, so it throws this ArgumentException. It indicates an unhandled DataView type in the label-column generation path.

Source

Thrown at src/Microsoft.ML.CodeGenerator/Utils.cs:171

            {
                case Microsoft.ML.Data.DataKind.String:
                    return typeof(string);
                case Microsoft.ML.Data.DataKind.Boolean:
                    return typeof(bool);
                case Microsoft.ML.Data.DataKind.Single:
                    return typeof(float);
                case Microsoft.ML.Data.DataKind.Double:
                    return typeof(double);
                case Microsoft.ML.Data.DataKind.Int32:
                    return typeof(int);
                case Microsoft.ML.Data.DataKind.UInt32:
                    return typeof(uint);
                case Microsoft.ML.Data.DataKind.Int64:
                    return typeof(long);
                case Microsoft.ML.Data.DataKind.UInt64:
                    return typeof(ulong);
                default:
                    throw new ArgumentException($"The data type '{labelType}' is not handled currently.");
            }
        }

        internal static void WriteOutputToFiles(string fileContent, string fileName, string outPutBaseDir)
        {
            if (!Directory.Exists(outPutBaseDir))
            {
                Directory.CreateDirectory(outPutBaseDir);
            }
            File.WriteAllText($"{outPutBaseDir}/{fileName}", fileContent);
        }

        internal static string FormatCode(string trainProgramCSFileContent)
        {
            //Format
            var tree = CSharpSyntaxTree.ParseText(trainProgramCSFileContent);
            var syntaxNode = tree.GetRoot();
            trainProgramCSFileContent = Formatter.Format(syntaxNode, new AdhocWorkspace()).ToFullString();

View on GitHub (pinned to 7b76e69cf9)

Solutions

  1. Change the label column to a supported type (cast to Single/Double/Int32/UInt32/Int64/UInt64 as appropriate) before code generation.
  2. Upgrade Microsoft.ML.CodeGenerator to a version whose GetCSharpType covers your DataKind.
  3. Inspect the label column's schema (schema.GetColumnOrNull(label).Value.Type) before generating and reject unsupported kinds early.
  4. If you own the source, extend the switch to map the missing DataKind.

Example fix

// before: date label column passes through unchanged
pipeline.Append(mlContext.Regression.Trainers.Sdca(labelColumnName: "OrderDate"));
// after: convert label to a supported numeric kind first
var converted = mlContext.Transforms.Conversion.ConvertType("OrderDateNum", "OrderDate", outputKind: DataKind.Int32)
    .Append(pipeline);
Defensive patterns

Strategy: validation

Validate before calling

var colType = schema.GetColumnOrNull(labelColumn)?.Value.Type;
var ok = colType is NumberDataViewType ||
         colType.RawType == typeof(uint) || colType.RawType == typeof(long) ||
         colType.RawType == typeof(ulong);
if (!ok) throw new InvalidOperationException($"Label column '{labelColumn}' has unsupported kind {colType}");

Type guard

static bool HasSupportedLabelType(DataViewSchema.Column col) =>
    col.Type is NumberDataViewType n &&
    (n == NumberDataViewType.Single || n == NumberDataViewType.Double ||
     n == NumberDataViewType.Int32 || n == NumberDataViewType.UInt32 ||
     n == NumberDataViewType.Int64 || n == NumberDataViewType.UInt64);

Try / catch

try
{
    var clrType = Utils.GetCSharpType(labelType);
}
catch (ArgumentException ex) when (ex.Message.StartsWith("The data type"))
{
    // fall back to object / abort codegen for this column
    clrType = typeof(object);
}

Prevention

When it happens

Trigger: Code generation encounters a label column whose DataKind is one not covered by the switch (e.g. DateTime, TimeSpan, Boolean, String or numeric kinds not enumerated) when generating trainer/evaluator code.

Common situations: Datasets with date/time or string label columns fed to AutoML; new DataKind values added in a newer Microsoft.Data.DataView/Microsoft.ML version while an older CodeGenerator switches on them; custom loaders producing exotic column types.

Related errors

AI-assisted analysis of dotnet/machinelearning@7b76e69cf9 (2026-09-11). Data as JSON: /api/errors/5e6d47b4c9850f6f. Report an issue: GitHub.