dotnet/machinelearning · error · NotImplementedException

nameof(T)

Error message

nameof(T)

What it means

PrimitiveDataFrameColumn<T>.GetArrowType() throws NotImplementedException(nameof(T)) when the column's generic type parameter T has no mapping to an Apache Arrow type. The method maps supported primitives (including ulong, ushort, DateTime) to Arrow types; any other T reaches the else branch. The message is just the type parameter name, so it identifies T only indirectly.

Source

Thrown at src/Microsoft.Data.Analysis/PrimitiveDataFrameColumn.cs:111

                return Int8Type.Default;
            else if (typeof(T) == typeof(int))
                return Int32Type.Default;
            else if (typeof(T) == typeof(long))
                return Int64Type.Default;
            else if (typeof(T) == typeof(short))
                return Int16Type.Default;
            else if (typeof(T) == typeof(byte))
                return UInt8Type.Default;
            else if (typeof(T) == typeof(uint))
                return UInt32Type.Default;
            else if (typeof(T) == typeof(ulong))
                return UInt64Type.Default;
            else if (typeof(T) == typeof(ushort))
                return UInt16Type.Default;
            else if (typeof(T) == typeof(DateTime))
                return Date64Type.Default;
            else
                throw new NotImplementedException(nameof(T));
        }

        protected internal override Field GetArrowField() => new Field(Name, GetArrowType(), NullCount != 0);

        protected internal override int GetMaxRecordBatchLength(long startIndex) => _columnContainer.MaxRecordBatchLength(startIndex);

        private int GetNullCount(long startIndex, int numberOfRows)
        {
            int nullCount = 0;
            for (long i = startIndex; i < numberOfRows; i++)
            {
                if (!IsValid(i))
                    nullCount++;
            }
            return nullCount;
        }

        protected internal override Apache.Arrow.Array ToArrowArray(long startIndex, int numberOfRows)

View on GitHub (pinned to 7b76e69cf9)

Solutions

1. Only create columns whose element type is an Arrow-supported primitive (bool, byte, char, decimal, double, float, int, long, sbyte, short, uint, ulong, ushort, DateTime)
2. Convert unsupported columns to a supported primitive type before Arrow export
3. Remove or skip non-primitive columns from the DataFrame before calling the Arrow conversion APIs
4. Upgrade the Microsoft.Data.Analysis package in case a newer version added the missing Arrow mapping

Example fix

// before
var col = new PrimitiveDataFrameColumn<MyStruct>("data");
var field = col.GetArrowField(); // NotImplementedException
// after
var col = new PrimitiveDataFrameColumn<int>("data"); // supported Arrow type
var field = col.GetArrowField();

static readonly HashSet<Type> ArrowSupported = new() { typeof(bool), typeof(byte), typeof(char), typeof(decimal), typeof(double), typeof(float), typeof(int), typeof(long), typeof(sbyte), typeof(short), typeof(uint), typeof(ulong), typeof(ushort), typeof(DateTime) };
if (!ArrowSupported.Contains(column.DataType)) throw new NotSupportedException($"{column.DataType} has no Arrow mapping");

static bool IsArrowExportable<T>(PrimitiveDataFrameColumn<T> col) where T : unmanaged =>
    col.DataType == typeof(bool) || col.DataType == typeof(byte) || col.DataType == typeof(char)
    || col.DataType == typeof(decimal) || col.DataType == typeof(double) || col.DataType == typeof(float)
    || col.DataType == typeof(int) || col.DataType == typeof(long) || col.DataType == typeof(sbyte)
    || col.DataType == typeof(short) || col.DataType == typeof(uint) || col.DataType == typeof(ulong)
    || col.DataType == typeof(ushort) || col.DataType == typeof(DateTime);

try { var field = column.GetArrowField(); }
catch (NotImplementedException) { /* skip column or convert before export */ }

When it happens

Trigger: Calling GetArrowField()/GetArrowType() on a PrimitiveDataFrameColumn<T> where T is not one of the explicitly mapped Arrow-supported primitives, e.g. a PrimitiveDataFrameColumn<MyStruct> or T=TimeSpan/bool-dependent conversions.

Common situations: Serializing a DataFrame to Arrow/IPC after building columns with unusual element types; passing custom struct columns into the Arrow export pipeline; library version where a primitive's Arrow mapping was not yet added.

Related errors

• type.ToString()
• {fieldType.Name}
• The method or operation is not implemented.
• NotImplementedException
• offsetBuffer

AI-assisted analysis of dotnet/machinelearning@7b76e69cf9 (2026-09-11). Data as JSON: /api/errors/bc2ef20c6216c96e. Report an issue: GitHub.