When using a RowTableSource, unprojected columns are properly handled by SetRowValue:

However, DataChunk which is exposed to the FillChunk method on ChunkTableSource does not have any knowledge of which columns are projected. This results in errors like invalid column count: expected 1, got 1 when calling chunk.SetValue for too many columns.

Potential solutions:

• Give DataChunk internal knowledge about projected column indices and ignore unprojected columns when SetValue is called. This would mimic the Row type.
  • Cons: this field would not be relevant in all use cases of DataChunk and could cause confusion
• Wrap DataChunk with a table UDF specific type that handles this
  • Pros: Doesn't have the con from above
  • Cons: Breaking change to table UDFs