Consider CHAR->INT conversion for a SimpleColumn. The current algorithm includes two calls:

• SimpleColumn::evaluate() that:
• - detects NULL value
• - runs the actual conversion based on the SimpleColumn result data type
• - stores the converted value into the corresponding Result fields
• and TreeNode::getIntValue() that returns the corresponding Result field based on the conversion target type. Unfortunately there are cases when getIntValue() calls for a conversion.
  The flaws of this approach raise a number of questions:

• what if SimpleColumn::evaluate() is never called prior to TreeNode::getIntValue()?
• what if conversion source type or target type are not accurate?
• what if NULL value isn't detected in SimpleColumn::evaluate() and produced later in TreeNode::getIntValue()? How to propage NULL upwards?
  This issue is dedicated for the research on how to refactor such simple conversion to build a basis for complete runtime type conversion refactoring.