To make the long story short, we have a situation where there can be different datatypes for a column. For example, 'AttributeValue' can be a datetime, int, float etc. My DBA is suggesting the following design,
For any AttributeId, only one of the value columns will have a value and the rest will be NULLs. I feel that this is a bad design. (FYI, We decided against SQLVariant for specific reasons). I have suggested, we create a seperate table for each datatype, like
tblAttribute: AttributeId, AttributeDataTypeId
tblAttributeValueFloat: AttributeId INT, AttributeValue FLOAT
tblAttibuteValueDateTime: AttributeId INT, AttributeValue DATETIME
etc for each datatype.
Depending on the AttributeDataTypeId in the Attribute table go to the specific table and pick up the value. This way, we don't need to have too many NULLs in a row. Which of the 2 do you think is a better design? And if you agree with me, what would be the points to support it.
Because AttributeValueInt, AttributeValueFloat, AttributeValueDateTime etc will have to allow nulls and even if you index them, the execution plan may be based on index scan which in your case will be almost identical to table scan.