Forum OpenACS Development: Re: complex survey datamodel suggestion/question
In the table where we store returned data (as_item_data -- see https://openacs.org/projects/openacs/packages/assessment/design/data), we use a single numeric_answer column to stuff the actual values for integer and numeric data, but we know from the corresponding as_item_types value which type it really is. Similarly, we can stuff an "exponential" value in the varchar_answer column of as_item_data but we'll know that it's a "number string" from as_item_types.
Thanks for your offer of the procs. Hopefully we'll start cranking out code for Assessment soon.
Rephrased: Is there a compelling need for the database to make sure of integrity, thereby adding additional columns to the table and confusing things, instead of having the API make sure that only items that have been checked for validity (pending on the item_type as Stan pointed out) are stored.