>Hello,
>
>In the past I have broken data normalization rules but I am swearing that off :) I also want some table columns to be defined in a data driven way.
>
>I'm asking for help to see if I am missing something important. Please give me your comments and criticism as I need to make a choice soon and don't want to make a bad one.
>
>Say you have the following tables:
>1. Header Table
> header_id, other fields
>2. Child Table
> child_id, header_id, other fields
>3. Table with name and value of "virtual fields" of child table - One record for each virtual column that has a non default value
> Virtual_id, child_id, ColumnName, ColumnValue
>4. Data dictionary with name of all possible "virtual fields" of child table - Contains description of all possible virtual columns
> ColumnName,ColumnType,ColumnWidth,ColumnPrecision
>
>I am considering the following approach to integrate data from tables 2 and 3 during user edits:
>I use Codemine framework which makes it relatively easy to go the cursor approach.
>- Create either an object or a cursor with the following properties / fields
> child_id, header_id, other fields from child table, VirtualColumn1, VirtualColumn2, VirtualColumn3 etc
>- Update base tables as user makes changes
>
Instead of step 3 you might consider the route of NoSQL,
which can to a part be simulated in vfp utilizing 1 memo field filled with key/value pairs

Formats well adapted are Ini,JSon or Xml.
Saves a lot of Join work but leaves you with packing worry for high-write setups.
There is no rule that you cannot employ DD setups like step 4 such NoSQL/EVA hybrids
- I even think it is a good policy ;-)

regards


thomas