Designing a data-driven logic system

Question

I'm developing a tax calculation system that applies various taxes based on a set of supplied criteria.
The information frequently changes, so I'm trying to create a way to store all these logic rules in the database.

As you can imagine, there is a lot of compound logic involved in applying taxes.
For example, a tax might only apply if A is true, B is less than 100, and C equals 7 .

My current design is terrible.
I have a few database columns for very common criteria filtering, such as location and tax year.
For more complex logic, I have a column that holds JavaScript, and in code, I run an interpreter to filter the results. Performance and maintainability suck.

I'd like to improve this design by making the logic entirely data-driven, but I'm having trouble figuring out how to correctly represent this logic within a relational database. What is a good way to model this logic in the database?

Answer 1

I have worked on this similar issue for over a year now for a manufacturing cost generation application. Similarly, it takes in loads of product design data input and base on the design, and other inventory considerations such as quantity, bulk purchase options, part supplier, electrical ratings etc. The result is a list of direct materials, labour and costs.

I knew from the onset that what I need is some kind of query language instead of a computational one, and it has to be scripted, not compiled. But I have yet to find a perfect solution:

METHOD 1 - SQL I created tables that represents my objects and columns that represents properties and then manually typed in the all the SQL SELECT statments required in an item_rules table. What I did was to first save the object into the database, then then I did

rules = SELECT * FROM item_rules
foreach(rules as _rule)
{
   count = SELECT COUNT(*) FROM (_rule[select_statement]) as T1
   if(count > 1) itemlist.add(_rule[item_that_satisfy_rule])
}

What it does is it takes each rule in the item_rules table and run it against my object that is now in the tables. eg SELECT * FROM my_object WHERE A=5 AND B>10. If I successfully pick it up, I get a positive count and then I know I should include the corresponding rule item to my items list.

METHOD 2 - NCALC Instead of storing the queries in SQL format, I found the NCALC opensource expression parsing library. NCALC takes a string expression and option variable and computes a result. The string expressions can be stored in plain text on the filesystem.

METHOD 3 - EXCEL EXCEL is actually a very good piece of software for doing data lookups. You can create the formulas in excel and then feed data from your application into excel and then let excel run the formulas to give you the results. Advantage is that many people knows how to use excel, so different people can maintain it.

But like I say, none of these are perfect for me. I am just sharing and hopefully we can get better recommedations.

Answer 2

如果要使用Jake的方法，那么也可以使用Dynamic Sql。