通过不断编辑维护订单数据的完整性

Question

hypothetically, if i were with a system that keeps track of products and order information across multiple tables (orders, order_items, product)

orders
    id              INT(11)
    shipping_name   VARCHAR(255)
    shipping_street VARCHAR(255)
    shipping_city   VARCHAR(255)
    [etc] 

order_details
    id              INT(11)
    order_id        INT(11)
    product_id      INT(11)

products
    id              INT(11)
    name            VARCHAR(255)
    description     VARCHAR(255)
    price           DECIMAL(8,2)

structure is very simple order has multiple order_items , order_items has one product .

the problem is that when someone edits a product, those edits modify the data of previous orders. if an employee were to go back and look at that information later on, they may not have the same information that the customer received at the time the order was placed.

What would be best practice? should i add a 'display_item' field to the products table, and on edit/delete set display to 0 and add edited product as new row? should i duplicate the name, description, and price in order_details?

Answer 1

I think this is one of those cases where database normalization "breaks".

Some possible solutions:

1. Keep a copy of the product attributes for each order. This is expensive in terms of storage, but it makes it easier to track down the product data stored in the order.
2. Create a log of attributes that can be changed in time. Product attributes can change over time, so a log which stores the modification date can help you filter out the product attributes to the moment the order was made.

Proposal for option 1

Create a copy of the products table and create a relation (one-to-one) to the order_details table for each order and order detail.

Proposal for option 2

Split the products table in two: product_general_info and product_attributes . Product general info is meant to be stable through time (a product's general info will not change), as any modification to the data in this table will propagate to the whole orders set. Product attributes must have a date or timestamp value to define when the attributes changed. Then you can query the database and return the last record that is before or on the order date.