基础数据库设计方案

Question

Problem definition: a company wants to measure customers satisfaction of their products by asking some questions. company's prducts data stores in "products" table, customers data of each product stores in "customers" table and questions of each product stores in questions table. As simple as below:

products: [ id, name ]

customers: [ id, product_id(fk), name ]

questions: [ id, product_id(fk), text ]

the question is, how can we design a table to store answers of each customer by considering that each customer can only answer its product's question. For example, if your answer is something like this:

customer_answer:[ id, customer_id(fk), question_id(fk), answer ]

then its wrong because if our table's sample data is like the following:

products: [ 1, "Chair" ], [ 2, "Table" ]

customers:[ 1, 1, "Anthony Quinn" ], [ 2, 2, "Marilyn Monroe" ]

question: [ 1, 1, "Any suggestion for this product ?" ], [ 2, 1, "How do you estimate this chair ?" ]

then there might be a possible row with below data:

customer_answer: [ 1, 2, 1, 'No, thanks' ]

that is wrong, because the customer with id #2 don't have the product_id #1, so it can be wrong table for this problem.

What can we do in such situations ?

it doesn't matter but BTW I'm using mysql for solving this problem.

Answer 1

You should define a foreign key constraint on customer_id and product_id in the customer_answer table referencing the id and product_id columns in the customer table to enforce storing answers to only customer's questions.

CREATE TABLE customer_answer
(
    id  INT,
    customer_id INT,
    product_id INT, 
    answer VARCHAR(128),
)
ALTER TABLE customer_answer
ADD CONSTRAINT FK_Customer_Answer_To_Customer FOREIGN KEY (customer_id, product_id) REFERENCES customer(id, product_id)

Answer 2

This is a case of surrogate keys hurting, rather than helping, your database design. Also, you should probably have a separate Customers table to uniquely list customers.

Here's a quick design that removes the surrogate keys from the tables in which the existence of the surrogate keys makes it harder to enforce referential integrity. The critical change is that CustomerProducts (what I've renamed your Customers table) now uses CustomerID, ProductID as the PRIMARY KEY, making it possible to enforce the integrity of the relationship between answers and products owned.

 Products:         ProductID, Name 
                       (PK = ProductID)
 Customers:        CustomerID, Name 
                       (PK = CustomerID)
 CustomerProducts: CustomerID, ProductID 
                       (PK = CustomerID, ProductID)
                       (FK = CustomerID REFERENCES Customers)
                       (FK = ProductID  REFERENCES Products)
 Questions:        ProductID, QuestionText 
                       (PK = ProductID, QuestionText)
                       (FK = ProductID REFERENCES Products)
 Answers:   CustomerID, ProductID, QuestionText 
                       (FK CustomerID, ProductID REFERENCES CustomerProducts)
                       (FK ProductID, QuestionText REFERENCES Questions)

Note: If you feel uncomfortable using the large VARCHAR field QuestionText as part of the PK on Questions and the FK on Answers (especially since the contents may change), you can reintroduce a surrogate key here or, perhaps better, add a column QuestionNo to provide ordering of questions within a single product; then make ProductID, QuestionNo the primary and foreign key.

Answer 3

Note 2 FKs on CustomerQuestion table: FK1 {CustomerID, ProductID} and FK2 {ProductID, QuestionID}

Answer 4

我认为这应该是应用程序数据层（例如MVC体系结构中的模型）在插入数据之前进行必要检查的工作。

Answer 5

或者，您可以尝试在插入“ customer_answer”时编写MySQL触发器，该触发器将检查是否要在客户表中找到要添加的新组合（产品ID和客户ID），如果是，则继续操作，如果不，你可以还原。