MySQL - 超类型/子类型设计

Question

我需要创建以下数据库:

对于半卡车我不需要额外的亚型,而对于Car我只需要那3种亚型而且对于轿车我也需要四种亚型.对于SELECT,我将使用JOIN(规范化数据库),但我需要找到一种简单的方法来进行INSERT.

• 车辆表存储公共信息
• 半卡车存储半成品的特定信息
• 汽车表具有汽车的特定字段和与三个子类型链接的car_type字段
• Van,Suv和Sedan(以及其他类型,如果我需要它们)应该在一个表CAR_TYPE中
• 但是,对于Sedan类型,我需要有另外的子类型,这些子类型可能应该包含在另一个表中.Suvs和Vans不需要这些子类型(在现实生活中,厢式货车可以与轿车具有相同的子类型但在我的情况下不是这样).

我需要完全按照图中的方式创建此数据库.

到目前为止,我的第一种方法是使用以下表格:

• 车辆:veh_id,veh_type(半汽车,汽车),...,other_fields
• Vehicle_semis:veh_id,...,other_semis_fields
• Vehicle_car:veh_id,car_type(Van,Suv,Sedan),other_car_specific_fields
• Car_type:car_type_id,类型
• Sedan_type:sedan_type_id,输入

我的问题是我不确定这是正确的方法,我不知道如何在表之间创建关系.

有任何想法吗？

谢谢!

更新:

下图基于@Mike的答案:

Answer 1

Before I get started, I want to point out that "gas" describes either fuel or a kind of engine, not a kind of sedan. Think hard before you keep going down this path. (Semantics are more important in database design than most people think.)

What you want to do is fairly simple, but not necessarily easy. The important point in this kind of supertype/subtype design (also known as an exclusive arc) is to make it impossible to have rows about sedans referencing rows about semi-trucks, etc..

MySQL makes the code more verbose, because it doesn't enforce CHECK constraints. You're lucky; in your application, the CHECK constraints can be replaced by additional tables and foreign key constraints. Comments refer to the SQL above them.

create table vehicle_types (
  veh_type_code char(1) not null,
  veh_type_name varchar(10) not null,
  primary key (veh_type_code),
  unique (veh_type_name)
);

insert into vehicle_types values
('s', 'Semi-truck'), ('c', 'Car');

This is the kind of thing I might implement as a CHECK constraint on other platforms. You can do that when the meaning of the codes is obvious to users. I'd expect users to know or to figure out that 's' is for semis and 'c' is for cars, or that views/application code would hide the codes from users.

create table vehicles (
  veh_id integer not null,
  veh_type_code char(1) not null,
  other_columns char(1) default 'x',
  primary key (veh_id),
  unique (veh_id, veh_type_code),
  foreign key (veh_type_code) references vehicle_types (veh_type_code)
);

The UNIQUE constraint lets the pair of columns {veh_id, veh_type_code} be the target of a foreign key reference. That means a "car" row can't possibly reference a "semi" row, even by mistake.

insert into vehicles (veh_id, veh_type_code) values
(1, 's'), (2, 'c'), (3, 'c'), (4, 'c'), (5, 'c'), 
(6, 'c'), (7, 'c');

create table car_types (
  car_type char(3) not null,
  primary key (car_type)
);

insert into car_types values
('Van'), ('SUV'), ('Sed');

create table veh_type_is_car (
  veh_type_car char(1) not null,
  primary key (veh_type_car)
);

Something else I'd implement as a CHECK constraint on other platforms. (See below.)

insert into veh_type_is_car values ('c');

Only one row ever.

create table cars (
  veh_id integer not null,
  veh_type_code char(1) not null default 'c',
  car_type char(3) not null,
  other_columns char(1) not null default 'x',
  primary key (veh_id ),
  unique (veh_id, veh_type_code, car_type),
  foreign key (veh_id, veh_type_code) references vehicles (veh_id, veh_type_code),
  foreign key (car_type) references car_types (car_type),
  foreign key (veh_type_code) references veh_type_is_car (veh_type_car)
);

The default value for veh_type_code, along with the foreign key reference to veh_type_is_car, guarantees that this rows in this table can be only about cars, and can only reference vehicles that are cars. On other platforms, I'd just declare the column veh_type_code as veh_type_code char(1) not null default 'c' check (veh_type_code = 'c').

insert into cars (veh_id, veh_type_code, car_type) values
(2, 'c', 'Van'), (3, 'c', 'SUV'), (4, 'c', 'Sed'),
(5, 'c', 'Sed'), (6, 'c', 'Sed'), (7, 'c', 'Sed');

create table sedan_types (
  sedan_type_code char(1) not null,
  primary key (sedan_type_code)
);

insert into sedan_types values
('g'), ('d'), ('h'), ('e');

create table sedans (
  veh_id integer not null,
  veh_type_code char(1) not null,
  car_type char(3) not null,
  sedan_type char(1) not null,
  other_columns char(1) not null default 'x',
  primary key (veh_id),
  foreign key (sedan_type) references sedan_types (sedan_type_code),
  foreign key (veh_id, veh_type_code, car_type) references cars (veh_id, veh_type_code, car_type)
);

insert into sedans (veh_id, veh_type_code, car_type, sedan_type) values 
(4, 'c', 'Sed', 'g'), (5, 'c', 'Sed', 'd'), (6, 'c', 'Sed', 'h'),
(7, 'c', 'Sed', 'e');

If you have to build additional tables that reference sedans, such as gas_sedans, diesel_sedans, etc., then you need to build one-row tables similar to "veh_type_is_car" and set foreign key references to them.

In production, I'd revoke permissions on the base tables, and either use

• updatable views to do the inserts and updates, or
• stored procedures to do the inserts and updates.