How do I represent object classification hierarchy in a RDBMS

问题

My data, IF it would be represented by objects, it would look like:

abstract class A{
   int a;
   int b;
   string c;
}

class B inherits A{
  string D;
}

class C inherits A{
  int e;
  int f;
}

My question: Do I create a separate table for entities B and C, Or do I create one main table, and for each entity type I do different joins to fetch the data.
In the real world, I will have around 15 similar fields for all entities, and about 1-3 unique field for each entity.
I expect a max of 100K records.

Any insights?

回答1:

You'll find no shortage of opinions on this topic. Many people advocate concrete table inheritance, where (as you describe in your first option) you define the data explicitly in each table.

Given your second statement, I would not recommend this pattern. I would go with the idea of having a "main" table representing your parent and auxiliary tables representing the children. You may want to include a type identifier of some kind in your main table in order to indicate what sort of entity it is, but this may or may not be necessary. At a minimum, you need something like...

tableA
(
    ID (primary),
    A,
    B,
    C
)

tableB
(
    ID (primary and foreign->table_A),
    D
)

tableC
(
    ID (primary and foreign->table_A),
    E,
    F
)

回答2:

table_a contains all similar fields,table_B and table_c contain the unique fields

table_A
   PKa
   a int
   b int
   c string


table_B
  PKb
  FKa
  D string


table_C
  PKc
  FKa
  e int
  f int

来源：https://stackoverflow.com/questions/836383/how-do-i-represent-object-classification-hierarchy-in-a-rdbms