PostgreSQL 用户定义类型枚举与外键 - 维护和性能

Question

I am trying to decide between two approaches to user defined set of values for column in PostgreSQL table.

User defined data type:

CREATE TYPE track_class AS ENUM(
  'Easy', 'Medium', 'Difficult', 'Very difficult', 'NA'
);

Foreign key:

CREATE TABLE track_class (
    class TEXT PRIMARY KEY
);

INSERT INTO track_class (class) VALUES ('Easy'), ('Medium'), ('Difficult'), ('Very difficult'), ('NA');

CREATE TABLE tracks (
    track_class TEXT REFERENCES track_class (class) ON UPDATE CASCADE
);

Is any one superior to the other from maintenance (add / edit / replace value to domain later) and/or performance (where, group by, order by) perspective talking about millions of records, or it does not make much difference?

Answer 1

I would not recommend using an enum. Espacially deleting enum values is not possible without dropping the type.

I guess very easy to handle and best in terms of modelling would be the new table with foreign key you mentioned. Performance should be no problem here, if you put an index on the foreign-key-column.

Another option would be to use a CHECK -constraint:

ALTER TABLE mytable
    ADD CONSTRAINT myconstraint_only_some_values
        CHECK (mycolumn IN ('Easy', 'Medium', 'Difficult', 'Very difficult', 'NA'));

Perhaps this is more "visible" in your data and you don't have to think about joining. If you do not forget the index on mycolumn , this version should be just as performant as the one with the foreign key.