Question

I have an InnoDB MySQL database with a table that needs to be able to connect to one of 26 other tables via a foreign key. Each record will only connect to one of these 26 at a time. The table will probably consist of no more than 10,000 records. Is there an alternative way to do this?

-- -----------------------------------------------------
-- Table `db_mydb`.`tb_job`
-- -----------------------------------------------------
CREATE  TABLE IF NOT EXISTS `db_mydb`.`tb_job` (
  `job_id` INT(11) NOT NULL AUTO_INCREMENT ,
  // Removed 26 other fields that the table requires

  `job_foreignkey_a_id` INT(11) NULL DEFAULT NULL ,
  `job_foreignkey_b_id` INT(11) NULL DEFAULT NULL ,
  `job_foreignkey_c_id` INT(11) NULL DEFAULT NULL ,
  // Removed the other 23 foreign keys fields that are the same

  PRIMARY KEY (`job_id`) ,

  CONSTRAINT `fka_tb_job_tb`
    FOREIGN KEY (`job_foreignkey_a_id` )
    REFERENCES `db_mydb`.`tb_foreignkey_a` (`foreignkey_a_id` )
    ON DELETE NO ACTION
    ON UPDATE NO ACTION,
  CONSTRAINT `fkb_tb_job_tb`
    FOREIGN KEY (`job_foreignkey_b_id` )
    REFERENCES `db_mydb`.`tb_foreignkey_b` (`foreignkey_b_id` )
    ON DELETE NO ACTION
    ON UPDATE NO ACTION,
  CONSTRAINT `fkc_tb_job_tb`
    FOREIGN KEY (`job_foreignkey_c_id` )
    REFERENCES `db_mydb`.`tb_foreignkey_c` (`foreignkey_c_id` )
    ON DELETE NO ACTION
    ON UPDATE NO ACTION)
  // Removed the other 23 foreign keys constraints that are the same

ENGINE = InnoDB
DEFAULT CHARACTER SET = utf8;

CREATE INDEX `fka_tb_job_tb` ON `db_mydb`.`tb_job` (`job_foreignkey_a_id` ASC) ;
CREATE INDEX `fkb_tb_job_tb` ON `db_mydb`.`tb_job` (`job_foreignkey_b_id` ASC) ;
CREATE INDEX `fkc_tb_job_tb` ON `db_mydb`.`tb_job` (`job_foreignkey_c_id` ASC) ;
// Removed the other 23 foreign keys indexes that are the same

Answer 1

This is the problem of generic foreign keys, which MySQL and friends tend not to support. There are two ways you can do this.

The first, as you have done, is nullable foreign keys, one for every type.

The other, as in Django’s Content Types, is to have a join table, each row having a row id and a field that specifies the table to look up on. Your code then has to formulate the SQL query depending on the contents of the field. It works well, but has limitations:

The downside of the first one is bloat, but it brings you the upsides of normal FKs, i.e. referential integrity and SQL joins etc, both of which are very valuable. You can’t get those with the second method.