Question

Is it possible to have a many to many relationship between two tables, and enforce that all the members of a group are to have a particular attribute in common?

For example, a worker can be in several groups, and a group can have several workers, but all the workers in a group must be on the same site. There are enough workers and sites that I can’t make a new table for each site.

–EDIT–

This is the simplified schema. I’m using mySQL workbench, but I think this is right:

-- -----------------------------------------------------
-- Table `DB`.`Worker`
-- -----------------------------------------------------
CREATE  TABLE IF NOT EXISTS `DB`.`Worker` (
  `workerID` INT NOT NULL ,
  `site` VARCHAR(45) ,
  PRIMARY KEY (`workerID`) ;

-- -----------------------------------------------------
-- Table `DB`.`Group`
-- -----------------------------------------------------
CREATE  TABLE IF NOT EXISTS `DB`.`Group` (
  `groupID` INT NOT NULL ,
  PRIMARY KEY (`groupID`) ;

-- -----------------------------------------------------
-- Table `DB`.`Worker_Group`
-- -----------------------------------------------------
CREATE  TABLE IF NOT EXISTS `DB`.`workerGroup` (
  `workerID` INT NOT NULL ,
  `groupID` INT NOT NULL ,
  PRIMARY KEY (`workerID`, `groupID`) ,
  INDEX `fk_Group` (`groupID` ASC) ,
  CONSTRAINT `fk_Worker`
    FOREIGN KEY (`workerID` )
    REFERENCES `DB`.`Worker` (`workerID` )
    ON DELETE NO ACTION
    ON UPDATE NO ACTION,
  CONSTRAINT `fk_Group`
    FOREIGN KEY (`groupID`)
    REFERENCES `DB`.`Group` (`groupID`)
    ON DELETE NO ACTION
    ON UPDATE NO ACTION) ;

Answer 1

You do that by using composite foreign keys.

Instead of foreign keying on just one field, you use a foreign key made out of several (two in this case) fields.

CREATE TABLE site (
  id         INT AUTO_INCREMENT,
  name       VARCHAR(256),
  PRIMARY KEY (id)
)

CREATE TABLE worker (
  id         INT AUTO_INCREMENT,
  name       VARCHAR(256),
  site_id    INT,
  PRIMARY KEY (id, site_id),
  FOREIGN KEY (site_id) REFERENCES site (id)
)

CREATE TABLE group (
  id         INT AUTO_INCREMENT,
  name       VARCHAR(256),
  site_id    INT,
  PRIMARY KEY (id, site_id),
  FOREIGN KEY (site_id) REFERENCES site (id)
)

CREATE TABLE map_worker_group (
  worker_id  INT,
  site_id    INT,
  group_id   INT,
  PRIMARY KEY (worker_id, group_id, site_id),
  FOREIGN KEY (group_id,  site_id) REFERENCES group  (id, site_id),
  FOREIGN KEY (worker_id, site_id) REFERENCES worker (id, site_id)
)

Now a group is specific to a site, as is a worker. And when populating the worker:group mapping table you reference the sites of both. This means that a worker can only be in a group with the same site_id.