Table level constraint to prevent overlapping date ranges

问题

Below is the schema where I need to put a constraint such that a second new entry could be put in for a room number, even before the existing depDt for the same room number. Could any of you please help me with this??

CREATE TABLE Accomodation (
  roomNo INTEGER NOT NULL,
  arrDt DATE NOT NULL,
  depDt DATE NOT NULL,
  PRIMARY KEY (roomNo, arrDt), 
  CONSTRAINT date_chk CHECK (arrDt < depDt)
);

INSERT INTO HotelStays(roomNo, arrDt, depDt) VALUES 
  (123, to_date('20160202', 'YYYYMMDD'),to_date('20160206','YYYYMMDD')),
  (123, to_date('20160205', 'YYYYMMDD'), to_date('20160208','YYYYMMDD'));

I have tried giving a sub query under WHERE in CONSTRAINTS but its is not working in SQL Fiddle.

回答1:

This can be done using an exclusion constraint on the date range:

alter table Accomodation
  add constraint no_overlap 
  exclude using gist (roomno with =, daterange(arrdt, depdt) with &&);

Note that you need the btree_gist extension to support the = operator in a GiST index.

回答2:

Note: This does not solve the problem of race conditions.

Create a function that checks whether a room is available based on your conditions and returns a scalar boolean value which can be used in CHECK constraint.

Here you can preview how it works (remember to uncomment the last insert statement):SQL FIDDLE

CREATE FUNCTION is_room_available(int, date)
RETURNS boolean
STABLE
LANGUAGE plpgsql
AS 
$$
BEGIN
  IF EXISTS ( SELECT 1 FROM Accomodation WHERE roomNo = $1 AND $2 BETWEEN arrDt AND depDt ) THEN
    RETURN false;
  END IF;

  RETURN true;
END;
$$;

Create table with new constraint

CREATE TABLE Accomodation (
  roomNo INTEGER NOT NULL,
  arrDt DATE NOT NULL,
  depDt DATE NOT NULL,
  PRIMARY KEY (roomNo, arrDt), 
  CONSTRAINT date_chk CHECK (arrDt<depDt),
  CONSTRAINT room_avail CHECK (is_room_available(roomNo, arrDt)) -- added
  );

Try inserting two rows in separate statements

INSERT INTO Accomodation(roomNo, arrDt, depDt)
VALUES 
(123, to_date('20160202', 'YYYYMMDD'), to_date('20160206','YYYYMMDD'));

INSERT INTO Accomodation(roomNo, arrDt, depDt)
VALUES 
(123, to_date('20160205', 'YYYYMMDD'), to_date('20160208','YYYYMMDD'));

First value is inserted, while when issuing the second insert statement you get a check constraint violation

ERROR: new row for relation "accomodation" violates check constraint "room_avail" Detail: Failing row contains (123, 2016-02-05, 2016-02-08).

Note: This could be easily implemented using triggers as well. You just need to modify the function a little and issue a CREATE TRIGGER statement.

来源：https://stackoverflow.com/questions/35521073/table-level-constraint-to-prevent-overlapping-date-ranges

标签

sql

postgresql

constraints