How can I define a one to many constraint between columns of the same table?

Question

I'm creating an IDs table in PostgreSQL to represent the relationship between multiple foreign IDs - call them a, b and c.

I happen to know that the foreign IDs have one-to-many relationships between them:

a 1..* b b 1..* c

I'd like to represent them as a single table (see below), but enforce the one-to-many invariant between a & b, & between b and c.

Is that possible?

Examples: The following insertions would be illegal:

pk	a	b	c	reason
p15	a1	b11	c111	c111 already exists
p16	a1	b21	c213	b21 already exists for a2
p17	a2	b11	c214	b11 already exists for a1

Given the following legal table:

pk	a	b	c
p1	a1	b11	c111
p2	a1	b11	c112
p3	a1	b12	c121
p4	a1	b12	c122
p5	a2	b21	c211
p6	a2	b21	c212
p7	a2	b22	c221
p8	a2	b22	c222
p9	null	b31	c311
p10	a3	null	c312
p11	a3	b31	null
p12	null	null	c314
p13	null	b31	null
p14	a3	null	null

Answer 1

This has to be done by TRIGGER, because it is too complex for CHEKCconstraint

For updating you need a similar trigger/function

CREATE TABLE tabl1
    ("pk" varchar(5), "a" varchar(6), "b" varchar(6), "c" varchar(6))
;
    
INSERT INTO tabl1
    ("pk", "a", "b", "c")
VALUES
    ('p1', 'a1', 'b11', 'c111'),
    ('p2', 'a1', 'b11', 'c112'),
    ('p3', 'a1', 'b12', 'c121'),
    ('p4', 'a1', 'b12', 'c122'),
    ('p5', 'a2', 'b21', 'c211'),
    ('p6', 'a2', 'b21', 'c212'),
    ('p7', 'a2', 'b22', 'c221'),
    ('p8', 'a2', 'b22', 'c222'),
    ('p9', NULL, 'b31', 'c311'),
    ('p10', 'a3', NULL, 'c312'),
    ('p11', 'a3', 'b31', NULL),
    ('p12', NULL, NULL, 'c314'),
    ('p13', NULL, 'b31', NULL),
    ('p14', 'a3', NULL, NULL)
;

CREATE OR REPLACE FUNCTION check_insert()
RETURNS TRIGGER AS $$
   BEGIN
   IF EXISTS(SELECT 1 FROM tabl1 WHERE "c" = NEW."c" AND "b" = NEW."b" ) then
       raise EXCEPTION 'The column %  already  with %',NEW."c",NEW."b";
   ELSIF EXISTS(SELECT 1 FROM tabl1 WHERE "a" <> NEW."a" AND "b" = NEW."b" ) then
           raise EXCEPTION 'The column %  already  has already a column',NEW."b";
   END IF;
   RETURN NEW;
   END;
$$ LANGUAGE plpgsql;

CREATE  TRIGGER trigger_name BEFORE
 INSERT
ON tabl1
FOR EACH ROW
EXECUTE PROCEDURE check_insert();

INSERT INTO tabl1
    ("pk", "a", "b", "c")
VALUES
    ('p15', 'a1', 'b11', 'c111')

ERROR:  The column c111  already  with b11

CONTEXT: PL/pgSQL function check_insert() line 4 at RAISE

INSERT INTO tabl1
    ("pk", "a", "b", "c")
VALUES
('p16' , 'a1'   , 'b21'  , 'c213' )

ERROR:  The column b21  already  has already a column

CONTEXT: PL/pgSQL function check_insert() line 6 at RAISE

INSERT INTO tabl1
    ("pk", "a", "b", "c")
VALUES
('p17',   'a2',   'b11'   ,'c214')

ERROR:  The column b11  already  has already a column

CONTEXT: PL/pgSQL function check_insert() line 6 at RAISE

SELECT * FROM tabl1

pk  | a    | b    | c   
:-- | :--- | :--- | :---
p1  | a1   | b11  | c111
p2  | a1   | b11  | c112
p3  | a1   | b12  | c121
p4  | a1   | b12  | c122
p5  | a2   | b21  | c211
p6  | a2   | b21  | c212
p7  | a2   | b22  | c221
p8  | a2   | b22  | c222
p9  | null | b31  | c311
p10 | a3   | null | c312
p11 | a3   | b31  | null
p12 | null | null | c314
p13 | null | b31  | null
p14 | a3   | null | null

db<>fiddle here