Question

I wonder how to solve this problem.

I got two tables

Table1 (t1_prim, t1_int)
with t1_prim as primary key and t1_int not null.

Table2 (t2_prim1, t2prim2, t2_int)
with t2_prim1 and t2_prim2 as primary key and foreign key constraint t2_prim1 references t1_prim.

how can I define a constraint that makes sure, that nobody enters an integer in t2_int that is bigger than the corresponding entry in t1_int?

I tried it like this (doesn’t work because you can’t enter subqueries in check constraint):

CREATE TABLE table2 
  (t2_prim1 TEXT,
  t2_prim2 INTEGER,
  t2_int INTEGER NOT NULL,
  PRIMARY KEY (t2_prim1, t2_prim2),
  FOREIGN KEY (t2_prim1) REFERENCES table1(t1_prim),
  CHECK (t2_int2 <= (SELECT t1_int2 FROM table1 WHERE t1_int1=t2_int1)));

And I think there would be another problem if it would work like this. How do I check, that this constraint is still fullfilled when changing t1_int?

Answer 1

Finally found an answer to my question. You can solve the problem by using triggers. Here’s a working example:

CREATE TABLE table1
  (t1_prim TEXT PRIMARY KEY,
   t1_int INTEGER NOT NULL);

CREATE TABLE table2
  (t2_prim1 TEXT,
  t2_prim2 INTEGER,
  t2_int INTEGER NOT NULL,
  PRIMARY KEY (t2_prim1, t2_prim2),
  FOREIGN KEY (t2_prim1) REFERENCES table1(t1_prim));

Creating the triggers for updating and inserting:

CREATE TRIGGER t1_int_update_constraint
BEFORE UPDATE OF t1_int ON table1
  BEGIN
    SELECT CASE 
      WHEN new.t1_int < (SELECT max(t2_int) FROM table2 WHERE t2_prim1=old.t1_prim) 
        THEN (SELECT RAISE(ABORT, 
          'Input smaller than maximum of all values for t2_int in database!'))
      END;
  END;

CREATE TRIGGER t2_int_update_constraint
BEFORE UPDATE OF t2_int ON table2
  BEGIN
    SELECT CASE 
      WHEN new.t2_int > (SELECT t1_int FROM table1 WHERE t1_prim=old.t2_prim1) 
        THEN (SELECT RAISE(ABORT, 
         'Input bigger than value in t1_int!'))
      END;
    END;

CREATE TRIGGER t2_int_insert_constraint
BEFORE INSERT ON table2
  BEGIN
    SELECT CASE 
      WHEN new.t2_int > (SELECT t1_int FROM table1 WHERE t1_prim=new.t2_prim1) 
        THEN (SELECT RAISE(ABORT, 
         'Input bigger than value in t1_int!'))
      END;
  END;

If you have now the following tables:

table1

t1_prim     t1_int
----------  ----------
one         5         
two         10   

table2

t2_prim1    t2_prim2    t2_int    
----------  ----------  ----------
one         1           5         
one         2           4         
two         1           7         
two         2           5         
two         3           1         

You get this output:

UPDATE table1 SET t1_int=4 WHERE t1_prim='one';

Error: Input smaller than maximum of all values for t2_int in database!

UPDATE table1 SET t1_int=6 WHERE t1_prim='two';

Error: Input smaller than maximum of all values for t2_int in database!

UPDATE table2 SET t2_int=8 WHERE t2_prim1='one' AND t2_prim2=1;

Error: Input bigger than value in t1_int!

UPDATE table2 SET t2_int=11 WHERE t2_prim1='two' AND t2_prim2=2;

Error: Input bigger than value in t1_int!

INSERT INTO table2 VALUES ('one', 3, 6);

Error: Input bigger than value in t1_int!

While those work perfectly well:

INSERT INTO table2 VALUES ('one', 3, 6);
UPDATE table2 SET t2_int=1 WHERE t2_prim1='one' AND t2_prim2=1;
UPDATE table1 SET t1_int=8 WHERE t1_prim='two';