CREATE UNIQUE INDEX idx_table0_unique ON table0 ( field0, coalesce ( field1, 'INDEX_COLUMN_NULL' ) );
I created function for insert only unique record (part of code):
BEGIN INSERT INTO table0 ( field0, field1 ) VALUES ( p_field0, p_field1 ) RETURNING id INTO v_table0_id; EXCEPTION WHEN unique_violation THEN SELECT id FROM table0 WHERE field0 = p_field0 AND field1 = p_field1; END;
I use this function for add data to table and sometimes I getting false unique violation (in block above). Why?
if p_field1 is NULL, then the select cannot return any rows, but can still violate the constraint.
Also, you may have a race, where the row exists when the insert was attempted, but was gone by the time it tried to do the select.