一个选项在列元组的最大/最小值上使用唯一索引：

create unique index ingredient_to_ingredient_idx 
on ingredient_to_ingredient( 
    greatest(ingredient_a_id, ingredient_b_id),
    least(ingredient_a_id, ingredient_b_id)
);

Demo on DB Fiddle：

-- I want this to produce a constraint error
insert into ingredient_to_ingredient (ingredient_a_id, ingredient_b_id) values (2, 1); 

-- ERROR:  duplicate key value violates unique constraint "ingredient_to_ingredient_idx"
-- DETAIL:  Key (GREATEST(ingredient_a_id, ingredient_b_id), LEAST(ingredient_a_id, ingredient_b_id))=(2, 1) already exists.

另一种方法是排除约束：

alter table ingredient_to_ingredient
add constraint ingredient_to_ingredient_excl exclude using gist(
        least(ingredient_a_id, ingredient_b_id)    with =,
        greatest(ingredient_a_id, ingredient_b_id) with =
);

这将为错误的insert语句产生以下错误：

-- ERROR:  conflicting key value violates exclusion constraint "ingredient_to_ingredient_excl"
-- DETAIL:  Key (LEAST(ingredient_a_id, ingredient_b_id), GREATEST(ingredient_a_id, ingredient_b_id))=(1, 2) conflicts with existing key (LEAST(ingredient_a_id, ingredient_b_id), GREATEST(ingredient_a_id, ingredient_b_id))=(1, 2).

Fiddle