PostgreSQL重复键值在冲突时使用时违反唯一约束插入时不执行任何操作

表def：

CREATE SEQUENCE IF NOT EXISTS lazy_product_stock_id_seq;
CREATE TABLE "public"."lazy_product_stock" (
    "id" int4 NOT NULL DEFAULT nextval('lazy_product_stock_id_seq'::regclass),
    "product_id" int4,
    "hold" int4 DEFAULT 0,
    "quantity" int4 DEFAULT 0,
    "warehouse_id" int4,
    PRIMARY KEY ("id")
);
CREATE UNIQUE INDEX lazy_product_stock_pkey ON public.lazy_product_stock USING btree (id)
CREATE INDEX lazy_product_stock_product_id_idx ON public.lazy_product_stock USING btree (product_id)
CREATE INDEX lazy_product_stock_warehouse_id_idx ON public.lazy_product_stock USING btree (warehouse_id)
CREATE UNIQUE INDEX CONCURRENTLY "lazy_product_stock_comb_idx2" ON "public"."lazy_product_stock" USING BTREE ("product_id","warehouse_id");

我有一个将新行插入数据库的函数：

CREATE OR REPLACE FUNCTION sp_lazystock_i_f(_product_id int4, site_id int4) RETURNS VOID AS $$
declare
warehouse record;
BEGIN
FOR warehouse IN select id from warehouse where siteid = site_id LOOP
    insert into lazy_product_stock (product_id, warehouse_id) VALUES (_product_id, warehouse.id) ON CONFLICT (product_id,warehouse_id) DO NOTHING;
END LOOP;
RETURN;
END;
$$ LANGUAGE plpgsql;

它失败了

duplicate key value violates unique constraint "lazy_product_stock_comb_idx2"

索引

CREATE UNIQUE INDEX CONCURRENTLY "lazy_product_stock_comb_idx2" ON "public"."lazy_product_stock" USING BTREE ("product_id","warehouse_id");

尽管使用重复的值单独运行相关的insert，但它没有问题

insert into lazy_product_stock (product_id, warehouse_id) VALUES (123, 1234) ON CONFLICT (product_id,warehouse_id) DO NOTHING;

Query 1 OK: INSERT 0 0, 0 rows affected

我不明白函数和单个语句之间的区别是什么

这也没有问题：

do $$
declare
warehouse record;
begin
    FOR warehouse IN select id from warehouse where siteid = 123 LOOP
    insert into lazy_product_stock (product_id, warehouse_id) VALUES (12345, warehouse.id) ON CONFLICT (product_id,warehouse_id) DO NOTHING;
END LOOP;
end; $$ LANGUAGE plpgsql;

同样的错误

insert into lazy_product_stock (product_id, warehouse_id)
    select _product_id, warehouse.id
    from warehouse where siteid = site_id
    on conflict (product_id, warehouse_id) do nothing;

---

我正在使用Postgresql 12.3，为什么要为此使用循环？那么：

insert into lazy_product_stock (product_id, warehouse_id)
    select _product_id, w.id
    from warehouse where siteid = site_id
    on conflict (product_id, warehouse_id) do nothing;

---

如果不指定目标，它是否有效

insert into lazy_product_stock (product_id, warehouse_id)
select _product_id, warehouse.id
from warehouse 
where siteid = site_id
on conflict do nothing;

---

什么都不做

是唯一一个不需要指定目标的冲突操作，这使得此解决方案在这里成为可能。

似乎我有两个这样的函数在导致问题的相同逻辑上触发。谢谢大家的帮助。

使用插入重复键值时出现相同错误，违反了唯一约束“lazy\u product\u stock\u comb\u idx2”编辑您的问题并显示表的定义。我怀疑您还有其他唯一的约束。更新了问题以包含表定义无关，但是：循环完全没有必要，只会使事情变得非常缓慢。我无法重现：不，它失败，出现相同的约束错误错误：重复键值违反唯一约束“lazy\u product\u stock\u comb\u idx2”