Postgres - Distinct + jsonb_object_keys 给出了在不能接受集合的上下文中调用的错误集合值函数

Question

我正在尝试运行以下查询

SELECT 
DISTINCT jsonb_object_keys("classif_ai_fileregion"."defects") AS "defect_ids" 
FROM "classif_ai_fileregion" 
WHERE 
("classif_ai_fileregion"."file_id" IN 
  (SELECT U0."id" FROM "classif_ai_file" U0 ) 
AND 
"classif_ai_fileregion"."ml_model_id" IN 
  (SELECT U0."id" FROM "classif_ai_mlmodel" U0 WHERE U0."id" IN (2)))

我收到以下错误

SELECT DISTINCT jsonb_object_keys("classif_ai_fileregion"."defects") AS "defect_ids" FROM "classif_ai_fileregion" 
WHERE 
("classif_ai_fileregion"."fi...

ERROR: set-valued function called in context that cannot accept a set
  Position: 17

SELECT DISTINCT jsonb_object_keys("classif_ai_fileregion"."defects") AS "defect_ids" FROM "classif_ai_fileregion" 
                ^

现在，我尝试了上述查询的一些修改版本，它们似乎都可以工作。从上面的查询中，我将 U0."id" IN (2) 替换为 U0."id" IN (1,2) 或 U0."id" IN (1) 并且查询有效。或者我删除 "classif_ai_fileregion"."file_id" IN (SELECT U0."id" FROM "classif_ai_file" U0 ) AND 并且它可以工作。

我无法理解为什么它适用于这些修改以及为什么它不适用于原始查询。

为便于理解查询，以下是表格。

1. classif_ai_file

   • 身份证

2. classif_ai_mlmodel

   • 身份证

3. classif_ai_fileregion

   • 身份证
   • file_id
   • ml_model_id
   • 缺陷 - jsonb 类型

缺陷列的格式为 {1: {}, 2: {}}

注意：我在这一行下面写了一些类似的查询，它们可以完成一些工作，而问题中的初始查询却没有。

SELECT 
DISTINCT jsonb_object_keys("classif_ai_fileregion"."defects") AS "defect_ids" 
FROM "classif_ai_fileregion" 
WHERE 
("classif_ai_fileregion"."file_id" IN 
  (SELECT U0."id" FROM "classif_ai_file" U0 ) 
AND 
"classif_ai_fileregion"."ml_model_id" IN 
  (SELECT U0."id" FROM "classif_ai_mlmodel" U0 WHERE U0."id" IN (1)))

您可以注意到我只是在此查询中将 2 更改为 1。

SELECT 
DISTINCT jsonb_object_keys("classif_ai_fileregion"."defects") AS "defect_ids" 
FROM "classif_ai_fileregion" 
WHERE 
("classif_ai_fileregion"."file_id" IN 
  (SELECT U0."id" FROM "classif_ai_file" U0 ) 
AND 
"classif_ai_fileregion"."ml_model_id" IN 
  (SELECT U0."id" FROM "classif_ai_mlmodel" U0 WHERE U0."id" IN (1,2)))

您可以注意到我只是在此查询中将 2 更改为 1,2。

SELECT 
DISTINCT jsonb_object_keys("classif_ai_fileregion"."defects") AS "defect_ids" 
FROM "classif_ai_fileregion" 
WHERE  
"classif_ai_fileregion"."ml_model_id" IN 
  (SELECT U0."id" FROM "classif_ai_mlmodel" U0 WHERE U0."id" IN (2)))

我在这个查询中删除了一个额外的 where 子句。

SELECT 
DISTINCT jsonb_object_keys("classif_ai_fileregion"."defects") AS "defect_ids" 
FROM "classif_ai_fileregion" 
WHERE 
("classif_ai_fileregion"."file_id" IN 
  (SELECT U0."id" FROM "classif_ai_file" U0 ) 
AND 
"classif_ai_fileregion"."ml_model_id" IN 
   (2))

我只是在这里简化了 ml_model_id 的 where 子句

Answer 1

将函数移至 FROM 子句：

SELECT DISTINCT df.defect_id
FROM "classif_ai_fileregion" 
  cross join lateral jsonb_object_keys(classif_ai_fileregion.defects) AS df(defect_id)
WHERE ("classif_ai_fileregion"."file_id" IN (SELECT U0."id" FROM "classif_ai_file" U0 ) 
  AND "classif_ai_fileregion"."ml_model_id" IN (SELECT U0."id" 
                                                FROM "classif_ai_mlmodel" U0 
                                                WHERE U0."id" IN (2)))

请注意，您的第二个 WHERE 条件可以简化为

AND "classif_ai_fileregion"."ml_model_id" IN (2)

Answer 2

jsonb_object_keys() 返回一组记录，这意味着它返回多行。在这种情况下，SELECT 列表中不允许这样做。

在不知道您的数据的情况下，我猜您的更改以这种方式过滤了您的数据集，只剩下一条记录。当且仅当返回不超过一条记录时，SELECT 列表中才允许设置返回函数。

因此，您应该尝试将函数移入FROM 列表：

SELECT DISTINCT 
    *
FROM "classif_ai_fileregion",
    jsonb_object_keys("classif_ai_fileregion"."defects") AS "defect_ids"  
WHERE ...

逗号创建了一个CROSS JOIN LATERAL，它表示该函数会分别为每条记录调用，这应该与您最初的预期输出相同。