RedShift：不支持领导者上的 Unnest 子查询结果

Question

我尝试在 RedShift 中解析 JSON。

我在“输入”列中的字符串是：

[{"desc": "How many people does the video contain?", "name": "Number of People", "type": "dropdown", "values": ["", "Only 1", "2-3", "3+"]}, {"desc": "What is the camera position?", "name": "Movement", "type": "dropdown", "values": ["", "Fixed position", "Moving"]}, {"desc": "From which angle did you shoot the video?", "name": "Shoot Angle", "type": "dropdown", "values": ["", "Frontal recording", "Tight angle: 10-40 degree", "Wide angle: 40-70 degree"]}, {"desc": "From which distance did you shoot the video?", "name": "Distance", "type": "dropdown", "values": ["", "Near/Selfie", "Mid (3-6 ft)", "Far (>6 ft)"]}, {"desc": "What is the video lighting direction?", "name": "Lighting Direction", "type": "dropdown", "values": ["", "Front lit", "Side lit", "Back lit"]}, {"desc": "What is the video background?", "name": "Background", "type": "dropdown", "values": ["", "Outdoors", "In office", "At home", "Plain background"]}, {"desc": "What is the topic in your speech?", "name": "Topic", "type": "dropdown", "values": ["", "Arts and Media", "Business", "Education", "Entertainment", "Food/Eating", "Nutrition", "Healthcare ", "High School Life", "Mental Health", "News", "Technology", "Morals and Ethics", "Phones and Apps", "Sports", "Science"]}]

我的任务是：“JSON 中的每个值、名称、desc 都需要在表中存储为一行”。

示例： id: 1, desc: "视频包含多少人？" id: 2, desc: "相机位置是什么？" 等等

我使用查询：

SELECT c.*, d.desc, d.name, d.values FROM source.table AS c, c.inputs AS d; 

得到一个错误：不允许在“输入”列上导航，因为它不是超级类型

并查询：

SELECT c.*, d.desc, d.name, d.values FROM source.table AS c, JSON_PARSE(c.inputs) AS d;

给我另一个错误：“FROM 中的函数表达式可能不引用相同查询级别的其他关系”

但是当我将测试 JSON 创建为：

CREATE TABLE test_parse_json_super
(
  id smallint,
  details super
);
 
INSERT INTO test_parse_json_super VALUES(1, JSON_PARSE('[{"desc": "How many people does the video contain?", "name": "Number of People", "type": "dropdown", "values": ["", "Only 1", "2-3", "3+"]}, {"desc": "What is the camera position?", "name": "Movement", "type": "dropdown", "values": ["", "Fixed position", "Moving"]}, {"desc": "From which angle did you shoot the video?", "name": "Shoot Angle", "type": "dropdown", "values": ["", "Frontal recording", "Tight angle: 10-40 degree", "Wide angle: 40-70 degree"]}, {"desc": "From which distance did you shoot the video?", "name": "Distance", "type": "dropdown", "values": ["", "Near/Selfie", "Mid (3-6 ft)", "Far (>6 ft)"]}, {"desc": "What is the video lighting direction?", "name": "Lighting Direction", "type": "dropdown", "values": ["", "Front lit", "Side lit", "Back lit"]}, {"desc": "What is the video background?", "name": "Background", "type": "dropdown", "values": ["", "Outdoors", "In office", "At home", "Plain background"]}, {"desc": "What is the topic in your speech?", "name": "Topic", "type": "dropdown", "values": ["", "Arts and Media", "Business", "Education", "Entertainment", "Food/Eating", "Nutrition", "Healthcare ", "High School Life", "Mental Health", "News", "Technology", "Morals and Ethics", "Phones and Apps", "Sports", "Science"]}]'));
                                               

并使用来自官方 RedShift 文档的查询 "SELECT c.*, d.desc, d.name, d.values FROM test_parse_json_super AS c, c.details AS d;" - 它工作正常，来自 JSON 的所有数据都解析到每一行，并且 JSON 是正确的。

如何修复查询以使用我的真实数据？

谢谢。

Answer 1

您似乎对子选择的 SUPER 数据类型的概念感到困惑。您不能将字符串强制转换为 super（使用 JSON_PARSE），然后在查询的同一级别将其用作 FROM 子句的源。我不完全了解您的情况，但我认为这样的事情应该让您更接近：

SELECT c.*, d.desc, d.name, d.values 
FROM (
    SELECT *, JSON_PARSE(inputs) AS inputs_super
    FROM source.table 
    ) AS c, 
JSON_PARSE(c.inputs_super) AS d
;

（这是对显示结构的即兴响应，因此请原谅任何语法问题）

Answer 2

会更正确：

SELECT c.*, d.desc, d.name, d.values 
FROM (
    SELECT id, created, JSON_PARSE(inputs) AS inputs_super
    FROM source.table
    WHERE prompttype = 'input'
    ) AS c, 
c.inputs_super AS d                                                      
;