将字符串强制转换为结构数组，然后在另一个SQL表中查找值

我当前有一个字符串，它表示表中的结构列表。我想根据结构中元素的值在另一个表中查找值

例如，下面的car info结构是[spare，carType，carcolor]

╔═══════════════════════════╗
║          CarInfo          ║
╠═══════════════════════════╣
║ “[1,1,1]”                 ║
║ “[1,2,1] [1,1,2]”         ║
║ null                      ║
║ “[1,2,1] [1,1,2] [1,1,1]” ║
╚═══════════════════════════╝

我想查一下表格：

╔═══════════╦═══════════════╦═════════════╦═════════════════╦══╗
║ CarTypeId ║ CarTypeString ║ CarColourId ║ CarColourString ║  ║
╠═══════════╬═══════════════╬═════════════╬═════════════════╬══╣
║         1 ║ "Hyundai"     ║           1 ║ "Red"           ║  ║
║         1 ║ "Hyundai"     ║           2 ║ "Blue"          ║  ║
║         2 ║ "Toyota"      ║           1 ║ "Green"         ║  ║
║         2 ║ "Toyota"      ║           2 ║ "Yellow"        ║  ║
╚═══════════╩═══════════════╩═════════════╩═════════════════╩══╝

并得到以下结果：

╔═════════════════════════════════════════════════════╗
║                       CarInfo                       ║
╠═════════════════════════════════════════════════════╣
║ “[1,Hyundai,Red]”                                   ║
║ “[1,Toyota,Green] [1,Hyundai,Blue]”                 ║
║ null                                                ║
║ “[1,Toyota,Green] [1,Hyundai,Blue] [1,Hyundai,Red]” ║
╚═════════════════════════════════════════════════════╝

---

我发现我可以使用someString.split（CarInfo，''）将字符串拆分为数组，但此后我不确定如何执行结构转换或之后的“循环”左连接。

下面是针对BigQuery标准SQL的

#standardSQL
SELECT STRING_AGG('[' || spare || ',' || carTypeString || ',' || carColourString || ']', ' ') AS CarInfo
FROM `project.dataset.cars` t
LEFT JOIN UNNEST(SPLIT(CarInfo, ' ')) info,
UNNEST([STRUCT(
  SPLIT(TRIM(info, '[]'))[OFFSET(0)] AS spare, 
  CAST(SPLIT(TRIM(info, '[]'))[OFFSET(1)] AS INT64) AS carTypeId, 
  CAST(SPLIT(TRIM(info, '[]'))[OFFSET(2)] AS INT64) AS carColourId
)])
LEFT JOIN `project.dataset.lookup` l
USING(carTypeId, carColourId)
GROUP BY FORMAT('%t', t)   

如果要应用于您问题中的样本数据，请参见下面的示例

#standardSQL
WITH `project.dataset.cars` AS (
  SELECT '[1,1,1]' CarInfo UNION ALL
  SELECT '[1,2,1] [1,1,2]' UNION ALL
  SELECT NULL UNION ALL
  SELECT '[1,2,1] [1,1,2] [1,1,1]'
), `project.dataset.lookup` AS (
  SELECT 1 CarTypeId, 'Hyundai' CarTypeString, 1 CarColourId, 'Red' CarColourString UNION ALL
  SELECT 1, 'Hyundai', 2, 'Blue' UNION ALL
  SELECT 2, 'Toyota', 1, 'Green' UNION ALL
  SELECT 2, 'Toyota', 2, 'Yellow'
)
SELECT STRING_AGG('[' || spare || ',' || carTypeString || ',' || carColourString || ']', ' ') AS CarInfo
FROM `project.dataset.cars` t
LEFT JOIN UNNEST(SPLIT(CarInfo, ' ')) info,
UNNEST([STRUCT(
  SPLIT(TRIM(info, '[]'))[OFFSET(0)] AS spare, 
  CAST(SPLIT(TRIM(info, '[]'))[OFFSET(1)] AS INT64) AS carTypeId, 
  CAST(SPLIT(TRIM(info, '[]'))[OFFSET(2)] AS INT64) AS carColourId
)])
LEFT JOIN `project.dataset.lookup` l
USING(carTypeId, carColourId)
GROUP BY FORMAT('%t', t)      

输出为

Row CarInfo  
1   [1,Hyundai,Red]  
2   [1,Toyota,Green] [1,Hyundai,Blue]    
3   null     
4   [1,Toyota,Green] [1,Hyundai,Blue] [1,Hyundai,Red]    

---

澄清：在carInfo表中，列的具体数据类型是什么？拥有精确的表模式将有助于理解您的用例！它是一个字符串。所以双引号是字符串的一部分，或者您将它们添加到了强调符号中，它们是字符串？另外，在查找表

“现代”

-中，双引号实际上是值的一部分，还是仅仅是表示它是字符串的一种方式？在这两种情况下，我都将它们添加到了强调符号中，它们是字符串，对不起！非常感谢，太棒了。我还有一个简短的问题。如果我在CarInfo表中有另一列要开始，并且希望将其保留在末尾。使用GroupBy时，保存它的最佳方式是什么？我在其他地方看到其他人把Max（additionalCol）作为additionalCol来做？这正是应该做的：o）只需将它添加到最外层的选择中