bigquery提取的信息来自存储为字符串数据类型的JSON字符串

Question

MyBQ表由BQ模块（Load_table_from_dataframe）填充到一个表中，并将其作为字符串类型（包括`````````````````````````````````````````````````````````````''）'

```json{   
"number": 123,   
"date": "2020-11-01",   
"last_date": "2020-12-30"
 }```

我正在尝试使用多种方法提取键，但我的零件是无效的，因此我必须强行将``在空白中替换为空白并提取信息。我认为这个论坛可以比我更好地帮助我。 below代码有效，但我认为这不是从JSON中提取的正确方法

SELECT

REPLACE(json_extract(json_extract((replace(REPLACE(FIELD_NAME,"```json",""),"```","")),"$"),"$[number]"),"\"","") as number

FROM
Table

您可以使用以下查询：

with table AS (
select json_object(
    'number', 123,
    'date', '2020-11-01',
    'last_date','2020-12-30',
    'array', ['0','1','2']
) as json
)
select 
 json_value(json, '$.number') as number -- To get the number
 , json_value(json, '$.date') as date_ -- To get the date
 , json_value_array(json, '$.array') as array_ -- To get the array
 , json_value(json, '$.object.foo') as foo -- To get the value of foo
 , json_value(json, '$.object.bar') as bar -- To get the value of bar
from table

Answer 1

json_value

（），对于数组，您可以使用

JSON_VALUE_ARRAY

（）。如果您有任何高级案例，只需阅读示例并稍微播放。
    

今天使用ml.generate_text时，我遇到了类似的问题。 提取我使用的响应

WITH

myjson AS (
SELECT '```json\n{"field_a": "a string", "field_b": {"a_secondary": 0, "b_secondary": 1}}\n```' AS a_weird_json_string
),

extract_json AS (
  SELECT
    *,
    REGEXP_EXTRACT(a_weird_json_string, r'({.*})') AS json_content
  FROM myjson
  
)

SELECT
  a_weird_json_string,
  JSON_VALUE(json_content, '$.field_a') AS field_a_preprocessed,
  JSON_VALUE(json_content, '$.field_b.a_secondary') AS field_b_a_secondary_preprocessed,
  -- On original string we get Null
  JSON_VALUE(a_weird_json_string, '$.field_a') AS field_a_value,
  JSON_EXTRACT(a_weird_json_string, '$.field_a') AS field_a_extract,
  -- PARSE_JSON(a_weird_json_string) AS parsed_json -- Yield an Invalid input: syntax error while parsing value - invalid literal; last read: '`'; error in PARSE_JSON expression
  
FROM extract_json

不幸的是，在原始字符串上均以json

Answer 2