为什么在使用 pandas 规范化 JSON 时，访问嵌套元数据会得到 NaN？

huangapple
go评论114阅读模式
英文:

Why reaching nested meta gives NaN when normalizing a json with pandas?

问题

以下是您要翻译的内容：

  1. 我的输入是一个Python字典（类似JSON）：
  3. d = {
  4.     "type": "type1",
  5.     "details": {
  6.         "name": "foo",
  7.         "date": {
  8.             "timestamp": "01/02/2023 21:42:44",
  9.             "components": {
  10.                 "day": 2,
  11.                 "month": 1,
  12.                 "year": 2023,
  13.                 "time": "21:42:44"
  14.             }
  15.         }
  16.     },
  17.     "infos": {
  18.         "records": [
  19.             {
  20.                 "field1": "qux",
  21.                 "field2": "baz",
  22.             }
  23.         ],
  24.         "class": "P"
  25.     }
  26. }
  28. 我使用以下代码：
  30. df = pd.json_normalize(
  31.     d,
  32.     record_path=["infos", "records"],
  33.     meta=[
  34.         "type",
  35.         ["details", "date", "timestamp"],
  36.         ["details", "date", "components", "year"],
  37.         ["infos", "class"]
  38.     ],
  39.     errors="ignore"
  40. )
  42. 这给了我以下输出：
  44. field1 field2   type details.date.timestamp details.date.components.year infos.class
  45. 0    qux    baz  type1                    NaN                          NaN           P
  47. 但我期望得到这个输出：
  49. field1 field2   type details.date.timestamp details.date.components.year infos.class
  50. 0    qux    baz  type1    01/02/2023 21:42:44                         2023           P
  52. 老实说，我对`meta`参数感到非常困惑！我不知道我做错了什么...
  54. 您能解释一下它的逻辑吗？
英文:

My input is a Python dictionnary (json-like) :

  1. d = {
  2.     "type": "type1",
  3.     "details": {
  4.         "name": "foo",
  5.         "date": {
  6.             "timestamp": "01/02/2023 21:42:44",
  7.             "components": {
  8.                 "day": 2,
  9.                 "month": 1,
  10.                 "year": 2023,
  11.                 "time": "21:42:44"
  12.             }
  13.         }
  14.     },
  15.     "infos": {
  16.         "records": [
  17.             {
  18.                 "field1": "qux",
  19.                 "field2": "baz",
  20.             }
  21.         ],
  22.         "class": "P"
  23.     }
  24. }

I'm using the code below :

  1. df = pd.json_normalize(
  2.     d,
  3.     record_path=["infos", "records"],
  4.     meta=[
  5.         "type",
  6.         ["details", "date", "timestamp"],
  7.         ["details", "date", "components", "year"],
  8.         ["infos", "class"]
  9.     ],
  10.     errors="ignore"
  11. )

Which gives me this output :

  1.   field1 field2   type details.date.timestamp details.date.components.year infos.class
  2. 0    qux    baz  type1                    NaN                          NaN           P

But I'm expecting this one :

  1.   field1 field2   type details.date.timestamp details.date.components.year infos.class
  2. 0    qux    baz  type1    01/02/2023 21:42:44                         2023           P

To be honest, I'm going crazy with the meta parameter! I ignore what I'm doing wrong..

Can you explain its logic, please ?

答案1

得分: 2

  1. 我认为你应该在`record_path=`中额外添加`[]`：
  3. ```py
  4. df = pd.json_normalize(
  5.     d,
  6.     record_path=[["infos", "records"]],  # <-- 在这里加上 []
  7.     meta=[
  8.         "type",
  9.         ["details", "date", "timestamp"],
  10.         ["details", "date", "components", "year"],
  11.         ["infos", "class"],
  12.     ],
  13.     errors="ignore",
  14. )
  16. print(df)

打印：

  1.   field1 field2   type details.date.timestamp details.date.components.year infos.class
  2. 0    qux    baz  type1    01/02/2023 21:42:44                         2023           P
  2. <details>
  3. <summary>英文:</summary>
  5. I think you should put extra `[]` in `record_path=`:
  7. ```py
  8. df = pd.json_normalize(
  9.     d,
  10.     record_path=[[&quot;infos&quot;, &quot;records&quot;]],  # &lt;-- put [] here
  11.     meta=[
  12.         &quot;type&quot;,
  13.         [&quot;details&quot;, &quot;date&quot;, &quot;timestamp&quot;],
  14.         [&quot;details&quot;, &quot;date&quot;, &quot;components&quot;, &quot;year&quot;],
  15.         [&quot;infos&quot;, &quot;class&quot;],
  16.     ],
  17.     errors=&quot;ignore&quot;,
  18. )
  20. print(df)

Prints:

  1.   field1 field2   type details.date.timestamp details.date.components.year infos.class
  2. 0    qux    baz  type1    01/02/2023 21:42:44                         2023           P

huangapple
  • 本文由 发表于 2023年7月24日 15:05:22
  • 转载请务必保留本文链接：https://go.coder-hub.com/76752107.html
  • json
  • pandas
  • python
匿名

发表评论

匿名网友

:?: :razz: :sad: :evil: :!: :smile: :oops: :grin: :eek: :shock: :???: :cool: :lol: :mad: :twisted: :roll: :wink: :idea: :arrow: :neutral: :cry: :mrgreen:

确定