ndarray在工作脚本上不是JSON可序列化的TypeError (Tableau TabPy)

Question

我有使用测试数据的工作代码，如下所示：(示例数据匹配我以后将使用此脚本的输入类型和格式，包括日期作为附加‘Z’的字符串)

    from sklearn import linear_model
    import pandas as pd
    import numpy as np
     
     
    d = {'Start Sunday of FW': ['2022-03-02Z', '2022-03-03Z', '2022-03-04Z', '2022-03-05Z', '2022-03-06Z', '2022-03-01Z',
                  '2022-03-02Z', '2022-03-03Z', '2022-03-04Z', '2022-03-05Z', '2022-03-06Z', '2022-03-01Z'],
         'Store': [1111, 1111, 1111, 1111, 1111, 1111, 2222, 2222, 2222, 2222, 2222, 2222],
         'Sales': [3163, 4298, 2498, 4356, 4056, 3931, 3163, 4298, 2498, 4356, 4056, 1]}
     
    df = pd.DataFrame(data=d)
     
    def get_coef(input):
         def model1(df):
              y = df[['Sales']].values
              x = df[['Start Sunday of FW']].values
              return np.squeeze(linear_model.LinearRegression().fit(x,y).coef_)
     
         cnames = {'Store': 'Store', 0: 'Coef'}
     
         def prep_input(df):
              df['Start Sunday of FW'] = df['Start Sunday of FW'].astype('string').str.rstrip('Z').astype('datetime64[ns]')
              return df
     
         return pd.DataFrame(prep_input(input).groupby('Store').apply(model1)).reset_index().rename(columns=cnames)
     
    print(get_coef(df))

这段代码本身运行良好。

我安装了Tableau和TabPy，并根据指令正确设置它。

但是，当我试图从Prep在下面的代码块中运行版本时，Prep流失败了，我得到了错误：

```2022-04-06,13:08:58 [ERROR] (base_handler.py:base_handler:115): Responding with status=500, message="Error processing script", info="TypeError : Object of type ndarray is not JSON serializable"```

If I instead have it `print()` the current `return` line from `get_coef()`, return the `input` instead, and remove the `get_output_schema` function: the return is printed correctly, the output does flow to Tableau Prep in the live previews, **but** the error still raises and the flow still won't work, which is baffling.

```javascript

from sklearn import linear_model

import pandas as pd

import numpy as np

def get_coef(input):

     def model1(df):

          y = df[['Sales']].values

          x = df[['Start Sunday of FW']].values

          return np.squeeze(linear_model.LinearRegression().fit(x,y).coef_)

     cnames = {'Store': 'Store', 0: 'Coef'}

     def prep_input(df):

          df['Start Sunday of FW'] = df['Start Sunday of FW'].astype('string').str.rstrip('Z').astype('datetime64[ns]')

          return df

     return pd.DataFrame(prep_input(input).groupby('Store').apply(model1)).reset_index().rename(columns=cnames)

def get_output_schema():

     return pd.DataFrame({

          'Store' : prep_string(),

          'Coef' : prep_decimal()

     })

有人能帮我理解这个问题吗？我一开始对JSON序列化一无所知，所以像这这样的帖子对我没有多大帮助；我甚至无法评估相关性。

Answer 1

result = pd.DataFrame(prep_input(input).groupby('Store').apply(model1)).reset_index().rename(columns=cnames)
     result['Coef'] = result['Coef'].astype('double')

     return result

Coef是dtype对象，TabPy需要双倍。