Pandas DataFrame 中带有类的数据时间索引错误

Question

我有一个带类和不带类的 Pandas DataFrame。如果有类，我有错误的数据时间索引，但如果没有类，那就可以了。请帮忙。

from datetime import datetime
import pandas as pd


class Records:
    def __init__(self):
        columns = ['moment', 'val']

        # Create an empty DataFrame with columns
        self.dfv = pd.DataFrame(columns=columns)

    def add(self, moment, val):

        # Record
        rec = {'moment': moment, 'val': val}

        # Convert 'moment' to datetime
        rec['moment'] = pd.to_datetime(rec['moment'], format='%Y-%m-%d %H:%M:%S')

        # Add the record to DataFrame
        self.dfv = self.dfv.append(rec, ignore_index=True)

        self.dfv['moment'] = pd.to_datetime(self.dfv['moment'])
        self.dfv.set_index('moment', inplace=True)

        return self.dfv

if __name__ == '__main__':
    print('=============================Class records===================')
    r = Records()

    df = r.add('2023-11-01 10:00:00', 100.0)
    print(df)
    df = r.add('2023-11-01 11:00:00', 120.0)

    # Print the updated DataFrame
    print(df)
    df.info()

    print('=============================Single records==================')
    # Test first column datatime as an index

    columns = ['moment', 'val']
    dfv = pd.DataFrame(columns=columns)
    dfv['moment'] = pd.to_datetime(dfv['moment'])
    dfv.set_index('moment', inplace=True)

    rec = {'moment': '2023-11-01 10:00:00', 'val': 100.0}

    # Convert 'moment' to datetime
    rec['moment'] = pd.to_datetime(rec['moment'], format='%Y-%m-%d %H:%M:%S')

    # Add the record to DataFrame
    dfv = dfv.append(rec, ignore_index=True)

    # Set 'moment' as the index
    rec2 = {'moment': '2023-11-01 11:00:00', 'val': 120.0}
    dfv = dfv.append(rec2, ignore_index=True)

    # Set 'moment' as the index
    dfv.set_index('moment', inplace=True)

    print(dfv)
    dfv.info()
    exit(0)

输出

=============================Class records===================
                       val
moment                    
2023-11-01 10:00:00  100.0
                       val
moment                    
NaT                  100.0
2023-11-01 11:00:00  120.0
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 2 entries, NaT to 2023-11-01 11:00:00
Data columns (total 1 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   val     2 non-null      float64
dtypes: float64(1)
memory usage: 32.0 bytes
=============================Single records==================
                       val
moment                    
2023-11-01 10:00:00  100.0
2023-11-01 11:00:00  120.0
<class 'pandas.core.frame.DataFrame'>
Index: 2 entries, 2023-11-01 10:00:00 to 2023-11-01 11:00:00
Data columns (total 1 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   val     2 non-null      float64
dtypes: float64(1)
memory usage: 32.0+ bytes

因此，在没有类的情况下，我完全有我想要得到的东西，但如果我使用类（并且我需要使用类来达到我的目的），我将 NaT 作为第一个记录索引值。请告知如何在课堂上获得正确的代码。谢谢。

Answer 1

您可以使用以下方法修复代码：

from datetime import datetime
import pandas as pd


class Records:
    def __init__(self):
        # Create a correctly typed dataframe
        self.dfv = pd.DataFrame(columns=pd.Series(name='val', dtype='float'),
                                index=pd.DatetimeIndex([], name='moment'))

    def add(self, moment, val):
        # Same remark as previous
        moment = pd.to_datetime(moment, format='%Y-%m-%d %H:%M:%S')
        rec = pd.DataFrame({'val': val}, index=[moment])
        # Safe concat, works with most recent versions of Pandas
        self.dfv = pd.concat([self.dfv, rec], axis=0)
        return self.dfv


print('=============================Class records===================')
r = Records()

df = r.add('2023-11-01 10:00:00', 100.0)
print(df)

df = r.add('2023-11-01 11:00:00', 120.0)
print(df)

df.info()

输出：

=============================Class records===================
                       val
2023-11-01 10:00:00  100.0
                       val
2023-11-01 10:00:00  100.0
2023-11-01 11:00:00  120.0
<class 'pandas.core.frame.DataFrame'>
DatetimeIndex: 2 entries, 2023-11-01 10:00:00 to 2023-11-01 11:00:00
Data columns (total 1 columns):
 #   Column  Non-Null Count  Dtype  
---  ------  --------------  -----  
 0   val     2 non-null      float64
dtypes: float64(1)
memory usage: 32.0 bytes

Answer 2

问题是，在您第一次调用

df

后，您的

r.add(...)

的形状有所不同。让我们在初始化后和第一次调用后打印

df

来清楚地看到这一点：

if __name__ == '__main__':
    print('=============================Class records===================')
    r = Records()

    df = r.dfv
    print("""df after initialization:\n""")
    print(df)
    
    df = r.add('2023-11-01 10:00:00', 100.0)
    print("""df after first `add` call:\n""")
    print(df)

=============================Class records===================
df after initialization:

Empty DataFrame
Columns: [moment, val]
Index: []

df after first `add` call:

                       val
moment                    
2023-11-01 10:00:00  100.0
C:\Users\bobco\AppData\Local\Temp\ipykernel_12940\2699656422.py:20: FutureWarning: The frame.append method is deprecated and will be removed from pandas in a future version. Use pandas.concat instead.
  self.dfv = self.dfv.append(rec, ignore_index=True)

因此，您的第一个调用正在运行，因为您在有 2 列 (df.append

) 的 

df 上使用 ['moment', 'val']。然而，对于所有后续调用，您的

df

只剩下 1 列 (

['val']

)，

moment

已成为

index

。

修复方法是使用

df

作为索引来初始化

'moment'

，然后在添加数据之前以相同的形状添加数据。请注意上面输出中的

FutureWarning

：由于自 1.4.0 版本以来

df.append

已被弃用，所以让我们使用

pd.concat

代替。

import pandas as pd

class Records:
    def __init__(self):
        columns = ['moment', 'val']

        # Create an empty DataFrame with a `DatetimeIndex`
        self.dfv = pd.DataFrame(columns=['val'],
                                index=pd.DatetimeIndex([]))

    def add(self, moment, val):

        # Record (changing to setup of your dict)
        rec = {'val': {pd.to_datetime(moment): val}}

        # Add the record to DataFrame (using `pd.concat`)
        self.dfv = pd.concat([self.dfv, pd.DataFrame(rec)])
        
        return self.dfv

if __name__ == '__main__':
    print('=============================Class records===================')
    r = Records()

    df = r.dfv
    print("""df after initialization:\n""")
    print(df)
    
    df = r.add('2023-11-01 10:00:00', 100.0)
    print("""\ndf after first `add` call:\n""")
    print(df)
    
    df = r.add('2023-11-01 11:00:00', 120.0)
    print("""\ndf after second `add` call:\n""")
    print(df)

=============================Class records===================
df after initialization:

Empty DataFrame
Columns: [val]
Index: []

df after first `add` call:

                       val
2023-11-01 10:00:00  100.0

df after second `add` call:

                       val
2023-11-01 10:00:00  100.0
2023-11-01 11:00:00  120.0