pivot 表 pandas 错误 - 值应该是“Timedelta”、“NaT”或这些的数组。 取而代之的是“int”

Question

while trying to pivot a table I get an error I don't understand how to fix.

My code is:

import numpy as np
import pandas as pd

df1=pd.read_csv(r'C:\Users\Documents\Python\Data.csv')
df_com = df1.groupby(['CommentOwner','DiscussionId'])
y=df_com.nunique()
y=y.reset_index()
p=y.pivot(index="CommentOwner", columns="DiscussionId", values=['CommentOwner','DiscussionId','CommentCreation_min','CommentCreation_max','CommentCreation_count','AnswerId']).fillna(0)

I used reset_index() so I can use the columns 'CommentOwner','DiscussionId' after they were removed during the group by.

when I run this code I get this mistake:

TypeError: value should be a 'Timedelta', 'NaT', or array of those. Got 'int' instead.

when I try this code it does works:

import numpy as np
import pandas as pd

df1=pd.read_csv(r'C:\Users\Documents\Python\Data.csv')
df_com = df1.groupby(['CommentOwner','DiscussionId'])
y=df_com.nunique()
y.to_csv(r'C:\Users\Documents\Python\y.csv')
y_x=pd.read_csv(r'C:\Users\Documents\Python\y.csv')
p=y_x.pivot(index="CommentOwner", columns="DiscussionId", values=['CommentOwner','DiscussionId','CommentCreation_min','CommentCreation_max','CommentCreation_count','AnswerId']).fillna(0)

The code worked when I didn't use reset_index(), but rather saved the data frame as csv and then read it again.

I hope my question is clear. Any idea why this happens?

There must be a nicer way to do it without saving the output and reuploading it.

Thanks!

Answer 1

The problem is you fill null values with 0 for all columns even for datetime64 columns.

You should do something like:

p = (y.pivot(...)
      .fillna({my_datetime_col1: pd.NaT, my_datetime_col2: pd.NaT})
      .fillna(0))

The first fillna replace null values for all DatetimeIndex columns then the second one replace other missing values.

Answer 2

IIUC use:

df1=pd.read_csv(r'C:\Users\Documents\Python\Data.csv')
p = df1.groupby(['CommentOwner','DiscussionId']).nunique().unstack(fill_value=0)

Btw, your solution should working if remove 'CommentOwner','DiscussionId' from parameter value like:

p=y.pivot(index="CommentOwner", 
          columns="DiscussionId", 
          values=['CommentCreation_min','CommentCreation_max',
                  'CommentCreation_count','AnswerId']).fillna(0)