熊猫，在保留订单的同时进行分类

Question

I've received a CSV file that is a combination of several other csv files.

It has a datetime index (in the format of '2017-01-16' , year, month, day) However, two problems arise.

1. The combination was not done in order.

     Date string number (different) 1 2017-01-16 stringvalue 90 2 2017-01-16 stringvalue 912 3 2017-01-16 stringvalue 29 4 2017-01-17 stringvalue 883 5 2017-01-17 stringvalue 223 6 2017-01-17 stringvalue 211 (...) 230 2015-04-30 stringvalue 908 231 2015-04-29 stringvalue 28 232 2015-04-29 stringvalue 9 233 2015-04-30 stringvalue 98 234 2015-04-30 stringvalue 909 (...) 450 2017-03-30 stringvalue 348 

2. No time has been provided (the actual day is the smallest number, yet each days holds around 10 values, that need to be kept in order)

I resolved the first problem by peforming

    df = df.reset_index()
    df = df.sort_values('Date')
    df = df.set_index('Date')

This correctly orders the index, but messes up the ordering within each day. Is there a way to sort the dates, but keep the original order within the days intact?

Answer 1

By using a new para and prefix the original order :

df['G']=df.groupby(level='Date').cumcount()
df
Out[125]: 
                 string  number  G
Date                              
2017-01-16  stringvalue      90  0
2017-01-16  stringvalue     912  1
2017-01-16  stringvalue      29  2
2017-01-17  stringvalue     883  0
2017-01-17  stringvalue     223  1
2017-01-17  stringvalue     211  2
2015-04-30  stringvalue     908  0
2017-03-30  stringvalue     348  0

df.sort_values('G').sort_index().drop('G',1)
Out[124]: 
                 string  number
Date                           
2015-04-30  stringvalue     908
2017-01-16  stringvalue      90
2017-01-16  stringvalue     912
2017-01-16  stringvalue      29
2017-01-17  stringvalue     883
2017-01-17  stringvalue     223
2017-01-17  stringvalue     211
2017-03-30  stringvalue     348

Answer 2

df['Date'] = pd.to_datetime(df.Date)

first convert the Date column to datetime type if needed

df = df.reset_index().sort_values(by=['Date', 'index']).drop(['index'], axis=1)

This will reset the index, creating a temporary column called index . Then sort using both the Date & index columns, and finally drop the index column. leaving the data frame sorted by Date and the order in which they appeared in the original CSV file.