CODEWITHSUNDEEP
pythonpython-3.xpandasdataframe
Evy555
Evy555

Reputation: 229

Sort Python DataFrame by Day of the Week

I'm trying to sort my dataframe by the day of the week in the index. I was able to create a workaround where I created an additional column that I manually assigned the correct order value to and then sorted. But there has to be a better way, especially when I'm dealing with potentially more date values. 

dict = {'Monday': Monday/MonCount, 'Tuesday': Tuesday/TueCount, 'Wednesday': Wednesday/WedCount, 'Thursday': Thursday/ThuCount, 'Friday': Friday/FriCount}
df = pd.Series(dict, name='DailyValue')
dff = DataFrame(df)
dff['Day'] = dff.index
dff['Sorter'] = [5,1,4,2,3]

dff.sort_values(by = ['Sorter'], inplace = True) 
#dff.sort(['Day'], ascending = True) 

dff.plot(kind='bar', grid = True, y = ['DailyValue'])
plt.show()

Upvotes: 1

Views: 2501

Answers (1)

Alexander
Alexander

Reputation: 109626

You can use a list comprehension to generate the weekdays (i.e. 0-6) in your index, and then create a series from these values. You sort this series by these weekday values and take the index. You then use ix to index your original series based on this sorted index.

import numpy as np
import pandas as pd

s = pd.Series(np.random.randn(14), index=pd.date_range('2016-1-1', periods=14))

s
Out[34]: 
2016-01-01    0.915488
2016-01-02   -1.053409
2016-01-03   -1.826033
2016-01-04    0.559250
2016-01-05   -0.278086
2016-01-06    0.041113
2016-01-07    1.076463
2016-01-08    0.942720
2016-01-09   -0.850388
2016-01-10   -0.649713
2016-01-11    2.769957
2016-01-12    0.498103
2016-01-13    1.098085
2016-01-14    0.699077
Freq: D, dtype: float64

idx = pd.Series([d.weekday() for d in s.index]).sort_values().index

s.ix[idx]
Out[36]: 
2016-01-04    0.559250
2016-01-11    2.769957
2016-01-05   -0.278086
2016-01-12    0.498103
2016-01-06    0.041113
2016-01-13    1.098085
2016-01-07    1.076463
2016-01-14    0.699077
2016-01-01    0.915488
2016-01-08    0.942720
2016-01-02   -1.053409
2016-01-09   -0.850388
2016-01-03   -1.826033
2016-01-10   -0.649713
dtype: float64

As a one liner...

s_new = s.ix[pd.Series([d.weekday() for d in s.index]).sort_values().index]

Exactly the same for a dataframe.

df_new = df.ix[pd.Series([d.weekday() for d in df.index]).sort_values().index]

Upvotes: 1

Related Questions