Question

I have the following Pandas dataframe in Python 2.7.

import pandas as pd
trial_num = [1,2,3,4,5]
sail_rem_time = [ 11:33:11 , 16:29:05 , 09:37:56 , 21:43:31 , 17:42:06 ]
dfc = pd.DataFrame(zip(*[trial_num,sail_rem_time]),columns=[ Temp_Reading , Time_of_Sail ])
print dfc

The dataframe looks like this:

  Temp_Reading Time_of_Sail
             1     11:33:11
             2     16:29:05
             3     09:37:56
             4     21:43:31
             5     17:42:06

This dataframe comes from a *.csv file. I use Pandas to read in the *.csv file as a Pandas dataframe. When I use print dfc.dtypes, it shows me that the column Time_of_Sail has a datatype object. I would like to convert this column to datetime datatype BUT I only want the time part - I don t want the year, month, date.

I can try this:

dfc[ Time_of_Sail ] = pd.to_datetime(dfc[ Time_of_Sail ])
dfc[ Time_of_Sail ] = [time.time() for time in dfc[ Time_of_Sail ]]

but the problem is that the when I run print dfc.dtypes it still shows that the column Time_of_Sail is object.

Is there a way to convert this column into a datetime format that only has the time?

Additional Information:

To create the above dataframe and output, this also works:

import pandas as pd
trial_num = [1,2,3,4,5]
sail_rem_time = [ 11:33:11 , 16:29:05 , 09:37:56 , 21:43:31 , 17:42:06 ]
data = [
    [trial_num[0],sail_rem_time[0]],
    [trial_num[1],sail_rem_time[1]],[trial_num[2],sail_rem_time[2]],
    [trial_num[3],sail_rem_time[3]]
    ]
dfc = pd.DataFrame(data,columns=[ Temp_Reading , Time_of_Sail ])
dfc[ Time_of_Sail ] = pd.to_datetime(dfc[ Time_of_Sail ])
dfc[ Time_of_Sail ] = [time.time() for time in dfc[ Time_of_Sail ]]
print dfc
print dfc.dtypes

Answer 1

These two lines:

dfc[ Time_of_Sail ] = pd.to_datetime(dfc[ Time_of_Sail ])
dfc[ Time_of_Sail ] = [time.time() for time in dfc[ Time_of_Sail ]]

Can be written as:

dfc[ Time_of_Sail ] = pd.to_datetime(dfc[ Time_of_Sail ],format=  %H:%M:%S  ).dt.time

Answer 2

Using to_timedelta,we can convert string to time format(timedelta64[ns]) by specifying units as second,min etc.,

dfc[ Time_of_Sail ] = pd.to_timedelta(dfc[ Time_of_Sail ], unit= s )

Answer 3

This seems to work:

dfc[ Time_of_Sail ] = pd.to_datetime(dfc[ Time_of_Sail ], format= %H:%M:%S ).apply(pd.Timestamp)

Answer 4

If anyone is searching for a more generalized answer try

dfc[ Time_of_Sail ]= pd.to_datetime(dfc[ Time_of_Sail ])

Answer 5

If you just want a simple conversion you can do the below:

import datetime as dt

dfc.Time_of_Sail = dfc.Time_of_Sail.astype(dt.datetime)

or you could add a holder string to your time column as below, and then convert afterwards using an apply function:

dfc.Time_of_Sail = dfc.Time_of_Sail.apply(lambda x:  2016-01-01   + str(x))
dfc.Time_of_Sail = pd.to_datetime(dfc.Time_of_Sail).apply(lambda x: dt.datetime.time(x))

Answer 6

(Python 3)
You can apply pd.to_datetime() and datetime.time to the Series with apply() function and dt accessor: dfc[ Time_of_Sail ].apply(pd.to_datetime).dt.time.

Recap

import pandas as pd

# Original DataFrame
>>> dfc = pd.DataFrame(
>>>     {
>>>          Temp_Reading : [1, 2, 3, 4, 5],
>>>          Time_of_Sail : [ 11:33:11 ,  16:29:05 ,  09:37:56 ,  21:43:31 ,  17:42:06 ]
>>>     }
>>>> )

# Convert to datetime.time object
>>> dfc[ Time_of_Sail ] = dfc[ Time_of_Sail ].apply(pd.to_datetime).dt.time
>>> dfc[ Time_of_Sail ]
    0    11:33:11
    1    16:29:05
    2    09:37:56
    3    21:43:31
    4    17:42:06
    Name: Time_of_Sail, dtype: object

Data type
Indeed dfc.dtypes returns Time_of_Sail object, but you ll see they are datetime.time objects at a closer look:

>>> dfc[ Time_of_Sail ][0]
datetime.time(11, 33, 11)

友情链接