Pandas - 将函数应用于具有来自不同列的多个参数的数据帧

Question

I would like to use apply() function to a dataframe to generate a date range thanks to pandas date_range() function.

The following code works, and does what I expect it to do.

import pandas as pd

def my_date_range(start, end, freq):
    return pd.date_range(start = start, end = end, freq = freq)

df = pd.DataFrame({'Start':[pd.Timestamp('1970-01-02 00:00:00')], 'End':[pd.Timestamp('1970-01-02 00:30:00')], 'Freq':[pd.Timedelta(5,'m')]})

df1 = df.apply(lambda x: my_date_range(x.Start, x.End, x.Freq), axis=1)

The result:

In [28]: df
Out[28]: 
       Start                 End     Freq
0 1970-01-02 1970-01-02 00:30:00 00:05:00

In[29] : df1[0]
Out[29]: 
DatetimeIndex(['1970-01-02 00:00:00', '1970-01-02 00:05:00',
               '1970-01-02 00:10:00', '1970-01-02 00:15:00',
               '1970-01-02 00:20:00', '1970-01-02 00:25:00',
               '1970-01-02 00:30:00'],
              dtype='datetime64[ns]', freq='5T')

So now my problem / my questions. I could read that it is possible to use apply() without lambda in this way, as I understand:

df2 = df[['Start', 'End', 'Freq']].apply(my_date_range, axis=1)

But above code produces following error.

TypeError: ("my_date_range() missing 2 required positional arguments: 'end' and 'freq'", 'occurred at index 0')

Please, what am I doing wrong?

And is it interesting to avoid the use of lambda? (better performances?)

Finally, is there also a way to use directly pd.date_range?

If I try with below code, I get the following error:

df1 = df.apply(lambda x: pd.date_range(x.Start, x.End, x.Freq), axis=1)

"periods must be a number, got {periods}".format(periods=periods)

TypeError: ('periods must be a number, got 0 days 00:05:00', 'occurred at index 0')

Thanks in advance for your help! Have a good day!

Answer 1

1

As you can see in the error message, if you want to use the function name to pandas.DataFrame.apply , the function should take a pandas.Series as a argument. So it should be like this.

def my_date_range(x):
    return pd.date_range(start = x.Start, end = x.End, freq = x.Freq)
df2 = df.apply(my_date_range, axis=1)

2

Well personally I think lambda makes things a lot more convenient. In your case, the original way you used of defining a function and then using another lambda is not convenient at all, since the point of lambda is not having to use def . However you can use lambda and make it more convenient as you tried in the last part of the question.

3

The reason of the error is because the function pd.date_range arguments goes like this. pandas.date_range(start=None, end=None, periods=None, ...) So if you just give it as a positional argument as you did, it thinks the third argument is period= . You should give it as a keyword argument (as you did in above).

df1 = df.apply(lambda x: pd.date_range(start = x.Start, end = x.End, freq = x.Freq), axis=1)

Answer 2

What about something like this:

import pandas as pd
start = pd.Timestamp('1970-01-02 00:00:00')
end = pd.Timestamp('1970-01-02 00:30:00')
pd.date_range(start, end, freq='5Min')