TypeError using timedelta, cannot sum times

问题

I have data that looks like this:

    user                in               out location  flag     Time
0    ron  12/21/2021 10:11  12/21/2016 17:50     home     0  4:19:03
1    ron  12/21/2016 13:26  12/21/2016 13:52   office     2  0:25:28
2  april   12/21/2016 8:12  12/21/2016 17:27   office     0  8:15:03
3  april  12/21/2016 18:54  12/21/2016 22:56   office     0  4:02:36
4   andy   12/21/2016 8:57  12/21/2016 12:15     home     0  2:59:40

I want to sum or take the max value of time per user based on the flag. So I converted the column to timedelta.

sample.loc[:,'Time'] = pd.to_timedelta(sample['Time'])

However, when I try to test this by summing the entire column

sum(sample['Time'])

I get the following error:

TypeError: unsupported operand type(s) for +: 'int' and 'Timedelta'

What am I missing here? I thought you could sum with Timedelta.

回答1:

Python's sum, by default, assumes you are summing integers. Hence it tries to start summing from 0, which is where this error comes from. It's impossible to add 0 to a timedelta.

This can be fixed in 2 ways:

Provide a different starting value to sum, perhaps an "empty" timedelta, as the second argument for sum:
```
from datetime import timedelta
...
sum(sample['Time'], timedelta())
```
Use Series.sum (which will probably have better performance anyway):
```
sample['Time'].sum()
```

回答2:

In order to take the sum of the number of days that you received using pd.to_timedelta(), you need to do the following:

sum(sample.Time.astype('int'))/8.64e+13

That is, you need to convert the 'Time' column integer to perform the sum() operation. 8.64e+13 is to convert the ns to days.

来源：https://stackoverflow.com/questions/60804162/typeerror-using-timedelta-cannot-sum-times

标签

python

pandas

timedelta