Pandas 重新采样以在所有值为 NaN 时返回 NaN

Question

I'm using resample to sum my data into hourly blocks. When all input data for the hour is NaN, resample is producing a value of 0 instead of NaN.

My raw data is this:

infile
Out[206]:
             Date_time  Rainfall
0  2019-02-02 14:18:00       NaN
1  2019-02-02 14:20:00       NaN
2  2019-02-02 14:25:00       NaN
3  2019-02-02 14:30:00       NaN
4  2019-02-02 14:35:00       NaN
5  2019-02-02 14:40:00       NaN
6  2019-02-02 14:45:00       NaN
7  2019-02-02 14:50:00       NaN
8  2019-02-02 14:55:00       NaN
9  2019-02-02 15:00:00       0.0
10 2019-02-02 15:05:00       NaN
11 2019-02-02 15:10:00       NaN
12 2019-02-02 15:15:00       NaN
13 2019-02-02 15:20:00       NaN
14 2019-02-02 15:25:00       NaN
15 2019-02-02 15:30:00       NaN
16 2019-02-02 15:35:00       NaN
17 2019-02-02 15:40:00       NaN
18 2019-02-02 15:45:00       NaN
19 2019-02-02 15:50:00       NaN
20 2019-02-02 15:55:00       NaN

I want my output to be this:

             Date_time  Rainfall  
0  2019-02-02 14:18:00       NaN
1  2019-02-02 15:00:00       0.0

But instead I'm getting this:

output[['Date_time', 'Rainfall']]
Out[208]: 
                Date_time  Rainfall
0     2019-02-02 14:18:00       0.0
1     2019-02-02 15:00:00       0.0

This is the code that I'm using to get there - it's a little more complicated than it needs to be for this example because I use it to iterate through a list of column names at other points:

def sum_calc(col_name):
    col =  infile[['Date_time', col_name]].copy()
    col.columns = ('A', 'B')
    col = col.resample('H', on='A').B.sum().reset_index(drop=True)
    output[col_name] = col.copy()

sum_calc('Rainfall')

Any clues on how to get this to work? I've had a look online and all the options seem to produce NaN if any value in group is NaN, rather than all values like I'm after.

Answer 1

Try:

>>> df.resample("H", on="Date_time")["Rainfall"].agg(pd.Series.sum, min_count=1)
Date_time
2021-12-17 14:00:00    NaN
2021-12-17 15:00:00    0.0
Freq: H, Name: Rainfall, dtype: float64