在 Pandas 中如何用最接近的非 nan 值替换零值？

Question

I have a dataframe where the col look like :

NaN
859.0
NaN
NaN
0.0
NaN

and I would like to change the zero by the previous non NaN value, and don't change the other NaN,id get this :

NaN
859.0
NaN
NaN
859.0
NaN

I've tried replace with ffill, but can't manage to get the right output.

Any help welcome !

Answer 1

.ffill().shift() will propagate the last non-null value forward, and then you can just assign any rows with value = 0 to that:

In [42]: s.ffill().shift()
Out[42]:
0      NaN
1      NaN
2    859.0
3    859.0
4    859.0
5      0.0
dtype: float64

In [43]: s[s==0] = s.ffill().shift()

In [44]: s
Out[44]:
0      NaN
1    859.0
2      NaN
3      NaN
4    859.0
5      NaN
dtype: float64

Answer 2

First replace 0 to missing values, use ffill for forward filling missing values and last replace missing values back by Series.mask :

df['col'] = df['col'].mask(df['col'].eq(0)).ffill().mask(df['col'].isna())
print (df)
     col
0    NaN
1  859.0
2    NaN
3    NaN
4  859.0
5    NaN

Answer 3

you could also do this with last_valid_index:

say your column is in df['col']

for i,_ in df.iterrows():
    if df.loc[i,'col'] == 0:
        df.at[i,'col'] = df.loc[df.loc[:i-1,'col'].last_valid_index(),'col']

output:

     col
0    NaN
1  859.0
2    NaN
3    NaN
4  859.0
5    NaN