How to do a fillna with zero values until data appears in each column, then use the forward fill for-CodePudding

I have the following data frame,

date          x1     x2     x3  
2001-01-01   nan    0.4    0.1 
2001-01-02   nan    0.3    nan
2021-01-03   nan    nan    0.5
...
2001-05-05   nan    0.1    0.2
2001-05-06   0.1    nan    0.3
...

So I want to first fill all nan values with zero until the first data point appears in each column, after that, I want the rest of the rows to use the fowardfill function.

So the above data frame should look like this,

date          x1     x2     x3  
2001-01-01     0    0.4    0.1 
2001-01-02     0    0.3    0.1
2021-01-03     0    0.3    0.5

...
2001-05-05     0    0.1    0.2
2001-05-06   0.1    0.1    0.3
...

If I do fillna with 0 first then do forwardfill, like this,

df = df.fillna(0)
df = df.ffill()

I just get all the na values to be zero, and I am unable to do ffill for the parts where the data starts.

Is there a way to do the ffill the way I want?

CodePudding user response：

Reverse the logic:

out = df.ffill().fillna(0)
print(out)

# Output
         date   x1   x2   x3
0  2001-01-01  0.0  0.4  0.1
1  2001-01-02  0.0  0.3  0.1
2  2021-01-03  0.0  0.3  0.5
3  2001-05-05  0.0  0.1  0.2
4  2001-05-06  0.1  0.1  0.3