Replace zeroes with nan in either data frame or array based on another element in the row-CodePudding

I have a dataset which can be in a numpy array, or a dataframe, here is a sample of it in a dataframe:

  totalsum totalmean raindiffsum raindiffmean       name                  bin
0        0       NaN           0          NaN  openguage  2021-11-01 00:00:00
1        0       NaN           0          NaN  openguage  2021-11-01 00:30:00
2        0       NaN           0          NaN  openguage  2021-11-01 01:00:00
3        0       NaN           0          NaN  openguage  2021-11-01 01:30:00
4        0       NaN           0          NaN  openguage  2021-11-01 02:00:00

I have the same data as a numpy array. I need to replace the zero values with nan, but only when there is a nan in the same row.

for clarity, this is further down the same dataframe, I DO NOT want to replace the zeroes in lines 1518 and 1519 with nan.

totalsum totalmean  ...       name                  bin
1515        0       NaN  ...  openguage  2021-12-02 13:30:00
1516        0       NaN  ...  openguage  2021-12-02 14:00:00
1517        0       NaN  ...  openguage  2021-12-02 14:30:00
1518      0.0       0.0  ...  openguage  2021-12-02 15:00:00
1519      0.0       0.0  ...  openguage  2021-12-02 15:30:00

[5 rows x 6 columns]

I have tried np.where() I have tried a for loop (on the numpy array), none of these loops throw an error, but have no effect:-

for i in range(len(dfbinarr)):
    if dfbinarr[i,1] is nan:
        dfbinarr[i,0]=nan
        dfbinarr[i,2]=nan

 for i in range(len(dfbinarr)):
    if dfbinarr[i,1] is nan:
        dfbinarr[i,0]=np.nan
        dfbinarr[i,2]=np.nan

for i in range(len(dfbinarr)):
    if dfbinarr[i,1] ==nan:
        dfbinarr[i,0]=np.nan
        dfbinarr[i,2]=np.nan

any help would be greatly appreciated!

CodePudding user response：

You can use .loc to do this. df['totalmean'].isna() returns a mask (just a Series) where each value is true if that item in totalmean is NaN, false otherwise.

df.loc[df['totalmean'].isna(), 'totalsum'] = np.nan

Output:

>>> df
   totalsum  totalmean  raindiffsum  raindiffmean       name                  bin
0       NaN        NaN            0           NaN  openguage  2021-11-01x00:00:00
1       NaN        NaN            0           NaN  openguage  2021-11-01x00:30:00
2       NaN        NaN            0           NaN  openguage  2021-11-01x01:00:00
3       0.0        0.0            0           NaN  openguage  2021-11-01x01:30:00
4       0.0        0.0            0           NaN  openguage  2021-11-01x02:00:00