CODEWITHSUNDEEP
pythonpandas
pythonRcpp
pythonRcpp

Reputation: 2146

Sum only numeric columns in pandas

I have a df like below, (initial 2 rows are text and 1st column is date)

In [4]: df
Out[4]: 
           test          bs         dv         if          ir         md         qb         sy          tb
0       TESTacc      a10900     a10900     a10900    IJJMKK11     a10900     a10900     a10900      a10900
1   01-Feb-2019  18.8668013  4.6021207  0.9330807  13.9766832  2.9002571  0.2824343  0.8280988   0.8587644
2   04-Feb-2019   16.187526  3.1000162  0.4145835  14.6465183   2.848472  0.2516608  0.8618771    0.218063

I need to get have this csv with 3 decimal precision Also I need to add a "Total" Column (rightmost column) I have tried the below things, but these are not proper

To add the total column I did:

ndf=df.iloc[2:,1:] #take only numerics in ndf
ndf = ndf.apply(pd.to_numeric)
ndf=ndf.round(3)
df['total']=ndf.sum(axis=1)

This is not a proper way of doing simple thing like adding a total column

So I tried df=df.apply(pd.to_numeric,errors='ignore') but round still wont work on df My intent is to just add a Total column and have all numbers rounded to 3 decimals. Additional: Once this is done I would add a last row as median row, having median for each column

Upvotes: 12

Views: 19350

Answers (2)

Noah Sheldon
Noah Sheldon

Reputation: 1652

According to the latest pandas documentation 1.0.3 you can sum only numeric columns with the following code:

df_sum = df.sum(numeric_only = True)

This will sum all numeric columns in df and assign it to variable df_sum.

Upvotes: 24

anky
anky

Reputation: 75140

IIUC, you may need:

df['sum']=df.apply(lambda x: pd.to_numeric(x,errors='coerce')).sum(axis=1).round(3)
#for median: df.apply(lambda x: pd.to_numeric(x,errors='coerce')).median(axis=1).round(3)
print(df)

          test          bs         dv         if          ir         md  \
0      TESTacc      a10900     a10900     a10900    IJJMKK11     a10900   
1  01-Feb-2019  18.8668013  4.6021207  0.9330807  13.9766832  2.9002571   
2  04-Feb-2019   16.187526  3.1000162  0.4145835  14.6465183   2.848472   

          qb         sy         tb     sum  
0     a10900     a10900     a10900   0.000  
1  0.2824343  0.8280988  0.8587644  43.248  
2  0.2516608  0.8618771   0.218063  38.529

EDIT you can use, df.where() to round all neumerics as :

df['sum']=df.apply(lambda x: pd.to_numeric(x,errors='coerce')).sum(axis=1)
df=(df.where(df.apply(lambda x: pd.to_numeric(x,errors='coerce')).isna(),
         df.apply(lambda x: pd.to_numeric(x,errors='coerce')).round(3)))
print(df)

          test      bs      dv      if        ir      md      qb      sy  \
0      TESTacc  a10900  a10900  a10900  IJJMKK11  a10900  a10900  a10900   
1  01-Feb-2019  18.867   4.602   0.933    13.977     2.9   0.282   0.828   
2  04-Feb-2019  16.188     3.1   0.415    14.647   2.848   0.252   0.862   

       tb     sum  
0  a10900   0.000  
1   0.859  86.496  
2   0.218  77.057

Upvotes: 2

Related Questions