merge/concatenate x columns with x columns in the same dafarame-CodePudding

I would like to combine a part of the columns from the dataframe with another part on the principle given in the screenshot. I have tried the pd.concat, pd.merge functions but I can't adjust them this way. Below I am posting a portion of the dataframe. The target I would like is that when there is a result in both parts it will select one of them based on the condition.

            ('C1Cs')    ('C1Co')    ('C1Cg')    ('C1Cmp')   ('C1Xs')    ('C1Xobs')  ('C1Xg')    ('C1Xmp')
HERS00GBR   38.05       684.0       3.2         0.0         Nan         Nan         Nan         Nan
HUEG00DEU   Nan         Nan         Nan         Nan         46.26       27372.9     96.9        0.0
JFNG00CHN   43.19       2879.5      0.4         0.0         43.17       27143.9     31.6        0.0
JOZE00POL   40.03       645.0       4.8         0.0         Nan         Nan         Nan         Nan
KIR000SWE   33.66       727.5       4.7         0.0         Nan         Nan         Nan         Nan
KIRU00SWE   33.12       827.5       34.2        0.0         Nan         Nan         Nan         Nan

CodePudding user response：

One option is to combine_first (or fillna) the left part with the right part.

For this you need to slice and rename the right-hand columns:

half = len(df.columns)//2
rename_dic = dict(zip(df.columns[half:], df.columns[:half]))
out = df.iloc[:, :half].combine_first(df.iloc[:, half:].rename(columns=rename_dic))

output:

          ('C1Cs') ('C1Co') ('C1Cg') ('C1Cmp')
HERS00GBR    38.05    684.0      3.2       0.0
HUEG00DEU    46.26  27372.9     96.9       0.0
JFNG00CHN    43.19   2879.5      0.4       0.0
JOZE00POL    40.03    645.0      4.8       0.0
KIR000SWE    33.66    727.5      4.7       0.0
KIRU00SWE    33.12    827.5     34.2       0.0

NB. first thing, ensure the NaNs are real NaNs and not 'Nan' strings: df = df.replace('Nan', float('nan'))

If you want in place modification of your dataframe (i.e. no output but modifying your input dataframe directly), @JonClements suggested a nice alternative with update and set_axis:

df.update(df.iloc[:, -4:].set_axis(df.columns[:4], axis=1), overwrite=False)