Use GroupBy.cumcount
for helper counter for MultiIndex
and reshape by DataFrame.unstack
, then for correct order is used DataFrame.sort_index
with map
for flatten MultiIndex
:
df = (df.set_index(['a',df.groupby('a').cumcount().add(1)])
.unstack()
.sort_index(axis=1, level=[1, 0], ascending=[True, False]))
df.columns = df.columns.map(lambda x: f'{x[0]}{x[1]}')
df = df.reset_index()
print (df)
a date1 c1 date2 c2 date3 c3 date4 c4
0 ABC 2020-06-01 0.1 2020-05-01 0.2 NaN NaN NaN NaN
1 DEF 2020-07-01 0.3 2020-01-01 0.4 2020-02-01 0.5 2020-07-01 0.6
Or if sorting is not possible because different columns names one idea is use DataFrame.reindex
:
df1 = df.set_index(['a',df.groupby('a').cumcount().add(1)])
mux = pd.MultiIndex.from_product([df1.index.levels[1], ['date','c']])
df = df1.unstack().swaplevel(1,0, axis=1).reindex(mux, axis=1)
df.columns = df.columns.map(lambda x: f'{x[1]}{x[0]}')
df = df.reset_index()
print (df)
a date1 c1 date2 c2 date3 c3 date4 c4
0 ABC 2020-06-01 0.1 2020-05-01 0.2 NaN NaN NaN NaN
1 DEF 2020-07-01 0.3 2020-01-01 0.4 2020-02-01 0.5 2020-07-01 0.6
与恶龙缠斗过久,自身亦成为恶龙;凝视深渊过久,深渊将回以凝视…