[FEA] Multiindex to Multicolumn

[ayushdg]

There should be a way to go from a Dataframe with a multi index to a multicolumn Dataframe. In pandas calling reset_index on a df with multi-index converts it to a multicolumn Dataframe.

[thomcom]

Fixed by #1542

[beckernick]

Re-opening this issue and widening the request to handle MultiIndex Series. This issue becomes relevant again because we can now call stack on DataFrames, which results in a Series.

In a 0.11 nightly runtime container:

import cudf
gdf = cudf.DataFrame({'a':range(3), 'b':range(5,8), 'c':range(10,13)})
print(gdf.stack().reset_index())
    0  1
0   0  a
1   5  b
2  10  c
3   1  a
4   6  b
5  11  c
6   2  a
7   7  b
8  12  c

The MultiIndex is reset but we lose information, when it should be multiple columns named by their level + the Series values:

import pandas as pd
df = pd.DataFrame({'a':range(3), 'b':range(5,8), 'c':range(10,13)})
print(df.stack().reset_index())
   level_0 level_1   0
0        0       a   0
1        0       b   5
2        0       c  10
3        1       a   1
4        1       b   6
5        1       c  11
6        2       a   2
7        2       b   7
8        2       c  12

[galipremsagar]

as of 0.15, this issue exists.

      1 import cudf
      2 gdf = cudf.DataFrame({'a':range(3), 'b':range(5,8), 'c':range(10,13)})
----> 3 print(gdf.stack().reset_index())

/conda/envs/cudf/lib/python3.7/site-packages/cudf/core/series.py in reset_index(self, drop, inplace)
    568                     "to create a DataFrame"
    569                 )
--> 570             return self.to_frame().reset_index(drop=drop)
    571         else:
    572             if inplace is True:

/conda/envs/cudf/lib/python3.7/site-packages/cudf/core/dataframe.py in reset_index(self, level, drop, inplace, col_level, col_fill)
   2456                 reversed(names), reversed(index_columns)
   2457             ):
-> 2458                 result.insert(0, name, index_column)
   2459         result.index = RangeIndex(len(self))
   2460         if inplace:

/conda/envs/cudf/lib/python3.7/contextlib.py in inner(*args, **kwds)
     72         def inner(*args, **kwds):
     73             with self._recreate_cm():
---> 74                 return func(*args, **kwds)
     75         return inner
     76 

/conda/envs/cudf/lib/python3.7/site-packages/cudf/core/dataframe.py in insert(self, loc, name, value)
   2544         num_cols = len(self._data)
   2545         if name in self._data:
-> 2546             raise NameError("duplicated column name {!r}".format(name))
   2547 
   2548         if loc < 0:

NameError: duplicated column name 0

In [2]: import pandas as pd 
   ...: df = pd.DataFrame({'a':range(3), 'b':range(5,8), 'c':range(10,13)}) 
   ...: print(df.stack().reset_index())                                                    
   level_0 level_1   0
0        0       a   0
1        0       b   5
2        0       c  10
3        1       a   1
4        1       b   6
5        1       c  11
6        2       a   2
7        2       b   7
8        2       c  12

In [3]:                                                                                    

[beckernick]

Closing. In the current 21.08 nightly, this is now fixed:

import cudf
gdf = cudf.DataFrame({'a':range(3), 'b':range(5,8), 'c':range(10,13)})
print(gdf.stack().reset_index())
   level_0 level_1   0
0        0       a   0
1        0       b   5
2        0       c  10
3        1       a   1
4        1       b   6
5        1       c  11
6        2       a   2
7        2       b   7
8        2       c  12

import pandas as pd
df = pd.DataFrame({'a':range(3), 'b':range(5,8), 'c':range(10,13)})
print(df.stack().reset_index())
   level_0 level_1   0
0        0       a   0
1        0       b   5
2        0       c  10
3        1       a   1
4        1       b   6
5        1       c  11
6        2       a   2
7        2       b   7
8        2       c  12