python pandas error when doing groupby counts

When doing groupby counts over multiple columns I get an error. Here is my dataframe and also an example that simply labels the distinct ‘b’ and ‘c’ groups.

df = pd.DataFrame(np.random.randint(0,2,(4,4)),
                  columns=['a', 'b', 'c', 'd'])
df['gr'] = df.groupby(['b', 'c']).grouper.group_info[0]
print df
   a  b  c  d  gr
0  0  1  0  0   1
1  1  1  1  0   2
2  0  0  1  0   0
3  1  1  1  1   2

However when the example is changed slightly so that count() is called instead of grouper.group_info[0], an error appear.

df = pd.DataFrame(np.random.randint(0,2,(4,4)),
                  columns=['a', 'b', 'c', 'd'])
df['gr'] = df.groupby(['b', 'c']).count()
print df

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-70-a46f632214e1> in <module>()
      1 df = pd.DataFrame(np.random.randint(0,2,(4,4)),
      2                   columns=['a', 'b', 'c', 'd'])
----> 3 df['gr'] = df.groupby(['b', 'c']).count()
      4 print df

C:\Python27\lib\site-packages\pandas\core\frame.pyc in __setitem__(self, key, value)
   2036         else:
   2037             # set column
-> 2038             self._set_item(key, value)
   2039 
   2040     def _setitem_slice(self, key, value):

C:\Python27\lib\site-packages\pandas\core\frame.pyc in _set_item(self, key, value)
   2082         ensure homogeneity.
   2083         """
-> 2084         value = self._sanitize_column(key, value)
   2085         NDFrame._set_item(self, key, value)
   2086 

C:\Python27\lib\site-packages\pandas\core\frame.pyc in _sanitize_column(self, key, value)
   2110                     value = value.values.copy()
   2111                 else:
-> 2112                     value = value.reindex(self.index).values
   2113 
   2114                 if is_frame:

C:\Python27\lib\site-packages\pandas\core\frame.pyc in reindex(self, index, columns, method, level, fill_value, limit, copy)
   2527         if index is not None:
   2528             frame = frame._reindex_index(index, method, copy, level,
-> 2529                                          fill_value, limit)
   2530 
   2531         return frame

C:\Python27\lib\site-packages\pandas\core\frame.pyc in _reindex_index(self, new_index, method, copy, level, fill_value, limit)
   2606                        limit=None):
   2607         new_index, indexer = self.index.reindex(new_index, method, level,
-> 2608                                                 limit=limit)
   2609         return self._reindex_with_indexers(new_index, indexer, None, None,
   2610                                            copy, fill_value)

C:\Python27\lib\site-packages\pandas\core\index.pyc in reindex(self, target, method, level, limit)
   2181             else:
   2182                 # hopefully?
-> 2183                 target = MultiIndex.from_tuples(target)
   2184 
   2185         return target, indexer

C:\Python27\lib\site-packages\pandas\core\index.pyc in from_tuples(cls, tuples, sortorder, names)
   1803                 tuples = tuples.values
   1804 
-> 1805             arrays = list(lib.tuples_to_object_array(tuples).T)
   1806         elif isinstance(tuples, list):
   1807             arrays = list(lib.to_object_array_tuples(tuples).T)

C:\Python27\lib\site-packages\pandas\lib.pyd in pandas.lib.tuples_to_object_array (pandas\lib.c:42342)()

ValueError: Buffer dtype mismatch, expected 'Python object' but got 'long long'

Best answer

Evaluate df.groupby(['b', 'c']).count() in an interactive session:

In [150]: df.groupby(['b', 'c']).count()
Out[150]: 
     a  b  c  d
b c            
0 0  1  1  1  1
  1  1  1  1  1
1 1  2  2  2  2

This is a whole DataFrame. It is probably not what you want to assign to a new column of df (in fact, you can not assign a column to a DataFrame, which is why an albeit cryptic exception is raised.).


If you wish to create a new column which counts the number of rows in each group, you could use

df['gr'] = df.groupby(['b', 'c'])['a'].transform('count')

For example,

import pandas as pd
import numpy as np
np.random.seed(1)
df = pd.DataFrame(np.random.randint(0, 2, (4, 4)),
                  columns=['a', 'b', 'c', 'd'])
print(df)
#    a  b  c  d
# 0  1  1  0  0
# 1  1  1  1  1
# 2  1  0  0  1
# 3  0  1  1  0

df['gr'] = df.groupby(['b', 'c'])['a'].transform('count')

df['comp_ids'] = df.groupby(['b', 'c']).grouper.group_info[0]
print(df)

yields

   a  b  c  d  gr  comp_ids
0  1  1  0  0   1         1
1  1  1  1  1   2         2
2  1  0  0  1   1         0
3  0  1  1  0   2         2

Notice that df.groupby(['b', 'c']).grouper.group_info[0] is returning something other than the counts of the number of rows in each group. Rather, it is returning a label for each group.