None handling in groupby, Series.unique

[wesm]

from mailing list. this is caused by the change in handling of None in PyObjectHashTable

import datetime
import numpy
import pandas

def simple_test2():
   data = [
           [1, 'string1', 1.0],
           [2, 'string2', 2.0],
           [3, None, 3.0]
          ]

   df = pandas.DataFrame({'key': [x[0] for x in data], 'grouper':
[x[1] for x in data], 'value': [x[2] for x in data]})
   df['weights'] = df['value']/df['value'].sum()
   gb = df.groupby('grouper').aggregate(numpy.sum)

   print
   print df
   print ''
   print gb

if __name__ == '__main__':
   simple_test2()




0.7rc1 yields:

 grouper  key  value  weights
0 string1  1    1      0.1667
1 string2  2    2      0.3333
2 NaN      3    3      0.5000

        key  value  weights
grouper
None     3    3      0.5000
string1  1    1      0.1667
string2  2    2      0.3333

0.7 final yields:
  grouper  key  value   weights
0  string1    1      1  0.166667
1  string2    2      2  0.333333
2     None    3      3  0.500000

        key  value   weights
grouper
string1    1      1  0.166667
string2    2      2  0.333333

missing the last line

[wesm]

Fixed Series.unique issue. groupby will be addressed by separate ticket #443