缺少数据转换规则和索引

While pandas supports storing arrays of integer and boolean type, these types are not capable of storing missing data. Until we can switch to using a native NA type in NumPy, we’ve established some “casting rules”. When a reindexing operation introduces missing data, the Series will be cast according to the rules introduced in the table below.

data typeCast to
integerfloat
booleanobject
floatno cast
objectno cast

For example:

  1. In [127]: s = pd.Series(np.random.randn(5), index=[0, 2, 4, 6, 7])
  3. In [128]: s > 0
  4. Out[128]: 
  5. 0    True
  6. 2    True
  7. 4    True
  8. 6    True
  9. 7    True
  10. dtype: bool
  12. In [129]: (s > 0).dtype
  13. Out[129]: dtype('bool')
  15. In [130]: crit = (s > 0).reindex(list(range(8)))
  17. In [131]: crit
  18. Out[131]: 
  19. 0    True
  20. 1     NaN
  21. 2    True
  22. 3     NaN
  23. 4    True
  24. 5     NaN
  25. 6    True
  26. 7    True
  27. dtype: object
  29. In [132]: crit.dtype
  30. Out[132]: dtype('O')

Ordinarily NumPy will complain if you try to use an object array (even if it contains boolean values) instead of a boolean array to get or set values from an ndarray (e.g. selecting values based on some criteria). If a boolean vector contains NAs, an exception will be generated:

  1. In [133]: reindexed = s.reindex(list(range(8))).fillna(0)
  3. In [134]: reindexed[crit]
  4. ---------------------------------------------------------------------------
  5. ValueError                                Traceback (most recent call last)
  6. <ipython-input-134-0dac417a4890> in <module>()
  7. ----> 1 reindexed[crit]
  9. /pandas/pandas/core/series.py in __getitem__(self, key)
  10.     805             key = list(key)
  11.     806 
  12. --> 807         if com.is_bool_indexer(key):
  13.     808             key = check_bool_indexer(self.index, key)
  14.     809 
  16. /pandas/pandas/core/common.py in is_bool_indexer(key)
  17.     105             if not lib.is_bool_array(key):
  18.     106                 if isna(key).any():
  19. --> 107                     raise ValueError('cannot index with vector containing '
  20.     108                                      'NA / NaN values')
  21.     109                 return False
  23. ValueError: cannot index with vector containing NA / NaN values

However, these can be filled in using fillna() and it will work fine:

  1. In [135]: reindexed[crit.fillna(False)]
  2. Out[135]: 
  3. 0    0.126504
  4. 2    0.696198
  5. 4    0.697416
  6. 6    0.601516
  7. 7    0.003659
  8. dtype: float64
  10. In [136]: reindexed[crit.fillna(True)]
  11. Out[136]: 
  12. 0    0.126504
  13. 1    0.000000
  14. 2    0.696198
  15. 3    0.000000
  16. 4    0.697416
  17. 5    0.000000
  18. 6    0.601516
  19. 7    0.003659
  20. dtype: float64