使用熊猫"where"方法时的"AttributeError: 'float' object has no attribute 'all'"

Question

我正在尝试使用Pandas的where方法，但我坚持了一个错误。这是一个非常小的例子。

提供文件

Chamber,Treatment
1,Sem palha
1,Sem palha
1,Sem palha
2,Sem palha
2,Sem palha

当我跑步时

import pandas
sample = pandas.read_csv('sample.csv')
sample.where(sample['Chamber'] == 1)

我得到

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-62-96ed02cb41da> in <module>()
----> 1 sample.where(sample['Chamber'] == 1)
/usr/lib/python3.4/site-packages/pandas/core/generic.py in where(self, cond, other, inplace, axis, level, try_cast, raise_on_error)
   3346             new_data = self._data.where(other=other, cond=cond, align=axis is None,
   3347                                         raise_on_error=raise_on_error,
-> 3348                                         try_cast=try_cast)
   3349 
   3350             return self._constructor(new_data).__finalize__(self)
/usr/lib/python3.4/site-packages/pandas/core/internals.py in where(self, **kwargs)
   2434 
   2435     def where(self, **kwargs):
-> 2436         return self.apply('where', **kwargs)
   2437 
   2438     def eval(self, **kwargs):
/usr/lib/python3.4/site-packages/pandas/core/internals.py in apply(self, f, axes, filter, do_integrity_check, **kwargs)
   2416                                                  copy=align_copy)
   2417 
-> 2418             applied = getattr(b, f)(**kwargs)
   2419 
   2420             if isinstance(applied, list):
/usr/lib/python3.4/site-packages/pandas/core/internals.py in where(self, other, cond, align, raise_on_error, try_cast)
   1043         axis = cond.ndim - 1
   1044         cond = cond.swapaxes(axis, 0)
-> 1045         mask = np.array([cond[i].all() for i in range(cond.shape[0])],
   1046                         dtype=bool)
   1047 
/usr/lib/python3.4/site-packages/pandas/core/internals.py in <listcomp>(.0)
   1043         axis = cond.ndim - 1
   1044         cond = cond.swapaxes(axis, 0)
-> 1045         mask = np.array([cond[i].all() for i in range(cond.shape[0])],
   1046                         dtype=bool)
   1047 
AttributeError: 'float' object has no attribute 'all'

Answer 1

你的语法实际上可以在新版本的熊猫中使用。您可以使用以下命令在 Ubuntu 上升级熊猫

sudo pip install --upgrade pandas

或者如果你使用的是python3，那么

sudo pip3 install --upgrade pandas

Answer 2

正如unutbu在评论中指出的那样，where似乎是一个"单元格明智"的选择。这些变体按预期工作：

sample[sample["Chamber"] == 1]

和

sample.query("Chamber == 1")