为清楚起见，我将从我的代码中提取摘录并使用通用名称。我有一个类Foo()，它将DataFrame存储到属性中。
import pandas as pd
import pandas.util.testing as pdt
class Foo():
def __init__(self, bar):
self.bar = bar # dict of dicts
self.df = pd.DataFrame(bar) # pandas object
def __eq__(self, other):
if isinstance(other, self.__class__):
return self.__dict__ == other.__dict__
return NotImplemented
def __ne__(self, other):
result = self.__eq__(other)
if result is NotImplemented:
return result
return not result
然而，当我尝试比较Foo的两个实例时，我得到了与比较两个DataFrames的模糊性相关的范围（比较应该可以正常工作而没有＆＃39; df＆＃39;键在{ {1}}）。
Foo.__dict__幸运的是，pandas具有用于断言两个DataFrame或Series是否为true的实用函数。如果可能，我想使用此功能的比较操作。
d1 = {'A' : pd.Series([1, 2], index=['a', 'b']),
'B' : pd.Series([1, 2], index=['a', 'b'])}
d2 = d1.copy()
foo1 = Foo(d1)
foo2 = Foo(d2)
foo1.bar # dict
foo1.df # pandas DataFrame
foo1 == foo2 # ValueError
[Out] ValueError: The truth value of a DataFrame is ambiguous. Use a.empty, a.bool(), a.item(), a.any() or a.all().
有几个选项可以解决两个pdt.assert_frame_equal(pd.DataFrame(d1), pd.DataFrame(d2)) # no raises
实例的比较：
Foo的副本，其中__dict__缺少df键new_dict删除df密钥（不理想）__dict__，但只有部分包含在元组中__dict__以促进pandas DataFrame比较 从长远来看，最后一个选项看起来最强劲，但我不确定最好的方法。最后，我想重构__eq__来比较__eq__中的所有项目，包括DataFrames（和系列）。有关如何完成此任务的任何想法？
答案 0 :(得分：2)
来自这些线程的解决方案
Comparing two pandas dataframes for differences
Pandas DataFrames with NaNs equality comparison
def df_equal(self):
try:
assert_frame_equal(csvdata, csvdata_old)
return True
except:
return False
对于数据帧字典：
def df_equal(df1, df2):
try:
assert_frame_equal(df1, df2)
return True
except:
return False
def __eq__(self, other):
if self.df.keys() != other.keys():
return False
for k in self.df.keys():
if not df_equal(self.df[k], other[k]):
return False
return True
答案 1 :(得分：0)
以下代码似乎完全符合我原来的问题。它处理大熊猫DataFrames和Series。欢迎简化。
这里的诀窍是__eq__已经实现，分别比较__dict__和pandas对象。最终比较每个人的真实性并返回结果。如果第一个值为and，True会返回第二个值，这里有一些有趣且被利用的内容。
使用错误处理和外部比较功能的想法受到@ ate50eggs提交的答案的启发。非常感谢。
import pandas as pd
import pandas.util.testing as pdt
def ndframe_equal(ndf1, ndf2):
try:
if isinstance(ndf1, pd.DataFrame) and isinstance(ndf2, pd.DataFrame):
pdt.assert_frame_equal(ndf1, ndf2)
#print('DataFrame check:', type(ndf1), type(ndf2))
elif isinstance(ndf1, pd.Series) and isinstance(ndf2, pd.Series):
pdt.assert_series_equal(ndf1, ndf2)
#print('Series check:', type(ndf1), type(ndf2))
return True
except (ValueError, AssertionError, AttributeError):
return False
class Foo(object):
def __init__(self, bar):
self.bar = bar
try:
self.ndf = pd.DataFrame(bar)
except(ValueError):
self.ndf = pd.Series(bar)
def __eq__(self, other):
if isinstance(other, self.__class__):
# Auto check attrs if assigned to DataFrames/Series, then add to list
blacklisted = [attr for attr in self.__dict__ if
isinstance(getattr(self, attr), pd.DataFrame)
or isinstance(getattr(self, attr), pd.Series)]
# Check DataFrames and Series
for attr in blacklisted:
ndf_eq = ndframe_equal(getattr(self, attr),
getattr(other, attr))
# Ignore pandas objects; check rest of __dict__ and build new dicts
self._dict = {
key: value
for key, value in self.__dict__.items()
if key not in blacklisted}
other._dict = {
key: value
for key, value in other.__dict__.items()
if key not in blacklisted}
return ndf_eq and self._dict == other._dict # order is important
return NotImplemented
def __ne__(self, other):
result = self.__eq__(other)
if result is NotImplemented:
return result
return not result
在DataFrames上测试后一段代码。
# Data for DataFrames
d1 = {'A' : pd.Series([1, 2], index=['a', 'b']),
'B' : pd.Series([1, 2], index=['a', 'b'])}
d2 = d1.copy()
d3 = {'A' : pd.Series([1, 2], index=['abc', 'b']),
'B' : pd.Series([9, 0], index=['abc', 'b'])}
# Test DataFrames
foo1 = Foo(d1)
foo2 = Foo(d2)
foo1.bar # dict of Series
foo1.ndf # pandas DataFrame
foo1 == foo2 # triggers _dict
#foo1.__dict__['_dict']
#foo1._dict
foo1 == foo2 # True
foo1 != foo2 # False
not foo1 == foo2 # False
not foo1 != foo2 # True
foo2 = Foo(d3)
foo1 == foo2 # False
foo1 != foo2 # True
not foo1 == foo2 # True
not foo1 != foo2 # False
最后测试另一个常见的pandas对象Series。
# Data for Series
s1 = {'a' : 0., 'b' : 1., 'c' : 2.}
s2 = s1.copy()
s3 = {'a' : 0., 'b' : 4, 'c' : 5}
# Test Series
foo3 = Foo(s1)
foo4 = Foo(s2)
foo3.bar # dict
foo4.ndf # pandas Series
foo3 == foo4 # True
foo3 != foo4 # False
not foo3 == foo4 # False
not foo3 != foo4 # True
foo4 = Foo(s3)
foo3 == foo4 # False
foo3 != foo4 # True
not foo3 == foo4 # True
not foo3 != foo4 # False