Does Index numbering matter in testing dataframe equality? I have 2 identical dataframes with exactly the same data and columns. The only difference is that the index numbers for each row is different and equals methods returns a False. How can I get around this? Here are my data frames
A B
0 87 54
1 87 75
2 87 22
3 87 69
A B
418 87 69
107 87 54
108 87 75
250 87 22
You can use np.array_equal to check the values, however the ordering is important, so in your example you have to sort by the index first.
In [11]: df1
Out[11]:
A B
0 87 54
1 87 75
2 87 22
3 87 69
In [12]: df2
Out[12]:
A B
418 87 69
107 87 54
108 87 75
250 87 22
In [13]: df3 = df2.sort()
In [14]: df3
Out[14]:
A B
107 87 54
108 87 75
250 87 22
418 87 69
In [15]: np.array_equal(df1, df3)
Out[15]: True
Note: You can't compare df1 and df2 as they have different indexes:
In [21]: df1 == df2
ValueError: Can only compare identically-labeled DataFrame object
You can reset the index, but be aware that an exception can be raised for that reason:
In [22]: df3.reset_index(drop=True)
Out[22]:
A B
0 87 54
1 87 75
2 87 22
3 87 69
In [23]: np.all(df1 == df3.reset_index(drop=True))
Out[23]: True
Another option is to have a try and except block around assert_frame_equals:
In [24]: pd.util.testing.assert_frame_equal(df1, df3.reset_index(drop=True))
as in this related answer.
As Jeff points out you can use .equals, which does this:
In [25]: df1.equals(df3.reset_index(drop=True))
Out[25]: True
