Do not change the order of columns when concatenating pandas data frames.

problem

When concatenating pandas data frames using pd.concat, the order of columns may change arbitrarily.

>>> df = pd.DataFrame([[1, 2], [3, 4]], index=[0, 1], columns=['B', 'A'])
>>> df2 = pd.DataFrame([[1, 2], [3, 4]], index=[0, 1], columns=['A', 'B'])
>>> pd.concat([df, df2])
   A  B #B A is good according to df!
0  2  1
1  4  3
0  1  2
1  3  4

• cf: https://github.com/pandas-dev/pandas/issues/4588

Solution

To keep the column order unchanged, use the DataFrame.append method.

>>> df.append(df2)[df.columns.tolist()]
   B  A
0  1  2
1  3  4
0  2  1
1  4  3

Now you can combine the data frames while keeping the order of the df columns. If you want to combine multiple data frames, just pass a list of data frames such as df.append ([df1, df2]).

However, DataFrame.append is slow and should not be used when joining many rows.

reference

• Why the column order is changing while appending pandas dataframes?
• Data concatenation / join processing in Python pandas diagram

Postscript

I forgot to write the execution environment, so I added it

• Python 3.6.1
• pandas 0.19.2