1

I am attempting to combine elements of a dataframe into a nested list. Say I have the following:

df = pd.DataFrame(np.random.randn(100,4), columns=list('abcd'))
df.head(4)

          a         b         c         d
0  0.455258  1.135895  0.573383 -0.637943
1  0.262079 -0.397168 -0.980062 -1.600837
2  0.921582  0.767232 -0.298590 -0.159964
3 -0.645110 -0.709058  1.223899  0.382212

Then, I would like to create a fifth column e that looks like:

          a         b         c         d         e
0  0.455258  1.135895  0.573383 -0.637943 [[0.455258  1.135895  0.573383 -0.637943]]
1  0.262079 -0.397168 -0.980062 -1.600837 [[0.262079 -0.397168 -0.980062 -1.600837]]
2  0.921582  0.767232 -0.298590 -0.159964 [[0.921582  0.767232 -0.298590 -0.159964]]
3 -0.645110 -0.709058  1.223899  0.382212 [[-0.645110 -0.709058  1.223899  0.382212]]

efficiently.

My most efficient but wrong guess so far has been to do

df['e'] = df.values.tolist()

But that just results in:

          a         b         c         d         e
0  0.455258  1.135895  0.573383 -0.637943 [0.455258  1.135895  0.573383 -0.637943]
1  0.262079 -0.397168 -0.980062 -1.600837 [0.262079 -0.397168 -0.980062 -1.600837]
2  0.921582  0.767232 -0.298590 -0.159964 [0.921582  0.767232 -0.298590 -0.159964]
3 -0.645110 -0.709058  1.223899  0.382212 [-0.645110 -0.709058  1.223899  0.382212]

My least efficient but correct guess has been:

a = []
for index, row in df.iterrows():
    a.append([[row['a'],row['b'],row['c'],row['d']]])

Is there a better way?

Improve this question

3 Answers 3

Reset to default
1

Another possible solution:

df['e'] = df.values.tolist()
df['e'] = df['e'].map(lambda x: [x])

Output:

          a         b         c         d  \
0 -1.594129  1.692562  0.602186 -1.620295   
1 -0.561567 -0.033658 -1.259215  1.054229   
2  0.450852 -0.483194  0.126173  0.354781   
3  2.060968 -0.428400 -0.973516 -0.201786   
4 -0.977307 -0.123215 -1.494138 -0.175432   

                                                   e  
0  [[-1.5941291794267378, 1.6925620764107292, 0.6...  
1  [[-0.5615669341251519, -0.03365818317800309, -...  
2  [[0.45085184068754164, -0.48319360005444034, 0...  
3  [[2.0609676606685086, -0.42839969840552594, -0...  
4  [[-0.9773067339895964, -0.12321466907036417, -...
Improve this answer
Sign up to request clarification or add additional context in comments.

1 Comment

Ah I was not far off, I should have guessed that a lambda would do it. I'll accept this answer as testing on my machine shows it to be ~3x faster than Scott's, though both produce the correct output.
1

Let's use np.array_split:

df['e'] = np.array_split(df.to_numpy(), df.shape[0], axis=0)

Output:

           a         b         c         d                                                  e
0  -0.164745 -0.498313 -0.247778 -1.531003  [[-0.16474534230721335, -0.49831346259483156, ...
1   0.079485  0.125790  0.002755 -0.182361  [[0.0794845071834397, 0.12579014367640728, 0.0...
2   0.790263  0.488152 -0.752555  0.432949  [[0.790263001866772, 0.48815219760288764, -0.7...
3  -0.139499 -1.493593 -1.708668 -2.495497  [[-0.13949904491921675, -1.493593498340277, -1...
4   2.662431  0.247559 -0.949407  2.746299  [[2.662430989009563, 0.2475588133223812, -0.94...
..       ...       ...       ...       ...                                                ...
95  0.252663  1.018614 -0.491736 -0.290786  [[0.252663350866794, 1.018613617727022, -0.491...
96  1.023089 -0.367463  0.437327 -0.017441  [[1.0230888404185123, -0.3674628009130751, 0.4...
97  0.571278  0.450803  0.441102  1.176884  [[0.5712775025212533, 0.4508029251387083, 0.44...
98  1.336477  0.166516  0.408941  0.972896  [[1.3364769455886123, 0.16651649771088423, 0.4...
99 -1.298205  1.868477 -0.174665  0.065565  [[-1.2982050517578514, 1.8684774453090633, -0....
Improve this answer

1 Comment

So this is a nice one-liner that produces the desired result, however Paul's suggestion was ~3x faster when I timed them both, so I've accepted his reply. Thank you for the insight into array_split however.
0

try:

df["e"]=df.apply(lambda x:[[x[column] for column in df.columns]],axis=1)
Improve this answer

2 Comments

This does not produce the same results as requested by the OP. Please verify and fix it.
Kindly improve the answer with output or reasoning on the method and functions.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.