moving a row of a pandas dataframe but index unchanged

Question

import numpy as np
import pandas as pd

d = np.random.rand(10,3)
df = pd.DataFrame(d)
df

          0         1         2
0  0.740335  0.957937  0.611504
1  0.164772  0.469355  0.310799
2  0.666506  0.030054  0.986393
3  0.191827  0.554556  0.707453
4  0.278617  0.532423  0.095124
5  0.659526  0.413681  0.588372
6  0.285892  0.466231  0.268647
7  0.559664  0.274913  0.618610
8  0.572839  0.490969  0.194019
9  0.028767  0.332626  0.290148

# move the third row to top
df.iloc[:3] = df.iloc[[2,0,1]]
df
          0         1         2
0  0.666506  0.030054  0.986393
1  0.740335  0.957937  0.611504
2  0.164772  0.469355  0.310799
3  0.191827  0.554556  0.707453
4  0.278617  0.532423  0.095124
5  0.659526  0.413681  0.588372
6  0.285892  0.466231  0.268647
7  0.559664  0.274913  0.618610
8  0.572839  0.490969  0.194019
9  0.028767  0.332626  0.290148

The index is unchanged.

I tried df.index[:3] = df.index[[2,0,1]] (raise error) or df.index.values[:3] = df.index.values[[2,0,1]], there is no error raised but the index is still unchanged.

Are you expecting the index of the first 3 rows to also be 2,0,1 in this case? — Asish M.
– Asish M., Commented May 29, 2021 at 22:33

Asish M. · Accepted Answer · 2021-05-29 22:34:45Z

3

If you want the index to also reflect the updated row, you need to use df.reindex

In [25]: np.random.seed(5)
    ...: d = np.random.rand(5, 3)
    ...: df = pd.DataFrame(d)
    ...: df
Out[25]:
          0         1         2
0  0.221993  0.870732  0.206719
1  0.918611  0.488411  0.611744
2  0.765908  0.518418  0.296801
3  0.187721  0.080741  0.738440
4  0.441309  0.158310  0.879937

In [26]: df.reindex([2,0,1,*df.index[3:]])
Out[26]:
          0         1         2
2  0.765908  0.518418  0.296801
0  0.221993  0.870732  0.206719
1  0.918611  0.488411  0.611744
3  0.187721  0.080741  0.738440
4  0.441309  0.158310  0.879937

answered May 29, 2021 at 22:34

Asish M.

2,6571 gold badge19 silver badges34 bronze badges

Sign up to request clarification or add additional context in comments.

Comments

blackraven · Accepted Answer · 2021-05-29 22:39:02Z

1

How about this?

np.random.seed(42)
df = pd.DataFrame(np.random.rand(10,3))

df = df.iloc[[2,0,1]].append(df.iloc[3:])

edited May 29, 2021 at 22:39

answered May 29, 2021 at 22:36

blackraven

5,6797 gold badges27 silver badges51 bronze badges

1 Comment

blackraven Over a year ago

you would like to arrange the first 3 rows and append the rest?

Henry Ecker · Accepted Answer · 2021-05-29 22:45:38Z

1

iloc will try to align the index.

For this reason, when the values are assigned back they will go right back to their initial location. Index 0 will align to index 0, index 1 will align to index 1, etc...

df.iloc[[2, 0, 1]]

          0         1         2
2  0.765908  0.518418  0.296801
0  0.221993  0.870732  0.206719
1  0.918611  0.488411  0.611744

(Note the index values)

To preserve the index values join the two DataFrame pieces together append or concat:

import numpy as np
import pandas as pd

np.random.seed(5)
d = np.random.rand(5, 3)
df = pd.DataFrame(d)

df = df.iloc[[2, 0, 1]].append(df.iloc[3:])

OR

df = pd.concat((df.iloc[[2, 0, 1]], df.iloc[3:]))

df Before:

          0         1         2
0  0.221993  0.870732  0.206719
1  0.918611  0.488411  0.611744
2  0.765908  0.518418  0.296801
3  0.187721  0.080741  0.738440
4  0.441309  0.158310  0.879937

df after:

          0         1         2
2  0.765908  0.518418  0.296801
0  0.221993  0.870732  0.206719
1  0.918611  0.488411  0.611744
3  0.187721  0.080741  0.738440
4  0.441309  0.158310  0.879937

(To match shown expected output)

To move without the indexes use values or to_numpy to prevent the aligning behaviour:

import numpy as np
import pandas as pd

np.random.seed(5)
d = np.random.rand(5, 3)
df = pd.DataFrame(d)

df.iloc[:3] = df.iloc[[2, 0, 1]].values

print(df)

df After:

          0         1         2
0  0.765908  0.518418  0.296801
1  0.221993  0.870732  0.206719
2  0.918611  0.488411  0.611744
3  0.187721  0.080741  0.738440
4  0.441309  0.158310  0.879937

edited May 29, 2021 at 22:45

answered May 29, 2021 at 22:24

Henry Ecker♦

35.8k19 gold badges48 silver badges67 bronze badges

4 Comments

wsdzbm Over a year ago

How about the index? The index doesn't move with the row. and manually set df.index.values[:3] = df.index.values[[2,0,1]] doesn't work

Henry Ecker Over a year ago

You'll either need to join the pieces as I've noted in my edit, or create a new index based on the other answer. See the edits. @ddzzbbwwmm

wsdzbm Over a year ago

For me, it's strange df.iloc[[2,0,1]] returns a dataframe with index also moved (as shown in your post), but the assignment only changes the values, not the index.

Henry Ecker Over a year ago

That's the point. iloc aligns indexes together. There are a ton of pandas operations where you want the order of your dataframe not to change when adding a new column (for example). You are specifically trying to get around one of the core data integrity features of pandas DataFrames.

Collectives™ on Stack Overflow

moving a row of a pandas dataframe but index unchanged

3 Answers 3

Comments

1 Comment

4 Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

3 Answers 3

Comments

1 Comment

4 Comments

Your Answer

Sign up or log in

Post as a guest

Related