split rows into columns in pandas dataframe

Question

I have table similar to this table:

     id      val  
0   abc_1    5
1   abc_1    3
2   abc_1    7
3   abc_2    12
4   abc_2    6
5   abc_2    9
...

I want to "split" the rows into columns based on the id, so the result should be:

     id      val1   val2   val3  
0   abc_1    5       3       7
1   abc_2    12      6       9

I was trying to do it by creating groups and then use pivot :

df['group'] = (df.index // 3) + 1

>>>
     id      val    group  
0   abc_1    5       1
1   abc_1    3       1
2   abc_1    7       1
3   abc_2    12      2
4   abc_2    6       2
5   abc_2    9       2

but then then when I was using pivot I got many new columns al with null values.

My goal is to get this desired output , open to other solutions as well.

…

     id      val1   val2   val3  
0   abc_1    5       3       7
1   abc_2    12      6       9

Asked By: Reut

||

Source

Answer 1

A possible solution:

(df.assign(val_names = [f'val{i}' for i in df.groupby('id').cumcount()+1])
 .pivot(index='id', columns='val_names', values='val')
 .rename_axis(None, axis=1)
 .reset_index())

Output:

      id  val1  val2  val3
0  abc_1     5     3     7
1  abc_2    12     6     9

Answered By: PaulS

Answer 2

You can use groupby + cumcount to determine which column a value should go into. Then pivot, then reset index and rename columns as demanded:

df["col_num"] = df.groupby("id").val.cumcount() + 1

df 
    .pivot(index="id", columns="col_num", values="val") 
    .reset_index() 
    .rename_axis(None, axis=1) 
    .rename({col: f"val_{col}" for col in df["col_num"]}, axis=1)

Output:

      id  val_1  val_2  val_3
0  abc_1      5      3      7
1  abc_2     12      6      9

Answered By: Maria K

split rows into columns in pandas dataframe

Question:

Answers: