How to iterate faster over rows in a DataFrame?

Question

I have a DataFrame from Pandas:

import pandas as pd
data = [{'c1':'aaa', 'c2':100, 'c3': 99, 'c4': 0}, {'c1':'bbb','c2':110, 'c3': 89, 'c4': 0}, 
{'c1':'aaa','c2':NaN,'c3': 93, 'c4': 0},{'c1':'ccc', 'c2':130,'c3': 77, 'c4': 0}, 
{'c1':'ddd','c2':140,'c3': 54, 'c4': 0}, {'c1':'bbb','c2':NaN,'c3': 76, 'c4': 0},
{'c1':'ddd', 'c2':NaN,'c3': 75, 'c4': 0}]
df = pd.DataFrame(data)
print df

Output:

   c1    c2   c3  c4
0 'aaa'  100  99  0
1 'bbb'  110  89  0
2 'aaa'  100  93  0
3 'ccc'  130  77  0
4 'ddd'  140  54  0
5 'bbb'  110  76  0
6 'ddd'  140  75  0

Now, I want for every row that matches the column c1, set the column c4 equals than the column c2 of the another row that matches the first field. The result:

   c1    c2   c3  c4
0 'aaa'  100  99  0
1 'bbb'  110  89  0
2 'aaa'  100  93  100
3 'ccc'  130  77  0
4 'ddd'  140  54  0
5 'bbb'  110  76  110
6 'ddd'  140  75  140

This dataframe is an example, the real dataframe has more columns and much more rows (around 4 million). My initial idea was this:

for index, row in df.iterrows(): 
    df[df.c1==row.c1].iloc[1].c4= row.c2

There can only be another matching row. Obviously, using iterrows the process is extremely slow.

Answer 1

Based on your latest edit,you can fillna with df.groupby followed by shift which will shift values 1 row down following the group:

df['c4'] = df.groupby("c1")['c2'].shift().fillna(df['c4'])

---

      c1   c2  c3     c4
0  'aaa'  100  99    0.0
1  'bbb'  110  89    0.0
2  'aaa'  100  93  100.0
3  'ccc'  130  77    0.0
4  'ddd'  140  54    0.0
5  'bbb'  110  76  110.0
6  'ddd'  140  75  140.0