1. Overview

In this quick guide, we're going to see how to iterate over rows in Pandas DataFrame.

Pandas offer several different methods for iterating over rows like:

This article will explain the most common ways.

Note:

Have in mind that iterating over rows is pretty slow operation and not needed in most cases.

There is even warning on Pandas docs:

Iterating through pandas objects is generally slow.

2. Setup

In the article, we'll use the small DataFrame, which consists of several rows:

import pandas as pd

df = pd.read_csv('https://raw.githubusercontent.com/softhints/Pandas-Tutorials/master/data/csv/extremes.csv')

DataFrame looks like:

Continent Highest point Elevation high Lowest point Elevation low
Asia Mount Everest 8848 Dead Sea −427
South America Aconcagua 6960 Laguna del Carbón −105
North America Denali 6198 Death Valley −86
Africa Mount Kilimanjaro 5895 Lake Assal −155
Europe Mount Elbrus 5642 Caspian Sea −28

3. Iterate Using iterrows()

Let's start by method iterrows() from the DataFrame class which iterates over rows and returns pairs of (index, Series).

This is the most popular way for iteration in Pandas DataFrame:

for index, row in df.iterrows():
    print(index, row)

This will result into index of the first row and all values, then the second etc:

0 Continent                  Asia
Highest point     Mount Everest
Elevation high             8848
Lowest point           Dead Sea
Elevation low              −427
Name: 0, dtype: object
1 Continent             South America
Highest point             Aconcagua
Elevation high                 6960
Lowest point      Laguna del Carbón
Elevation low                  −105
Name: 1, dtype: object

Row values are accessible with bracket notation: row['Continent']

for index, row in df.iterrows():
    print(index, row['Continent'], row['Elevation high'])

The output is:

0 Asia 8848
1 South America 6960
2 North America 6198
3 Africa 5895
4 Europe 5642
5 Antarctica 4892
6 Australia 4884

The image below demonstrates how the method works:

4. Using df.itertuples()

Another method which iterates over rows is: df.itertuples().

df.itertuples is a faster for iteration over rows in Pandas.

To loop over all rows in a DataFrame by itertuples() use the next syntax:

for row in df.itertuples():
      print(row)

this will result into(all rows are returned as namedtuples):

Pandas(Index=0, Continent='Asia', _2='Mount Everest', _3=8848, _4='Dead Sea', _5='−427')
Pandas(Index=1, Continent='South America', _2='Aconcagua', _3=6960, _4='Laguna del Carbón', _5='−105')

In order to access only the first row we need to use next(iter( - because generator is returned by this method:

next(iter(df.itertuples(index=True, name='Point')))

output:

Point(Index=0, Continent='Asia', _2='Mount Everest', _3=8848, _4='Dead Sea', _5='−427')
Note:

itertuples() have the parameter `name`. If it's missing then the default value is Pandas.

Accessing row data with itertuples() is available by indices - integers or slices:

for row in df.itertuples(index=True, name='Point'):
      print(row[3], row[2])

rows are returned as:

8848 Mount Everest
6960 Aconcagua
6198 Denali
Note:

namedtuples are subclasses of tuples. You can think of them like something between dict and tuple. It adds more features in comparison to tuples.

Check more for namedtuples: collections.namedtuple

5. Faster Iteration over rows

For bigger datasets a faster solution is required. There are many options available if you need to speed up the loop over rows.

For example you can use frameworks like:

  • Dask
  • Modin
    and others like: Vaex, Ray, RAPIDS

Best for pure Pandas is to use vectorization for your operations.

Another option for processing all rows is list comprehensions. In the example below you can iterate over each row and get values for 2 columns:

[print(x, y) for x, y in zip(df['Continent'], df['Highest point'])]

result:

Asia Mount Everest
South America Aconcagua
North America Denali

or applying some function:

def func(x, y):
    return x + ' : ' +  y

result = [func(x, y) for x, y in zip(df['Continent'], df['Highest point'])]
result

result:

['Asia : Mount Everest',
 'South America : Aconcagua',
 ...
 'Antarctica : Vinson Massif',
 'Australia : Puncak Jaya']

6. Conclusion

In this post, we looked at different ways for iterating over rows in Pandas. We focused on basic functionality, but also compared the advantages of different methods.

In general for beginners and medium datasets - df.iterrows() is the way to go. For bigger datasets and more advanced users - itertuples(), list comprehension or custom solution are better.

As usual the code examples are available on GitHub.