Selecting Rows and Columns

Before you can analyse data, you need to point at the part you care about. Pandas has three tools: square brackets for columns, loc for labels and iloc for positions.

We’ll use this table in every example:

import pandas as pd

df = pd.DataFrame({
    "name": ["Asha", "Rahul", "Meera", "Kabir", "Zoya"],
    "city": ["Pune", "Surat", "Delhi", "Pune", "Delhi"],
    "marks": [82, 64, 91, 58, 77],
    "age": [20, 21, 19, 22, 20],
})

Columns#

print(df["name"].tolist())
print(df[["name", "marks"]])

Output:

['Asha', 'Rahul', 'Meera', 'Kabir', 'Zoya']
    name  marks
0   Asha     82
1  Rahul     64
2  Meera     91
3  Kabir     58
4   Zoya     77

One name in brackets → a Series. A list of names → a DataFrame.

iloc: by position#

iloc works like NumPy: integer positions, starting at 0, end not included.

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

Output:

name     Asha
city     Pune
marks      82
age        20
Name: 0, dtype: object
    name   city
1  Rahul  Surat
2  Meera  Delhi

loc: by label#

loc uses the index labels and column names. Unlike iloc, the end of a label slice is included:

print(df.loc[2, "name"])
print(df.loc[1:3, ["name", "city"]])

Output:

Meera
    name   city
1  Rahul  Surat
2  Meera  Delhi
3  Kabir   Pune

It’s even more useful when the index has meaningful labels:

people = df.set_index("name")
print(people.loc["Zoya"])
print(people.loc[["Asha", "Meera"], "marks"])

Output:

city     Delhi
marks       77
age         20
Name: Zoya, dtype: object
name
Asha     82
Meera    91
Name: marks, dtype: int64

Changing a value safely#

Use loc with both row and column to update a cell:

df.loc[3, "marks"] = 61
print(df.loc[3])

Output:

name     Kabir
city      Pune
marks       61
age         22
Name: 3, dtype: object

Avoid

df["marks"][3] = 61 (two separate brackets) may change a temporary copy instead of your table, and Pandas warns about it. Always use df.loc[row, column] = value.

Summary#

You want	Code
One column	df["col"]
Several columns	df[["a", "b"]]
Rows by position	df.iloc[0:3]
Rows/columns by label	df.loc[1:3, ["a", "b"]] (end included)
Change a cell	df.loc[row, "col"] = value