Intro to Pandas

pandas is Python's library for working with tabular data — anything you'd naturally think of as rows and columns, like a spreadsheet or a database table. It's built on top of NumPy, so its columns are stored as fast, typed arrays under the hood, but it adds the things NumPy doesn't have: labelled rows and columns, mixed column types in a single table, missing-value handling, and a huge set of convenience methods for reading, cleaning, filtering, and summarizing data.

Installing pandas

Bash

pip install pandas

DataFrame and Series

pandas has two core objects. A Series is a single labelled column of data — essentially a NumPy array with an index attached. A DataFrame is a full table: a collection of Series sharing the same row index, one per column.

Python

import pandas as pd

# A Series — one labelled column
ages = pd.Series([25, 32, 18], name="age")

# A DataFrame — built from a dict of columns
df = pd.DataFrame({
    "name": ["Ada", "Grace", "Alan"],
    "age": [36, 30, 41],
    "city": ["London", "New York", "London"],
})
print(df)

    name  age      city
0    Ada   36    London
1  Grace   30  New York
2   Alan   41    London

Reading data from a file

In practice, you rarely build a DataFrame by hand — you load it from a file, most commonly a CSV:

Python

import pandas as pd

df = pd.read_csv("employees.csv")

pandas also has readers for many other formats out of the box, including read_excel(), read_json(), and read_sql().

Exploring a DataFrame

Before doing anything with a new dataset, it's standard practice to get a quick feel for its shape and contents:

Python

df.head()      # first 5 rows — quick sanity check
df.info()      # column names, dtypes, non-null counts
df.describe()  # count, mean, std, min/max, quartiles for numeric columns

.head(n) — the first n rows (default 5). There is also .tail(n) for the last rows.
.info() — column dtypes and how many non-null values each column has, useful for spotting missing data quickly.
.describe() — summary statistics (mean, standard deviation, min, max, percentiles) for every numeric column.
.shape — a (rows, columns) tuple, just like a NumPy array.

Selecting columns and rows

pandas gives you three main ways to select data, and it's worth knowing when to use each:

Python

df["name"]           # a single column, returned as a Series
df[["name", "age"]]   # multiple columns, returned as a DataFrame

df.loc[0]             # row selected by label/index value
df.loc[0, "age"]      # a specific cell, by row label + column name

df.iloc[0]            # row selected by integer position
df.iloc[0:2, 0:2]      # first 2 rows, first 2 columns, by position

[] — quick column access (or a boolean mask, see below).
.loc — select by label (row index value, column name). Use this when your rows have meaningful labels, not just 0, 1, 2....
.iloc — select by integer position, the same way you would index a plain Python list.

Filtering with boolean conditions

The most common pandas idiom is filtering rows with a boolean condition inside [] — this is often called boolean masking:

Python

# Rows where age is greater than 30
df[df["age"] > 30]

# Combine conditions with & and | (note the parentheses!)
df[(df["age"] > 30) & (df["city"] == "London")]

Parentheses are required

Because of Python's operator precedence, each condition combined with & or | must be wrapped in its own parentheses, otherwise the expression will not evaluate the way you expect.

Note

pandas is the de facto standard tool for data analysis and cleaning in the Python ecosystem — if a tutorial, job posting, or data science notebook mentions "wrangling" or "cleaning" data in Python, it's almost certainly using pandas.