Pandas with SQL databases

Introduction

Pandas helps you easily get data from SQL databases and work with it in Python. This makes data analysis faster and simpler.

You want to analyze data stored in a SQL database using Python.

You need to run SQL queries and then process the results with pandas.

You want to save your pandas DataFrame back into a SQL database.

You are combining SQL data with other data sources in Python.

You want to quickly explore database tables without leaving Python.

Syntax

Pandas

import pandas as pd
from sqlalchemy import create_engine

# Create a database connection
engine = create_engine('sqlite:///mydatabase.db')

# Read SQL query or table into a DataFrame
df = pd.read_sql('SELECT * FROM tablename', engine)

# Write DataFrame back to SQL table
df.to_sql('new_table', engine, if_exists='replace', index=False)

Use create_engine from SQLAlchemy to connect to your database.

pd.read_sql() can take a SQL query or a table name to load data.

Examples

This loads all rows from the 'users' table in the SQLite database 'example.db' and prints the first 5 rows.

Pandas

import pandas as pd
from sqlalchemy import create_engine

engine = create_engine('sqlite:///example.db')
df = pd.read_sql('SELECT * FROM users', engine)
print(df.head())

This saves the DataFrame df into a new table called 'backup_users' in the database, replacing it if it exists.

Pandas

df.to_sql('backup_users', engine, if_exists='replace', index=False)

This runs a SQL query to get names and ages of users older than 30, then prints the result.

Pandas

query = 'SELECT name, age FROM users WHERE age > 30'
df = pd.read_sql(query, engine)
print(df)

Sample Program

This program creates a small employees table in a temporary database, loads it into pandas, updates salaries for IT staff, saves changes back, and reads the updated table.

Pandas

import pandas as pd
from sqlalchemy import create_engine

# Create an in-memory SQLite database
engine = create_engine('sqlite:///:memory:')

# Create a sample table and insert data
with engine.connect() as conn:
    conn.execute("""
    CREATE TABLE employees (
        id INTEGER PRIMARY KEY AUTOINCREMENT,
        name TEXT,
        department TEXT,
        salary INTEGER
    )
    """)
    conn.execute("""
    INSERT INTO employees (name, department, salary) VALUES
    ('Alice', 'HR', 70000),
    ('Bob', 'IT', 80000),
    ('Charlie', 'Finance', 75000)
    """)

# Read the table into a pandas DataFrame
df = pd.read_sql('SELECT * FROM employees', engine)

# Show the DataFrame
print(df)

# Increase salary by 10% for IT department
df.loc[df['department'] == 'IT', 'salary'] = df.loc[df['department'] == 'IT', 'salary'] * 1.10

# Save updated data back to SQL
df.to_sql('employees', engine, if_exists='replace', index=False)

# Read again to confirm changes
df_updated = pd.read_sql('SELECT * FROM employees', engine)
print(df_updated)

Important Notes

Make sure to install SQLAlchemy with pip install sqlalchemy to use create_engine.

When writing DataFrames back to SQL, use if_exists='replace' to overwrite or if_exists='append' to add data.

SQLite is great for testing, but for bigger projects use databases like PostgreSQL or MySQL.

Summary

Pandas can read from and write to SQL databases easily using read_sql and to_sql.

Use SQLAlchemy's create_engine to connect pandas to your database.

This lets you combine SQL's power with pandas' easy data analysis in Python.

Practice

1. What is the main purpose of using pandas.read_sql() when working with SQL databases?

easy

A. To read data from a SQL database into a pandas DataFrame

B. To write a pandas DataFrame into a SQL database

C. To create a new SQL database

D. To delete tables from a SQL database

Solution

1. Step 1: Understand the function purpose

   pandas.read_sql() is designed to read data from a SQL query or table.

2. Step 2: Match function to action

   Reading data means loading it into a pandas DataFrame for analysis.

3. Final Answer:

   To read data from a SQL database into a pandas DataFrame -> Option A

4. Quick Check:

   read_sql() = read data [OK]

Hint: read_sql reads data into DataFrame fast [OK]

Common Mistakes:

• Confusing read_sql with to_sql
• Thinking it creates or deletes databases
• Assuming it writes data instead of reading

2. Which of the following is the correct way to create a connection engine to a SQLite database named data.db using SQLAlchemy for pandas?

easy

A. engine = create_engine('sqlite://data.db')

B. engine = create_engine('sqlite:///data.db')

C. engine = create_engine('sqlite:data.db')

D. engine = create_engine('sqlite:/data.db')

Solution

1. Step 1: Recall SQLite connection string format

   SQLite uses the format 'sqlite:///' followed by the database file path.

2. Step 2: Match correct syntax

   engine = create_engine('sqlite:///data.db') uses 'sqlite:///data.db' which is the correct format for a local file.

3. Final Answer:

   engine = create_engine('sqlite:///data.db') -> Option B

4. Quick Check:

   SQLite engine needs three slashes [OK]

Hint: SQLite needs three slashes in connection string [OK]

Common Mistakes:

• Using two slashes instead of three
• Missing slashes or colon
• Confusing with other database connection formats

3. Given the following code, what will be the output of print(df.shape)?

import pandas as pd
from sqlalchemy import create_engine
engine = create_engine('sqlite:///test.db')
df = pd.read_sql('SELECT * FROM users WHERE age > 30', engine)

medium

A. (0, 0) because no data is returned

B. (total_rows_in_users_table, number_of_columns_in_users_table)

C. An error because of incorrect SQL syntax

D. (number_of_rows_with_age_greater_than_30, number_of_columns_in_users_table)

Solution

1. Step 1: Understand the SQL query

   The query selects all columns from 'users' where age is greater than 30, so only matching rows are returned.

2. Step 2: Understand pandas DataFrame shape

   The DataFrame shape is (rows, columns). Rows = filtered rows, columns = all columns from 'users'.

3. Final Answer:

   (number_of_rows_with_age_greater_than_30, number_of_columns_in_users_table) -> Option D

4. Quick Check:

   Filtered rows and all columns = shape [OK]

Hint: DataFrame shape = filtered rows, all columns [OK]

Common Mistakes:

• Assuming all rows are returned
• Thinking shape returns data content
• Confusing SQL syntax with pandas syntax

4. What is wrong with this code snippet that tries to write a DataFrame df to a SQL table named sales?

from sqlalchemy import create_engine
engine = create_engine('sqlite:///sales.db')
df.to_sql('sales', engine)

medium

A. Missing the argument if_exists to handle existing tables

B. The engine connection string is incorrect

C. to_sql cannot write to SQLite databases

D. DataFrame df must be converted to a list first

Solution

1. Step 1: Check to_sql default behavior

   By default, to_sql will raise an error if the table exists unless if_exists is set.

2. Step 2: Identify missing argument

   The code lacks if_exists='replace' or if_exists='append' to avoid errors on existing tables.

3. Final Answer:

   Missing the argument if_exists to handle existing tables -> Option A

4. Quick Check:

   to_sql needs if_exists for existing tables [OK]

Hint: Add if_exists='replace' or 'append' to avoid errors [OK]

Common Mistakes:

• Assuming to_sql always overwrites
• Thinking engine string is wrong
• Believing DataFrame must convert before writing

5. You want to load data from a SQL database into pandas, filter rows where status = 'active', and then write only these filtered rows back to a new SQL table named active_users. Which sequence of commands correctly achieves this?

hard

A. df = pd.read_sql('SELECT * FROM users WHERE status = "active"', engine)
df.to_sql('active_users', engine, if_exists='append')

B. df = pd.read_sql('SELECT * FROM users WHERE status = active', engine)
df.to_sql('active_users', engine)

C. df = pd.read_sql('SELECT * FROM users', engine)
df_filtered = df[df['status'] == 'active']
df_filtered.to_sql('active_users', engine, if_exists='replace')

D. df = pd.read_sql('SELECT * FROM users', engine)
df_filtered = df[df.status = 'active']
df_filtered.to_sql('active_users', engine)

Solution

1. Step 1: Read all data and filter in pandas

   df = pd.read_sql('SELECT * FROM users', engine) df_filtered = df[df['status'] == 'active'] df_filtered.to_sql('active_users', engine, if_exists='replace') reads all users, then filters rows where status equals 'active' using correct pandas syntax.

2. Step 2: Write filtered data with proper if_exists argument

   df = pd.read_sql('SELECT * FROM users', engine) df_filtered = df[df['status'] == 'active'] df_filtered.to_sql('active_users', engine, if_exists='replace') writes filtered data to 'active_users' table, replacing if it exists, avoiding errors.

3. Final Answer:

   df = pd.read_sql('SELECT * FROM users', engine) df_filtered = df[df['status'] == 'active'] df_filtered.to_sql('active_users', engine, if_exists='replace') -> Option C

4. Quick Check:

   Read all, filter in pandas, write with if_exists [OK]

Hint: Filter in pandas, use if_exists='replace' when writing [OK]

Common Mistakes:

• Incorrect SQL syntax in WHERE clause
• Using single equals (=) in pandas filter
• Missing if_exists argument causing errors