24. How to Load SQL Query Results into a pandas DataFrame

24.5 Filter in SQL vs Filter in pandas

By Ravindra Bagale · Last updated: 23 October 2026

Why. You can WHERE city = 'Pune' in SQL, or load more rows and df.loc[df["city"] == "Pune"] in pandas. Wrong choice either floods memory or forces ten round-trips.

What.

• Filter in SQL when the database can drop rows early (city, date, status) and the table is large.
• Filter in pandas for flexible exploration, multiple what-if cuts, and logic that is awkward in SQL for you today.
• Aggregate in SQL (GROUP BY) when you only need a summary and the raw grain is huge.
• Aggregate in pandas when you will re-slice the same extract many ways (Chapter 7 groupby).

How.

Worked example A — Same answer, two places.

import sqlite3
import pandas as pd

con = sqlite3.connect("data/blinkit_lab.db")

sql_filtered = """
SELECT city, amount_inr
FROM orders
WHERE status = 'Delivered' AND city IN ('Pune', 'Nashik')
"""
df_sql = pd.read_sql_query(sql_filtered, con)

df_all = pd.read_sql_query(
    "SELECT city, status, amount_inr FROM orders",
    con,
)
con.close()

df_pd = df_all.loc[
    (df_all["status"] == "Delivered")
    & (df_all["city"].isin(["Pune", "Nashik"])),
    ["city", "amount_inr"],
]

print("SQL filter GMV", df_sql["amount_inr"].sum())
print("pandas filter GMV", df_pd["amount_inr"].sum())

Worked example B — Heavy table habit.

Zoya’s orders table has millions of rows. She never read_sql_table("orders"). She asks SQL for one festival week and six Maharashtra cities, then plots in seaborn.

Worked example C — Decision card.