Filtering and sorting rows

01The problem

The finance manager has three quick requests for Bisi:

"Which order lines had a 10% discount and more than 25 cartons? Show me our Lagos supermarkets. And what were the five biggest order lines in June 2026?"

Each one means keeping only the rows that meet a condition, and sometimes putting them in order. In Excel you'd click through filter menus and start again for the next question. In pandas each request is one line you can re-run, change and keep.

02The concept

A condition gives a True/False for every row

orders["discount_pct"] == 10

That doesn't filter anything yet: it returns a Series of 4,266 True/False values, one per row. It's called a boolean mask. Put the mask inside square brackets and pandas keeps the True rows:

big_discounts = orders[orders["discount_pct"] == 10]

Combining conditions

Each condition needs its own brackets, because & and | are worked out before == and >:

orders[(orders["discount_pct"] == 10) & (orders["quantity"] > 25)]

Without the inner brackets you get a confusing error about "truth value of a Series is ambiguous". When you see it, check your brackets.

Shortcuts for common conditions

• One of several values: customers["region"].isin(["Lagos", "South West"])
• A range, ends included: orders["quantity"].between(10, 20)
• Text: customers["customer_name"].str.contains("Wholesale"), .str.startswith("Ada"). Add case=False to ignore capitals.
• Missing values: .isna() and .notna().

Dates stored as text in YYYY-MM-DD form compare correctly as text, so orders["order_date"] >= "2026-06-01" works even before you convert dates (lesson 5).

Choosing rows and columns together: .loc

orders.loc[mask, ["order_id", "quantity"]] keeps the rows where the mask is True and only the columns you list. It's the clearest way to say "these rows, these columns", and the safe way to change values in a filtered part of a table.

Sorting

• orders.sort_values("quantity", ascending=False): largest first.
• Several columns: sort_values(["region", "credit_limit"], ascending=[True, False]).
• orders.nlargest(5, "quantity") and nsmallest are shortcuts for "sort and take the top 5".

Counting what's left

len(df) is the number of rows. Because True counts as 1, mask.sum() counts the matching rows without making a new table.

03Example

Request 1: 10% discount and more than 25 cartons.

import pandas as pd

base = "https://academy.cloudtechanalytics.com/datasets/sales/"
orders = pd.read_csv(base + "orders.csv")
customers = pd.read_csv(base + "customers.csv")

big_discounts = orders[(orders["discount_pct"] == 10) & (orders["quantity"] > 25)]
len(big_discounts)

95

Request 2: Lagos supermarkets, only the useful columns, biggest credit limits first.

lagos_supermarkets = customers.loc[
    (customers["region"] == "Lagos") & (customers["channel"] == "Supermarket"),
    ["customer_name", "city", "credit_limit"],
].sort_values("credit_limit", ascending=False)
lagos_supermarkets.head()

customer_name    city  credit_limit
38           Ada Supermarket   Ikeja       2000000
83        Olumide Superstore   Lekki       1950000
86  Yakubu Superstore Festac  Festac       1550000
22         Emeka Supermarket   Ikeja       1450000
68          Mama Nkechi Mart  Festac       1250000

Request 3: the five biggest order lines in June 2026, by quantity.

june_2026 = orders[orders["order_date"].between("2026-06-01", "2026-06-30")]
june_2026.nlargest(5, "quantity")

04Walkthrough

1. Load orders and customers as in the Example.
2. Run orders["discount_pct"] == 10 on its own and look at the result: a column of True and False. Then run (orders["discount_pct"] == 10).sum() to count the 10% lines (525).
3. Run the Request 1 filter, then remove the inner brackets and run it again to see the "ambiguous" error. Put them back.
4. Count lines that were not discounted: (~(orders["discount_pct"] > 0)).sum(). ~ flips the condition. (orders["discount_pct"] == 0 gives the same answer more simply.)
5. Use isin: customers[customers["region"].isin(["North West", "North Central"])] lists the northern customers.
6. Search text: customers[customers["customer_name"].str.contains("wholesale", case=False)].
7. Run Request 3 and check the dates in the result are all in June 2026.

05Practice

06Challenge

07More practice

Optional drills on a different dataset: the HR data. Load it with employees = pd.read_csv("https://academy.cloudtechanalytics.com/datasets/hr/employees.csv").

08Check your understanding