Sorting with key

sorted() never compares your objects directly — it compares the value your key= function produces. One idea, and you can sort anything:

words = ["banana", "fig", "cherry"]
sorted(words, key=len)              # by length: fig, banana, cherry

people = [
    {"name": "An", "age": 30},
    {"name": "Binh", "age": 25},
]
sorted(people, key=lambda p: p["age"])

Descending and multi-level sorting come free:

scores = [("An", 8), ("Binh", 9), ("Chi", 8)]
sorted(scores, key=lambda p: p[1], reverse=True)      # highest first

# sort by age, then by name — return a tuple from key:
sorted(people, key=lambda p: (p["age"], p["name"]))

Two facts professionals rely on:

• sorted() returns a new list; list.sort() mutates in place (and returns None — a classic bug).
• Sorts are stable: equal keys keep their original order, so you can sort by secondary key first, then primary key.

Why not compare functions?

key= decouples "what to compare" from "how to compare". Sorting records by a field, files by size, tasks by deadline — all become one-liners instead of hand-rolled comparison logic.