Functions

A function is a named block of code you can run again and again. Write the logic once, call it from ten places. This is how scripts turn into software.

def, parameters, return

def to_cad(usd, rate):
    return round(usd * rate)

print(to_cad(95000, 1.35))       # 128250

• def starts the function. The name follows the same lowercase_underscore style as variables.
• usd and rate are parameters. The values you pass in are arguments.
• return sends a value back. Without it, the function returns None.

Calling to_cad(95000, 1.35) does not print anything by itself. It gives you a value. You decide what to do with it: print it, store it, pass it on.

Default values

def to_cad(usd, rate=1.35):
    return round(usd * rate)

to_cad(95000)            # uses 1.35
to_cad(95000, 1.40)      # overrides it
to_cad(usd=95000)        # keyword argument, clearer at the call site

Real libraries lean on defaults heavily. pd.read_csv("file.csv", sep=";") is a function call with one required argument and dozens of optional ones.

Small functions, clear names

A good function does one thing and says what it does:

def clean_name(name):
    return name.strip().lower()

def is_target_market(country):
    return country in ["Canada", "Germany", "Poland"]

def average(numbers):
    return sum(numbers) / len(numbers)

Three lines each. When a bug appears, you know exactly where to look. In an Airflow DAG, each task usually calls one such function.

Docstrings

A string right under def explains what the function does. Editors and AI assistants read it.

def to_cad(usd, rate=1.35):
    """Convert a USD amount to CAD, rounded to whole dollars."""
    return round(usd * rate)

lambda: the one-line function

A lambda is a tiny anonymous function. You mostly see it passed into other functions:

offers = [("Shopify", 95000), ("Wealthsimple", 110000), ("Startup", 68000)]
sorted(offers, key=lambda o: o[1])            # sort by salary

lambda o: o[1] means “take o, return o[1]”. You will meet it in sorted(), map(), pandas .apply(), and Spark UDFs. Read it, write it when short. If it grows past one line, use def.