12.18 Best Practices

When to Use Sets

• Automatic deduplication is needed.
• Fast, repeated membership testing is needed.
• Mathematical set operations (union, intersection, difference) are needed.

Efficient Membership Tests

If a collection is checked with in more than a handful of times, convert it to a set once up front rather than testing against a list repeatedly (see 12.13 Performance).

# Less efficient -- O(n) per check, repeated many times
allowed = ["us-east-1", "us-west-2", "eu-west-1"]
for region in incoming_regions:
    if region in allowed:      # O(n) every single time
        ...

# More efficient -- convert once, O(1) per check afterward
allowed_set = set(allowed)
for region in incoming_regions:
    if region in allowed_set:  # O(1) average every time
        ...

Choosing Set vs. List vs. Frozenset

Quick Interview Answer

“The decision to reach for a set comes down to two questions: does this need to be deduplicated automatically, and will it be checked with in more than once or twice? If either is true, a set is the right default — and the moment a collection is checked repeatedly against a fixed set of allowed values, converting it to a set exactly once up front is one of the cheapest performance wins available. The one caveat: if that immutable, fixed collection itself needs to be a dict key or live inside another set, frozenset is the correct choice, not a regular set, since a plain set can never be hashed.”

Common Mistakes

• Testing membership against a list inside a loop that runs many times, instead of converting to a set once before the loop starts.
• Reaching for a set when the order elements were added actually matters for later output — a set silently discards that information.
• Using a regular set as a dict key or as a member of another set and hitting TypeError: unhashable type: 'set' — frozenset is the type built for exactly that use case (see 12.11 Frozen Sets).