Thinking ahead: plan the contract before the code

Give the operation a clear contract

An original trip-planning task estimates the fare for a group travelling to a museum. State the data type, units and meaning of each value: “a number” leaves too many assumptions unstated.

An output may be a displayed message, returned value, saved record or control signal. State which one is needed. Here, the calculation returns pence; a separate display component can format that result as pounds.

Check the state the operation depends on

Preconditions can describe the wider state, not just user input. Binary search requires the collection to be sorted according to its comparison rule. File reading may require an existing, readable file.

Cache the result of repeated work

A cache associates a key with a stored result. A matching request is a hit; an absent entry is a miss, so the original work runs and its result can be stored. A bounded cache eventually removes entries to stay within its limit.

This complete Python example demonstrates the mechanism using a tiny lookup. The lookup itself is cheap: caching this particular dictionary access is not a useful speed optimisation. A real route-planning calculation or a slow external lookup could justify the same pattern.

from functools import lru_cache

# Each version is a fixed price list, in pence.
prices = {"autumn": {"museum": 150}, "winter": {"museum": 180}}

@lru_cache(maxsize=4)
def fare_pence(destination, version):
    print("Lookup:", destination, version)
    return prices[version][destination]

print(fare_pence("museum", "autumn"))
print(fare_pence("museum", "autumn"))
print(fare_pence("museum", "winter"))

The exact output is Lookup: museum autumn, 150, 150, Lookup: museum winter, then 180, each on a new line. lru_cache uses the function arguments as the key; they must be hashable. Its maximum of four entries limits retained results.

Keep cached results fresh enough

The version works because each version names an unchanging price list. Changing prices["autumn"]["museum"] after the first call leaves the cached autumn result at 150.

• Clear or replace the affected cache.
• Use a new version key when the underlying data changes.
• Choose an expiry rule appropriate to the freshness requirement. Expiry limits staleness; it does not ensure every result is current.

Try it yourself

These four original questions total 16 marks. The allocations and answers are our indicative guidance, not an OCR mark scheme.