Python Caching: Choose Memoization Only for Pure Inputs

Only cache computations whose results remain valid for the same arguments.

What You Will Learn

Caching trades memory for time by reusing an earlier result. functools.cache or lru_cache are useful for deterministic computations with hashable inputs. Do not cache calls that read a changing clock, current exchange rate, database state or other hidden dependencies unless invalidation is deliberately managed.

Caching uses extra memory and can return stale results when a function relies on changing state outside its arguments.

Step-by-Step Example

The example below makes the behavior visible with small input values that you can change safely.

from functools import lru_cache

@lru_cache(maxsize=32)
def rectangle_area(width, height):
    print("Calculated")
    return width * height

print(rectangle_area(4, 5))
print(rectangle_area(4, 5))
print(rectangle_area.cache_info().hits)

How the Example Works

The imports provide the standard-library tools used in the demonstration: functools.
The function rectangle_area defines reusable behavior; its body executes when it is called, not when it is declared.
The print() calls make the return values or state visible so you can compare them with the explanation and check your prediction.

Expected output from the example:

Calculated
20
20
1

When You Would Use This

These tools are useful when composing small reusable functions, adapting callbacks, or eliminating repeated work. Pay attention to whether the operation returns a new object, updates state, or produces lazy results. Test a small success case and a boundary case before integrating it into an application.

Practice: Modify the Code

Call the function with different widths, then inspect misses and hits.

Verify the changed output against the behavior described above, including the relevant invalid or boundary case.

For complete parameters and version details, consult the official Python documentation.

Common Beginner Mistakes

Applying an unbounded cache to values that are nearly always unique.
Caching a function that reads changing external data and returning stale information.

Key Takeaways

Only cache computations whose results remain valid for the same arguments.
Caching trades memory for time by reusing an earlier result. functools.cache or lru_cache are useful for deterministic computations with hashable inputs.
Apply this operation only after checking the assumptions and input conditions shown in the example.