ReviseAlgo Logo

Functional Programming

functools Module

The functools standard library — partial application, reduce, lru_cache memoization, singledispatch, and wraps for building composable functional utilities.

Interview: reduce, partial, and lru_cache are common interview topics — shows mastery of Python functional tools.

Last Updated: June 12, 2026 • 10 min read

The functools module provides higher-order functions and tools for working with callable objects. Key functions include partial (partial application), reduce (fold), lru_cache (memoization), singledispatch (type-based dispatch), and wraps (preserving function metadata).

Key Functions

  • partial(func, *args) — freeze some arguments, returns a new callable
  • reduce(func, iterable) — apply function cumulatively to reduce to single value
  • @lru_cache(maxsize) — memoize function results for performance
  • @singledispatch — dispatch to different implementations based on type
  • @wraps(func) — preserve function name, docstring, and annotations

reduce (Fold)

  • reduce(lambda a, b: a + b, [1,2,3]) → 6 (sum)
  • Takes a binary function and applies it left-to-right across the iterable
  • Optional initializer parameter acts as default/seed value
  • Often a simple loop or sum()/max()/min() is more readable

lru_cache (Memoization)

  • Caches return values based on function arguments
  • maxsize=128 limits cache size (LRU eviction), maxsize=None is unbounded
  • Dramatically speeds up recursive functions and expensive computations
  • Only works with hashable arguments (no lists or dicts)

Interview Insight

Know when to use reduce vs a simple loop. lru_cache is the go-to solution for memoization in interviews (Fibonacci, dynamic programming). partial is useful for callback configuration.

Use Cases

Memoization — caching expensive computations with lru_cache

Callback configuration — partial to pre-fill arguments for event handlers

Data aggregation — reduce for custom folding operations

Type dispatch — singledispatch for polymorphic behavior without classes

Decorator writing — wraps to maintain function identity in decorators

Common Mistakes

Using reduce when sum()/max()/min() would be clearer and more Pythonic

Forgetting lru_cache only works with hashable arguments — no lists or dicts

Not setting maxsize on lru_cache for unbounded data — memory grows forever

Forgetting @wraps in decorators — breaks __name__, __doc__, and introspection

Using partial for everything when a lambda or nested function is clearer