demo · v143

Memoize cache

The original TC39 proposal motivation: turn the four-line "check, compute, set" memoization dance into one call. We memoize a pretend-slow Fibonacci with both styles side-by-side and count cache hits vs misses.

compute fib(n)

old: has + set dance

hits: 0
misses: 0
last call: —

new: getOrInsertComputed

hits: 0
misses: 0
last call: —

call log

the call

// Before: the four-line dance
function memoOld(n) {
  if (cache.has(n)) return cache.get(n);
  const v = compute(n);
  cache.set(n, v);
  return v;
}

// After: one line, factory only runs on miss
function memoNew(n) {
  return cache.getOrInsertComputed(n, compute);
}

why this angle

This is the canonical TC39 motivating example. The "check + compute + set" pattern is so common it inspired the whole proposal. Crucially, getOrInsertComputed only invokes the factory on a miss, so expensive computation is skipped on a hit — unlike map.get(k) ?? map.set(k, compute()).get(k) hacks that always run the right-hand side.

see also