Skip to content

Table of Contents

Exercises — Month 6 · Week 3 (Performance, profiling & hardening)

Standard-library-only Go packages with table-driven tests. Run all:

go test ./exercises/month-06/week-3/...
# allocation-sensitive tests are most reliable with the optimizer on (the default):
go test -race ./exercises/month-06/week-3/...

Prompts

1. tokenbucket — deterministic rate limiter

Implement a token-bucket Limiter with an injected clock (now func() time.Time) so refills are exact and testable without time.Sleep. It starts full (burst tokens), refills lazily at rate tokens/sec capped at burst, and Allow/AllowN consume tokens. The clock injection is the key idea — it makes time a value you control in tests.

  • Skills: lazy refill math, dependency injection for testability, floating-point bucket accounting.

2. allocfree — kill allocations with the append pattern

Implement AppendCSV(dst, fields) and AppendInt(dst, n) that write into a caller-provided buffer instead of returning a fresh string, so they allocate zero times when the buffer has capacity. Prove it with testing.AllocsPerRun. This mirrors the stdlib's own strconv.AppendInt / Time.AppendFormat style.

  • Skills: append-into-buffer pattern, strconv.AppendInt, testing.AllocsPerRun, reasoning about what escapes to the heap.

3. bufpool — sync.Pool used safely

Wrap a sync.Pool of *bytes.Buffer. Get must return a Reset buffer (a pooled object is not zeroed); Greet builds a string and returns it after Put (the string copies, so it doesn't alias pooled memory). Tests force reuse to catch a missing Reset and run concurrently for -race.

  • Skills: sync.Pool Get/Put/Reset discipline, why bytes.Buffer.String is safe after Put, concurrency hygiene.

Results

Exercise Focus go test
tokenbucket injected clock, lazy refill, burst cap ✅
allocfree append pattern, zero-alloc, AllocsPerRun ✅
bufpool sync.Pool Reset discipline, race-safe ✅