container

Go containers

This module contains minimal type-safe Ordered Map, Queue, Set and Stack implementations using Go generics.

The Ordered Map supports both stable (in-place) updates and recency-based ordering, making it suitable both for highest performance (in-place), and for LRU caches (recency).

Contents

See the available types by underlying storage

CAVEAT: In order to optimize performance, except for WaitableQueue, all of these implementations are unsafe for concurrent execution, so they need protection in concurrency situations.

WaitableQueue being designed for concurrent code, on the other hand, is concurrency-safe.

Generally speaking, in terms of performance:

• Slice > list+internal pool > plain List > list+sync.Pool
• Preallocated > not preallocated

See BENCHARKS.md for details.

Usage

See complete listings in:

• cmd/orderedmap
• cmd/queuestack
• cmd/set

Ordered Map

stable := true
om := orderedmap.NewSlice[Key, Value](sizeHint, stable) // OrderedMap and Countable
om.Store(k, v)
om.Range(func (k K, v V) bool { fmt.Println(k, v); return true })
v, loaded := om.Load(k)
if !loaded {
        fmt.Fprintf(w, "No entry for key %v\n", k)
}
om.Delete(k) // Idempotent: does not fail on nonexistent keys.

Classic Queues without flow control

var e Element
q := queue.NewSliceQueue[Element](sizeHint)
q.Enqueue(e)
if lq, ok := q.(container.Countable); ok {
        fmt.Fprintf(w, "elements in queue: %d\n", lq.Len())
}
for i := 0; i < 2; i++ {
        e, ok := q.Dequeue()
        fmt.Fprintf(w, "Element: %v, ok: %t\n", e, ok)
}

WaitableQueue: a concurrent queue with flow control

var e Element
q, _ := queue.NewWaitableQueue[Element](sizeHint, lowWatermark, highWatermark)
go func() {
        wqs := q.Enqueue(e)
        if lq, ok := q.(container.Countable); ok {
                fmt.Fprintf(w, "elements in queue: %d, status: %s\n", lq.Len(), wqs)
        }
}
<-q.WaitChan()                    // Wait for elements to be available to dequeue
for i := 0; i < 2; i++ {          // Then dequeue them
        e, ok, wqs := q.Dequeue() // Non-blocking, ok will be true for the first and false for the second 
        fmt.Fprintf(w, "Element: %v, ok: %t, status: %s\n", e, ok, wqs)
}
q.Close() // Only needed if consumers may still be waiting on <-q.WaitChan

Sets

var e Element
s := set.NewBasicMap[Element](sizeHint)
s.Add(e)
s.Add(e)
if cs, ok := q.(container.Countable); ok {
        fmt.Fprintf(w, "elements in set: %d\n", cs.Len()) // 1
}
for e := range s.Items() {
        fmt.Fprintln(w, e)
}

Stacks

s := stack.NewSliceStack[Element](sizeHint)
s.Push(e)
if ls, ok := s.(container.Countable); ok {
        fmt.Printf("elements in stack: %d\n", ls.Len())
}
for i := 0; i < 2; i++ {
        e, ok := s.Pop()
        fmt.Printf("Element: %v, ok: %t\n", e, ok)
}

Running tests

Normal tests: unit and benchmarks

The complete test coverage requires running not only the unit tests, but also the benchmarks, like:

    go test -race -run=. -bench=. -coverprofile=cover.out -covermode=atomic ./...

This will also run the fuzz tests in unit test mode, without triggering the fuzzing logic.

Fuzz tests

Fuzz tests are not run by CI, but you can run them on-demand during development with:

    go test -run='^$' -fuzz='^\QFuzzBasicMapAdd\E$'   -fuzztime=20s ./set
    go test -run='^$' -fuzz='^\QFuzzBasicMapItems\E$' -fuzztime=20s ./set
    go test -run='^$' -fuzz='^\QFuzzBasicMapUnion\E$' -fuzztime=20s ./set

• Adjust -fuzztime duration as relevant: 20 seconds is just a smoke test.
• Be sure to escape the ^$ and \Q\E characters in the -fuzz argument in your shell.