Documentation ¶
Index ¶
- func RunLayers(ctx context.Context, layers ...Layer) error
- type EventCallback
- type Layer
- type LayerFunc
- type LayerGoFunc
- type Listen
- type LogWhenDoneLayer
- type Mutex
- type SiblingLayers
- type Subscriptions
- type ValueHolder
- type WaitGroup
- type WorkSource
- type Worker
- type WorkerFactory
- type WorkerFunc
- type WorkerHolder
- type WorkerManager
- func NewProtoWorkerManager[ID comparable, C proto.Message](log *slog.Logger, fld func(ID) slog.Attr, workerFactory WorkerFactory[ID, C]) *WorkerManager[ID, C]
- func NewWorkerManager[ID comparable, C any](log *slog.Logger, fld func(ID) slog.Attr, workerFactory WorkerFactory[ID, C], ...) *WorkerManager[ID, C]
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
Types ¶
type EventCallback ¶
type Layer ¶ added in v17.4.0
Layer is a layer in a startup and shutdown stack. Each layer can expand into zero or more stages that are run first to last and shut down in reverse order.
Example ¶
|---------| | Layer A | |---------| | Layer B | |---------|
Let's say the above layers expand into the following:
|----------| | Stage A1 | |----------| | Stage A2 | |----------| | Stage B1 | |----------| | Stage B2 | |----------|
Stages are shut down in bottom up order: B2, B1, A2, A1.
While stages are started in the order they come, it doesn't matter much since they all start concurrently in independent goroutines and, hence, no real ordering is achieved. The purpose of this mechanism is to guarantee reverse shutdown order of the stages. This is ensured by the underlying github.com/ash2k/stager library. Layers provide more flexibility and composability on top of the library.
type LayerGoFunc ¶ added in v17.4.0
func (LayerGoFunc) ToStageFuncs ¶ added in v17.4.0
func (f LayerGoFunc) ToStageFuncs() ([]stager.StageFunc, error)
type Listen ¶ added in v17.1.0
type Listen[E any] func(cb EventCallback[E])
type LogWhenDoneLayer ¶ added in v17.4.0
func (*LogWhenDoneLayer) ToStageFuncs ¶ added in v17.4.0
func (l *LogWhenDoneLayer) ToStageFuncs() ([]stager.StageFunc, error)
type Mutex ¶
type Mutex struct {
// contains filtered or unexported fields
}
Mutex is a non-reentrant (like sync.Mutex) mutex that (unlike sync.Mutex) allows to acquire the mutex with a possibility to abort the attempt early if a context signals done.
A buffered channel of size 1 is used as the mutex. Think of it as of a box - the party that has put something into it has acquired the mutex. To unlock it, remove the contents from the box, so that someone else can use it. An empty box is created in the NewMutex() constructor.
TryLock, Lock, and Unlock provide memory access ordering guarantees by piggybacking on channel's "happens before" guarantees. See https://golang.org/ref/mem
type SiblingLayers ¶ added in v17.4.0
type SiblingLayers []Layer
SiblingLayers allows to compose multiple layers to be siblings rather than ordered on top of each other.
Example ¶
A is a simple layer and B and C are sibling layers. |---------|---------| | Layer A | |---------|---------| | Layer B | Layer C | |---------|---------|
Can expand into:
|----------|----------| | Stage A1 | |----------|----------| | Stage B1 | Stage C1 | | |----------| | | Stage C2 | |----------|----------|
Shutdown order: (B1 and (C2 then C1)) then A1. Where "and" means things happen concurrently and "then" means an ordering dependency.
func (SiblingLayers) ToStageFuncs ¶ added in v17.4.0
func (l SiblingLayers) ToStageFuncs() ([]stager.StageFunc, error)
type Subscriptions ¶
type Subscriptions[E any] struct { // contains filtered or unexported fields }
func (*Subscriptions[E]) Dispatch ¶
func (s *Subscriptions[E]) Dispatch(ctx context.Context, e E)
Dispatch dispatches the given event to all added subscriptions.
func (*Subscriptions[E]) Len ¶ added in v17.1.0
func (s *Subscriptions[E]) Len() int
Len returns the number of subscriptions.
func (*Subscriptions[E]) On ¶
func (s *Subscriptions[E]) On(ctx context.Context, cb EventCallback[E])
func (*Subscriptions[E]) Subscribe ¶ added in v17.1.0
func (s *Subscriptions[E]) Subscribe(ctx context.Context) Listen[E]
Subscribe subscribes to events but, unlike On, does not start listening immediately. It returns a function that must be called to listen to events. Make sure the function is always called as it performs cleanup once the passed context is done. The returned function must only be called once.
type ValueHolder ¶ added in v17.2.0
type ValueHolder[T comparable] struct { // contains filtered or unexported fields }
ValueHolder holds agent id of this agentk.
func NewValueHolder ¶ added in v17.2.0
func NewValueHolder[T comparable]() *ValueHolder[T]
func (*ValueHolder[T]) Get ¶ added in v17.2.0
func (h *ValueHolder[T]) Get(ctx context.Context) (T, error)
func (*ValueHolder[T]) Set ¶ added in v17.2.0
func (h *ValueHolder[T]) Set(value T) error
Set is not safe for concurrent use. It's ok since we don't need that.
func (*ValueHolder[T]) TryGet ¶ added in v17.2.0
func (h *ValueHolder[T]) TryGet() (T, bool)
type WaitGroup ¶ added in v17.5.0
type WaitGroup struct {
// contains filtered or unexported fields
}
func NewWaitGroup ¶ added in v17.5.0
func NewWaitGroup() *WaitGroup
type WorkSource ¶
type WorkSource[ID comparable, C any] struct { ID ID Configuration C }
type WorkerFactory ¶
type WorkerFactory[ID comparable, C any] interface { New(WorkSource[ID, C]) Worker }
type WorkerFunc ¶
func (WorkerFunc) Run ¶
func (wf WorkerFunc) Run(ctx context.Context)
type WorkerHolder ¶
type WorkerHolder[C any] struct { // contains filtered or unexported fields }
WorkerHolder holds a worker and restarts it when configuration changes.
func NewComparableWorkerHolder ¶
func NewComparableWorkerHolder[C comparable](factory func(C) Worker) *WorkerHolder[C]
func NewProtoWorkerHolder ¶
func NewProtoWorkerHolder[C proto.Message](factory func(C) Worker) *WorkerHolder[C]
func NewWorkerHolder ¶
func NewWorkerHolder[C any](factory func(C) Worker, isEqual func(config1, config2 C) bool) *WorkerHolder[C]
func (*WorkerHolder[C]) ApplyConfig ¶
func (w *WorkerHolder[C]) ApplyConfig(ctx context.Context, config C) bool
ApplyConfig ensures a worker is running with the provided or equal config.
This method starts a worker if it's not running already. If it is running and the config is not equal then the worker is stopped, a new worker is started then with the new config.
func (*WorkerHolder[C]) StopAndWait ¶
func (w *WorkerHolder[C]) StopAndWait()
type WorkerManager ¶
type WorkerManager[ID comparable, C any] struct { // contains filtered or unexported fields }
func NewProtoWorkerManager ¶
func NewProtoWorkerManager[ID comparable, C proto.Message](log *slog.Logger, fld func(ID) slog.Attr, workerFactory WorkerFactory[ID, C]) *WorkerManager[ID, C]
func NewWorkerManager ¶
func NewWorkerManager[ID comparable, C any](log *slog.Logger, fld func(ID) slog.Attr, workerFactory WorkerFactory[ID, C], equal func(c1, c2 C) bool) *WorkerManager[ID, C]
func (*WorkerManager[ID, C]) ApplyConfiguration ¶
func (m *WorkerManager[ID, C]) ApplyConfiguration(sources []WorkSource[ID, C]) error
func (*WorkerManager[ID, C]) Stop ¶ added in v17.4.0
func (m *WorkerManager[ID, C]) Stop()
Stop tells all workers to stop. It does not wait for workers to stop. Use Wait() for that.
func (*WorkerManager[ID, C]) StopAndWait ¶ added in v17.4.0
func (m *WorkerManager[ID, C]) StopAndWait()
func (*WorkerManager[ID, C]) Wait ¶ added in v17.4.0
func (m *WorkerManager[ID, C]) Wait()
Wait waits for all workers to stop.