ants/pool.go

// MIT License

// Copyright (c) 2018 Andy Pan

// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in all
// copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
// SOFTWARE.

package ants

import (
	"errors"
	"sync"
	"sync/atomic"
	"time"

	"github.com/panjf2000/ants/v2/internal"
)

// Pool accepts the tasks from client, it limits the total of goroutines to a given number by recycling goroutines.
type Pool struct {
	// capacity of the pool, a negative value means that the capacity of pool is limitless, an infinite pool is used to
	// avoid potential issue of endless blocking caused by nested usage of a pool: submitting a task to pool
	// which submits a new task to the same pool.
	capacity int32

	// running is the number of the currently running goroutines.
	running int32

	// lock for protecting the worker queue.
	lock sync.Locker

	// workers is a slice that store the available workers.
	workers workerArray

	// state is used to notice the pool to closed itself.
	state int32

	// cond for waiting to get an idle worker.
	cond *sync.Cond

	// workerCache speeds up the obtainment of a usable worker in function:retrieveWorker.
	workerCache sync.Pool

	// blockingNum is the number of the goroutines already been blocked on pool.Submit, protected by pool.lock
	blockingNum int

	stopHeartbeat chan struct{}

	options *Options
}

// purgePeriodically clears expired workers periodically which runs in an individual goroutine, as a scavenger.
func (p *Pool) purgePeriodically() {
	heartbeat := time.NewTicker(p.options.ExpiryDuration)
	defer heartbeat.Stop()

	for {
		select {
		case <-heartbeat.C:
		case <-p.stopHeartbeat:
			p.stopHeartbeat <- struct{}{}
			return
		}

		if p.IsClosed() {
			break
		}

		p.lock.Lock()
		expiredWorkers := p.workers.retrieveExpiry(p.options.ExpiryDuration)
		p.lock.Unlock()

		// Notify obsolete workers to stop.
		// This notification must be outside the p.lock, since w.task
		// may be blocking and may consume a lot of time if many workers
		// are located on non-local CPUs.
		for i := range expiredWorkers {
			expiredWorkers[i].task <- nil
			expiredWorkers[i] = nil
		}

		// There might be a situation that all workers have been cleaned up(no any worker is running)
		// while some invokers still get stuck in "p.cond.Wait()",
		// then it ought to wake all those invokers.
		if p.Running() == 0 {
			p.cond.Broadcast()
		}
	}
}

// NewPool generates an instance of ants pool.
func NewPool(size int, options ...Option) (*Pool, error) {
	opts := loadOptions(options...)

	if size <= 0 {
		size = -1
	}

	if expiry := opts.ExpiryDuration; expiry < 0 {
		return nil, ErrInvalidPoolExpiry
	} else if expiry == 0 {
		opts.ExpiryDuration = DefaultCleanIntervalTime
	}

	if opts.Logger == nil {
		opts.Logger = defaultLogger
	}

	p := &Pool{
		capacity:      int32(size),
		lock:          internal.NewSpinLock(),
		stopHeartbeat: make(chan struct{}, 1),
		options:       opts,
	}
	p.workerCache.New = func() interface{} {
		return &goWorker{
			pool: p,
			task: make(chan func(), workerChanCap),
		}
	}
	if p.options.PreAlloc {
		if size == -1 {
			return nil, ErrInvalidPreAllocSize
		}
		p.workers = newWorkerArray(loopQueueType, size)
	} else {
		p.workers = newWorkerArray(stackType, 0)
	}

	p.cond = sync.NewCond(p.lock)

	// Start a goroutine to clean up expired workers periodically.
	go p.purgePeriodically()

	return p, nil
}

// ---------------------------------------------------------------------------

// Submit submits a task to this pool.
//
// Note that you are allowed to call Pool.Submit() from the current Pool.Submit(),
// but what calls for special attention is that you will get blocked with the latest
// Pool.Submit() call once the current Pool runs out of its capacity, and to avoid this,
// you should instantiate a Pool with ants.WithNonblocking(true).
func (p *Pool) Submit(task func()) error {
	if p.IsClosed() {
		return ErrPoolClosed
	}
	var w *goWorker
	if w = p.retrieveWorker(); w == nil {
		return ErrPoolOverload
	}
	w.task <- task
	return nil
}

// Running returns the amount of the currently running goroutines.
func (p *Pool) Running() int {
	return int(atomic.LoadInt32(&p.running))
}

// Free returns the amount of available goroutines to work, -1 indicates this pool is unlimited.
func (p *Pool) Free() int {
	c := p.Cap()
	if c < 0 {
		return -1
	}
	return c - p.Running()
}

// Cap returns the capacity of this pool.
func (p *Pool) Cap() int {
	return int(atomic.LoadInt32(&p.capacity))
}

// Tune changes the capacity of this pool, note that it is noneffective to the infinite or pre-allocation pool.
func (p *Pool) Tune(size int) {
	capacity := p.Cap()
	if capacity == -1 || size <= 0 || size == capacity || p.options.PreAlloc {
		return
	}
	atomic.StoreInt32(&p.capacity, int32(size))
	if size > capacity {
		if size-capacity == 1 {
			p.cond.Signal()
			return
		}
		p.cond.Broadcast()
	}
}

// IsClosed indicates whether the pool is closed.
func (p *Pool) IsClosed() bool {
	return atomic.LoadInt32(&p.state) == CLOSED
}

// Release closes this pool and releases the worker queue.
func (p *Pool) Release() {
	if !atomic.CompareAndSwapInt32(&p.state, OPENED, CLOSED) {
		return
	}
	p.lock.Lock()
	p.workers.reset()
	p.lock.Unlock()
	// There might be some callers waiting in retrieveWorker(), so we need to wake them up to prevent
	// those callers blocking infinitely.
	p.cond.Broadcast()
}

// ReleaseTimeout is like Release but with a timeout, it waits all workers to exit before timing out.
func (p *Pool) ReleaseTimeout(timeout time.Duration) error {
	if p.IsClosed() {
		return errors.New("pool is already closed")
	}
	select {
	case p.stopHeartbeat <- struct{}{}:
		<-p.stopHeartbeat
	default:
	}
	p.Release()
	endTime := time.Now().Add(timeout)
	for time.Now().Before(endTime) {
		if p.Running() == 0 {
			return nil
		}
		time.Sleep(10 * time.Millisecond)
	}
	return ErrTimeout
}

// Reboot reboots a closed pool.
func (p *Pool) Reboot() {
	if atomic.CompareAndSwapInt32(&p.state, CLOSED, OPENED) {
		go p.purgePeriodically()
	}
}

// ---------------------------------------------------------------------------

// incRunning increases the number of the currently running goroutines.
func (p *Pool) incRunning() {
	atomic.AddInt32(&p.running, 1)
}

// decRunning decreases the number of the currently running goroutines.
func (p *Pool) decRunning() {
	atomic.AddInt32(&p.running, -1)
}

// retrieveWorker returns an available worker to run the tasks.
func (p *Pool) retrieveWorker() (w *goWorker) {
	spawnWorker := func() {
		w = p.workerCache.Get().(*goWorker)
		w.run()
	}

	p.lock.Lock()

	w = p.workers.detach()
	if w != nil { // first try to fetch the worker from the queue
		p.lock.Unlock()
	} else if capacity := p.Cap(); capacity == -1 || capacity > p.Running() {
		// if the worker queue is empty and we don't run out of the pool capacity,
		// then just spawn a new worker goroutine.
		p.lock.Unlock()
		spawnWorker()
	} else { // otherwise, we'll have to keep them blocked and wait for at least one worker to be put back into pool.
		if p.options.Nonblocking {
			p.lock.Unlock()
			return
		}
	retry:
		if p.options.MaxBlockingTasks != 0 && p.blockingNum >= p.options.MaxBlockingTasks {
			p.lock.Unlock()
			return
		}
		p.blockingNum++
		p.cond.Wait() // block and wait for an available worker
		p.blockingNum--
		var nw int
		if nw = p.Running(); nw == 0 { // awakened by the scavenger
			p.lock.Unlock()
			if !p.IsClosed() {
				spawnWorker()
			}
			return
		}
		if w = p.workers.detach(); w == nil {
			if nw < capacity {
				p.lock.Unlock()
				spawnWorker()
				return
			}
			goto retry
		}

		p.lock.Unlock()
	}
	return
}

// revertWorker puts a worker back into free pool, recycling the goroutines.
func (p *Pool) revertWorker(worker *goWorker) bool {
	if capacity := p.Cap(); (capacity > 0 && p.Running() > capacity) || p.IsClosed() {
		p.cond.Broadcast()
		return false
	}
	worker.recycleTime = time.Now()
	p.lock.Lock()

	// To avoid memory leaks, add a double check in the lock scope.
	// Issue: https://github.com/panjf2000/ants/issues/113
	if p.IsClosed() {
		p.lock.Unlock()
		return false
	}

	err := p.workers.insert(worker)
	if err != nil {
		p.lock.Unlock()
		return false
	}

	// Notify the invoker stuck in 'retrieveWorker()' of there is an available worker in the worker queue.
	p.cond.Signal()
	p.lock.Unlock()
	return true
}