如何使用Go来实现一个简单的基于磁盘的FIFO队列呢？我们来看看 go-diskqueue 的实现， 它是NSQ中用来持久化的一个库，我们借助它来了解一下，如何实现一个基于磁盘的队列。

这个库很简单，只有一个文件，我们来看看 diskqueue.go：

// diskQueue implements a filesystem backed FIFO queue
type diskQueue struct {
	// 64bit atomic vars need to be first for proper alignment on 32bit platforms

	// run-time state (also persisted to disk)
	readPos      int64
	writePos     int64
	readFileNum  int64
	writeFileNum int64
	depth        int64

	sync.RWMutex

	// instantiation time metadata
	name            string
	dataPath        string
	maxBytesPerFile int64 // currently this cannot change once created
	minMsgSize      int32
	maxMsgSize      int32
	syncEvery       int64         // number of writes per fsync
	syncTimeout     time.Duration // duration of time per fsync
	exitFlag        int32
	needSync        bool

	// keeps track of the position where we have read
	// (but not yet sent over readChan)
	nextReadPos     int64
	nextReadFileNum int64

	readFile  *os.File
	writeFile *os.File
	reader    *bufio.Reader
	writeBuf  bytes.Buffer

	// exposed via ReadChan()
	readChan chan []byte

	// internal channels
	depthChan         chan int64
	writeChan         chan []byte
	writeResponseChan chan error
	emptyChan         chan int
	emptyResponseChan chan error
	exitChan          chan int
	exitSyncChan      chan int

	logf AppLogFunc
}

可以看到，这就是主要的结构体，上面记录了当前读取文件的编号、写入文件的编号、读取的位置（偏移量），写入的位置等。

那么，NSQ的消息是乱序来的，我们无法预知什么时候会有消息来到，它是怎么做到处理成FIFO的呢？很简单，其实就是通过Go 的channel来实现的，可以看到 // internal channels 下面这一堆的channel。

接下来我们来看看读取和写入是如何做到的：

// Put writes a []byte to the queue
func (d *diskQueue) Put(data []byte) error {
	d.RLock()
	defer d.RUnlock()

	if d.exitFlag == 1 {
		return errors.New("exiting")
	}

	d.writeChan <- data
	return <-d.writeResponseChan
}

来看看 writeChan 在哪里被消费：

// ioLoop provides the backend for exposing a go channel (via ReadChan())
// in support of multiple concurrent queue consumers
//
// it works by looping and branching based on whether or not the queue has data
// to read and blocking until data is either read or written over the appropriate
// go channels
//
// conveniently this also means that we're asynchronously reading from the filesystem
func (d *diskQueue) ioLoop() {
	var dataRead []byte
	var err error
	var count int64
	var r chan []byte

	syncTicker := time.NewTicker(d.syncTimeout)

	for {
		// dont sync all the time :)
		if count == d.syncEvery {
			d.needSync = true
		}

		if d.needSync {
			err = d.sync()
			if err != nil {
				d.logf(ERROR, "DISKQUEUE(%s) failed to sync - %s", d.name, err)
			}
			count = 0
		}

		if (d.readFileNum < d.writeFileNum) || (d.readPos < d.writePos) {
			if d.nextReadPos == d.readPos {
				dataRead, err = d.readOne()
				if err != nil {
					d.logf(ERROR, "DISKQUEUE(%s) reading at %d of %s - %s",
						d.name, d.readPos, d.fileName(d.readFileNum), err)
					d.handleReadError()
					continue
				}
			}
			r = d.readChan
		} else {
			r = nil
		}

		select {
		// the Go channel spec dictates that nil channel operations (read or write)
		// in a select are skipped, we set r to d.readChan only when there is data to read
		case r <- dataRead:
			count++
			// moveForward sets needSync flag if a file is removed
			d.moveForward()
		case d.depthChan <- d.depth:
		case <-d.emptyChan:
			d.emptyResponseChan <- d.deleteAllFiles()
			count = 0
		case dataWrite := <-d.writeChan:
			count++
			d.writeResponseChan <- d.writeOne(dataWrite)
		case <-syncTicker.C:
			if count == 0 {
				// avoid sync when there's no activity
				continue
			}
			d.needSync = true
		case <-d.exitChan:
			goto exit
		}
	}

exit:
	d.logf(INFO, "DISKQUEUE(%s): closing ... ioLoop", d.name)
	syncTicker.Stop()
	d.exitSyncChan <- 1
}

ioLoop 是在 New 函数被调用时发起的。我们来看 case dataWrite := <-d.writeChan 分支，调用 d.writeOne，将其返回 结果放到 d.writeResponseChan 里：

// writeOne performs a low level filesystem write for a single []byte
// while advancing write positions and rolling files, if necessary
func (d *diskQueue) writeOne(data []byte) error {
	var err error

	if d.writeFile == nil {
		curFileName := d.fileName(d.writeFileNum)
		d.writeFile, err = os.OpenFile(curFileName, os.O_RDWR|os.O_CREATE, 0600)
		if err != nil {
			return err
		}

		d.logf(INFO, "DISKQUEUE(%s): writeOne() opened %s", d.name, curFileName)

		if d.writePos > 0 {
			_, err = d.writeFile.Seek(d.writePos, 0)
			if err != nil {
				d.writeFile.Close()
				d.writeFile = nil
				return err
			}
		}
	}

	dataLen := int32(len(data))

	if dataLen < d.minMsgSize || dataLen > d.maxMsgSize {
		return fmt.Errorf("invalid message write size (%d) maxMsgSize=%d", dataLen, d.maxMsgSize)
	}

	d.writeBuf.Reset()
	err = binary.Write(&d.writeBuf, binary.BigEndian, dataLen)
	if err != nil {
		return err
	}

	_, err = d.writeBuf.Write(data)
	if err != nil {
		return err
	}

	// only write to the file once
	_, err = d.writeFile.Write(d.writeBuf.Bytes())
	if err != nil {
		d.writeFile.Close()
		d.writeFile = nil
		return err
	}

	totalBytes := int64(4 + dataLen)
	d.writePos += totalBytes
	d.depth += 1

	if d.writePos >= d.maxBytesPerFile {
		d.writeFileNum++
		d.writePos = 0

		// sync every time we start writing to a new file
		err = d.sync()
		if err != nil {
			d.logf(ERROR, "DISKQUEUE(%s) failed to sync - %s", d.name, err)
		}

		if d.writeFile != nil {
			d.writeFile.Close()
			d.writeFile = nil
		}
	}

	return err
}

可以看到写入的细节，就是使用Go的 binary 库，用大端的方式，把数据写入，最后来一个判断，如果文件超长了，新开一个文件。 其中有一个细节，调用了 d.sync()，我们来看看：

// sync fsyncs the current writeFile and persists metadata
func (d *diskQueue) sync() error {
	if d.writeFile != nil {
		err := d.writeFile.Sync()
		if err != nil {
			d.writeFile.Close()
			d.writeFile = nil
			return err
		}
	}

	err := d.persistMetaData()
	if err != nil {
		return err
	}

	d.needSync = false
	return nil
}

// persistMetaData atomically writes state to the filesystem
func (d *diskQueue) persistMetaData() error {
	var f *os.File
	var err error

	fileName := d.metaDataFileName()
	tmpFileName := fmt.Sprintf("%s.%d.tmp", fileName, rand.Int())

	// write to tmp file
	f, err = os.OpenFile(tmpFileName, os.O_RDWR|os.O_CREATE, 0600)
	if err != nil {
		return err
	}

	_, err = fmt.Fprintf(f, "%d\n%d,%d\n%d,%d\n",
		d.depth,
		d.readFileNum, d.readPos,
		d.writeFileNum, d.writePos)
	if err != nil {
		f.Close()
		return err
	}
	f.Sync()
	f.Close()

	// atomically rename
	return os.Rename(tmpFileName, fileName)
}

可以看到调用了 persistMetaData，什么是metadata呢？就是我们最开始说的，读取的位置、写入的位置、读取的文件号， 写入的文件号等等。可以看到它的实现，就是先打开一个临时文件，把数据写入，并且sync之后，替换原来的metadata文件。

这样就达到了持久化metadata的作用。这就是写入文件和metadata的逻辑。至于读取，则是在ioLoop的for循环里：

		if (d.readFileNum < d.writeFileNum) || (d.readPos < d.writePos) {
			if d.nextReadPos == d.readPos {
				dataRead, err = d.readOne()

然后

		select {
		// the Go channel spec dictates that nil channel operations (read or write)
		// in a select are skipped, we set r to d.readChan only when there is data to read
		case r <- dataRead:
			count++
			// moveForward sets needSync flag if a file is removed
			d.moveForward()

放到 d.readChan 里，消费端则从这个channel里读取。

这就是 go-diskqueue 的实现。

Ref:

• https://github.com/nsqio/go-diskqueue