Search code examples
filegogenerator

"tail -f"-like generator


I had this convenient function in Python:

def follow(path):
    with open(self.path) as lines:
        lines.seek(0, 2)  # seek to EOF

        while True:
            line = lines.readline()
            if not line:
                time.sleep(0.1)
                    continue
                yield line 

It does something similar to UNIX tail -f: you get last lines of a file as they come. It's convenient because you can get the generator without blocking and pass it to another function.

Then I had to do the same thing in Go. I'm new to this language, so I'm not sure whether what I did is idiomatic/correct enough for Go.

Here is the code:

func Follow(fileName string) chan string {

    out_chan := make(chan string)

    file, err := os.Open(fileName)
    if err != nil {
        log.Fatal(err)
    }

    file.Seek(0, os.SEEK_END)
    bf := bufio.NewReader(file)

    go func() {
        for {
            line, _, _ := bf.ReadLine()

            if len(line) == 0 {
                time.Sleep(10 * time.Millisecond)
            } else {
                out_chan <- string(line)
            }
        }

        defer file.Close()
        close(out_chan)
    }()

    return out_chan
}

Is there any cleaner way to do this in Go? I have a feeling that using an asynchronous call for such a thing is an overkill, and it really bothers me.


Solution

  • Create a wrapper around a reader that sleeps on EOF:

    type tailReader struct {
        io.ReadCloser
    }
    
    func (t tailReader) Read(b []byte) (int, error) {
        for {
            n, err := t.ReadCloser.Read(b)
            if n > 0 {
                return n, nil
            } else if err != io.EOF {
                return n, err
            }
            time.Sleep(10 * time.Millisecond)
        }
    }
    
    func newTailReader(fileName string) (tailReader, error) {
        f, err := os.Open(fileName)
        if err != nil {
            return tailReader{}, err
        }
    
        if _, err := f.Seek(0, 2); err != nil {
            return tailReader{}, err
        }
        return tailReader{f}, nil
    }
    

    This reader can be used anywhere an io.Reader can be used. Here's how loop over lines using bufio.Scanner:

    t, err := newTailReader("somefile")
    if err != nil {
        log.Fatal(err)
    }
    defer t.Close()
    scanner := bufio.NewScanner(t)
    for scanner.Scan() {
        fmt.Println(scanner.Text())
    }
    if err := scanner.Err(); err != nil {
        fmt.Fprintln(os.Stderr, "reading:", err)
    }
    

    The reader can also be used to loop over JSON values appended to the file:

    t, err := newTailReader("somefile")
    if err != nil {
        log.Fatal(err)
    }
    defer t.Close()
    dec := json.NewDecoder(t)
    for {
        var v SomeType
        if err := dec.Decode(&v); err != nil {
           log.Fatal(err)
        }
        fmt.Println("the value is ", v)
    }
    

    There are a couple of advantages this approach has over the goroutine approach outlined in the question. The first is that shutdown is easy. Just close the file. There's no need to signal the goroutine that it should exit. The second advantage is that many packages work with io.Reader.

    The sleep time can be adjusted up or down to meet specific needs. Decrease the time for lower latency and increase the time to reduce CPU use. A sleep of 100ms is probably fast enough for data that's displayed to humans.