ProcessBuilder：在不阻塞主线程的情况下转发已启动进程的 stdout 和 stderr

Question

I'm building a process in Java using ProcessBuilder as follows:

ProcessBuilder pb = new ProcessBuilder()
        .command("somecommand", "arg1", "arg2")
        .redirectErrorStream(true);
Process p = pb.start();

InputStream stdOut = p.getInputStream();

Now my problem is the following: I would like to capture whatever is going through stdout and/or stderr of that process and redirect it to System.out asynchronously. I want the process and its output redirection to run in the background. So far, the only way I've found to do this is to manually spawn a new thread that will continuously read from stdOut and then call the appropriate write() method of System.out .

new Thread(new Runnable(){
    public void run(){
        byte[] buffer = new byte[8192];
        int len = -1;
        while((len = stdOut.read(buffer)) > 0){
            System.out.write(buffer, 0, len);
        }
    }
}).start();

While that approach kind of works, it feels a bit dirty. And on top of that, it gives me one more thread to manage and terminate correctly. Is there any better way to do this?

Answer 1

Use ProcessBuilder.inheritIO , it sets the source and destination for subprocess standard I/O to be the same as those of the current Java process.

Process p = new ProcessBuilder().inheritIO().command("command1").start();

If Java 7 is not an option

public static void main(String[] args) throws Exception {
    Process p = Runtime.getRuntime().exec("cmd /c dir");
    inheritIO(p.getInputStream(), System.out);
    inheritIO(p.getErrorStream(), System.err);

}

private static void inheritIO(final InputStream src, final PrintStream dest) {
    new Thread(new Runnable() {
        public void run() {
            Scanner sc = new Scanner(src);
            while (sc.hasNextLine()) {
                dest.println(sc.nextLine());
            }
        }
    }).start();
}

Threads will die automatically when subprocess finishes, because src will EOF.

Answer 2

For Java 7 and later , see Evgeniy Dorofeev's answer .

For Java 6 and earlier , create and use a StreamGobbler :

StreamGobbler errorGobbler = 
  new StreamGobbler(p.getErrorStream(), "ERROR");

// any output?
StreamGobbler outputGobbler = 
  new StreamGobbler(p.getInputStream(), "OUTPUT");

// start gobblers
outputGobbler.start();
errorGobbler.start();

...

private class StreamGobbler extends Thread {
    InputStream is;
    String type;

    private StreamGobbler(InputStream is, String type) {
        this.is = is;
        this.type = type;
    }

    @Override
    public void run() {
        try {
            InputStreamReader isr = new InputStreamReader(is);
            BufferedReader br = new BufferedReader(isr);
            String line = null;
            while ((line = br.readLine()) != null)
                System.out.println(type + "> " + line);
        }
        catch (IOException ioe) {
            ioe.printStackTrace();
        }
    }
}

Answer 3

A flexible solution with Java 8 lambda that lets you provide a Consumer that will process the output (eg. log it) line by line. run() is a one-liner with no checked exceptions thrown. Alternatively to implementing Runnable , it can extend Thread instead as other answers suggest.

class StreamGobbler implements Runnable {
    private InputStream inputStream;
    private Consumer<String> consumeInputLine;

    public StreamGobbler(InputStream inputStream, Consumer<String> consumeInputLine) {
        this.inputStream = inputStream;
        this.consumeInputLine = consumeInputLine;
    }

    public void run() {
        new BufferedReader(new InputStreamReader(inputStream)).lines().forEach(consumeInputLine);
    }
}

You can then use it for example like this:

public void runProcessWithGobblers() throws IOException, InterruptedException {
    Process p = new ProcessBuilder("...").start();
    Logger logger = LoggerFactory.getLogger(getClass());

    StreamGobbler outputGobbler = new StreamGobbler(p.getInputStream(), System.out::println);
    StreamGobbler errorGobbler = new StreamGobbler(p.getErrorStream(), logger::error);

    new Thread(outputGobbler).start();
    new Thread(errorGobbler).start();
    p.waitFor();
}

Here the output stream is redirected to System.out and the error stream is logged on the error level by the logger .

Answer 4

It's as simple as following:

    File logFile = new File(...);
    ProcessBuilder pb = new ProcessBuilder()
        .command("somecommand", "arg1", "arg2")
    processBuilder.redirectErrorStream(true);
    processBuilder.redirectOutput(logFile);

by .redirectErrorStream(true) you tell process to merge error and output stream and then by .redirectOutput(file) you redirect merged output to a file.

Update:

I did manage to do this as follows:

public static void main(String[] args) {
    // Async part
    Runnable r = () -> {
        ProcessBuilder pb = new ProcessBuilder().command("...");
        // Merge System.err and System.out
        pb.redirectErrorStream(true);
        // Inherit System.out as redirect output stream
        pb.redirectOutput(ProcessBuilder.Redirect.INHERIT);
        try {
            pb.start();
        } catch (IOException e) {
            e.printStackTrace();
        }
    };
    new Thread(r, "asyncOut").start();
    // here goes your main part
}

Now you're able to see both outputs from main and asyncOut threads in System.out

Answer 5

Simple java8 solution with capturing both outputs and reactive processing using CompletableFuture :

static CompletableFuture<String> readOutStream(InputStream is) {
    return CompletableFuture.supplyAsync(() -> {
        try (
                InputStreamReader isr = new InputStreamReader(is);
                BufferedReader br = new BufferedReader(isr);
        ) {
              StringBuilder res = new StringBuilder();
              String inputLine;
              while ((inputLine = br.readLine()) != null) {
                  res.append(inputLine).append(System.lineSeparator());
              }
              return res.toString();
        } catch (Throwable e) {
            throw new RuntimeException("problem with executing program", e);
        }
    });
}

And the usage:

Process p = Runtime.getRuntime().exec(cmd);
CompletableFuture<String> soutFut = readOutStream(p.getInputStream());
CompletableFuture<String> serrFut = readOutStream(p.getErrorStream());
CompletableFuture<String> resultFut = 
    soutFut.thenCombine(serrFut, (stdout, stderr) -> {
         // print to current stderr the stderr of process and return the stdout
         System.err.println(stderr);
        
         return stdout;
    });
// get stdout once ready, blocking
String result = resultFut.get();

Answer 6

There is a library that provides a better ProcessBuilder, zt-exec. This library can do exactly what you are asking for and more.

Here's what your code would look like with zt-exec instead of ProcessBuilder :

add the dependency :

<dependency>
  <groupId>org.zeroturnaround</groupId>
  <artifactId>zt-exec</artifactId>
  <version>1.11</version>
</dependency>

The code :

new ProcessExecutor()
  .command("somecommand", "arg1", "arg2")
  .redirectOutput(System.out)
  .redirectError(System.err)
  .execute();

Documentation of the library is here : https://github.com/zeroturnaround/zt-exec/

Answer 7

I too can use only Java 6. I used @EvgeniyDorofeev's thread scanner implementation. In my code, after a process finishes, I have to immediately execute two other processes that each compare the redirected output (a diff-based unit test to ensure stdout and stderr are the same as the blessed ones).

The scanner threads don't finish soon enough, even if I waitFor() the process to complete. To make the code work correctly, I have to make sure the threads are joined after the process finishes.

public static int runRedirect (String[] args, String stdout_redirect_to, String stderr_redirect_to) throws IOException, InterruptedException {
    ProcessBuilder b = new ProcessBuilder().command(args);
    Process p = b.start();
    Thread ot = null;
    PrintStream out = null;
    if (stdout_redirect_to != null) {
        out = new PrintStream(new BufferedOutputStream(new FileOutputStream(stdout_redirect_to)));
        ot = inheritIO(p.getInputStream(), out);
        ot.start();
    }
    Thread et = null;
    PrintStream err = null;
    if (stderr_redirect_to != null) {
        err = new PrintStream(new BufferedOutputStream(new FileOutputStream(stderr_redirect_to)));
        et = inheritIO(p.getErrorStream(), err);
        et.start();
    }
    p.waitFor();    // ensure the process finishes before proceeding
    if (ot != null)
        ot.join();  // ensure the thread finishes before proceeding
    if (et != null)
        et.join();  // ensure the thread finishes before proceeding
    int rc = p.exitValue();
    return rc;
}

private static Thread inheritIO (final InputStream src, final PrintStream dest) {
    return new Thread(new Runnable() {
        public void run() {
            Scanner sc = new Scanner(src);
            while (sc.hasNextLine())
                dest.println(sc.nextLine());
            dest.flush();
        }
    });
}

Answer 8

As an addition to msangel answer I would like to add the following code block:

private static CompletableFuture<Boolean> redirectToLogger(final InputStream inputStream, final Consumer<String> logLineConsumer) {
        return CompletableFuture.supplyAsync(() -> {
            try (
                InputStreamReader inputStreamReader = new InputStreamReader(inputStream);
                BufferedReader bufferedReader = new BufferedReader(inputStreamReader);
            ) {
                String line = null;
                while((line = bufferedReader.readLine()) != null) {
                    logLineConsumer.accept(line);
                }
                return true;
            } catch (IOException e) {
                return false;
            }
        });
    }

It allows to redirect the input stream (stdout, stderr) of the process to some other consumer. This might be System.out::println or anything else consuming strings.

Usage:

...
Process process = processBuilder.start()
CompletableFuture<Boolean> stdOutRes = redirectToLogger(process.getInputStream(), System.out::println);
CompletableFuture<Boolean> stdErrRes = redirectToLogger(process.getErrorStream(), System.out::println);
System.out.println(stdOutRes.get());
System.out.println(stdErrRes.get());
System.out.println(process.waitFor());

Answer 9

Thread thread = new Thread(() -> {
      new BufferedReader(
          new InputStreamReader(inputStream, 
                                StandardCharsets.UTF_8))
              .lines().forEach(...);
    });
    thread.start();

Your custom code goes instead of the ...

Answer 10

It's really surprising to me that the redirection methods in ProcessBuilder don't accept an OutputStream , only File . Yet another proof of forced boilerplate code that Java forces you to write.

That said, let's look at a list of comprehensive options:

1. If you want the process output to simply be redirected to its parent's output stream, inheritIO will do the job.
2. If you want the process output to go to a file, use redirect*(file) .
3. If you want the process output to go to a logger, you need to consume the process InputStream in a separate thread. See the answers that use a Runnable or CompletableFuture . You can also adapt the code below to do this.
4. If you want to the process output to go to a PrintWriter , that may or may not be the stdout (very useful for testing), you can do the following:

static int execute(List<String> args, PrintWriter out) {
    ProcessBuilder builder = new ProcessBuilder()
            .command(args)
            .redirectErrorStream(true);
    Process process = null;
    boolean complete = false;
    try {
        process = builder.start();
        redirectOut(process.getInputStream(), out)
                .orTimeout(TIMEOUT, TimeUnit.SECONDS);
        complete = process.waitFor(TIMEOUT, TimeUnit.SECONDS);
    } catch (IOException e) {
        throw new UncheckedIOException(e);
    } catch (InterruptedException e) {
        LOG.warn("Thread was interrupted", e);
    } finally {
        if (process != null && !complete) {
            LOG.warn("Process {} didn't finish within {} seconds", args.get(0), TIMEOUT);
            process = process.destroyForcibly();
        }
    }

    return process != null ? process.exitValue() : 1;
}

private static CompletableFuture<Void> redirectOut(InputStream in, PrintWriter out) {
    return CompletableFuture.runAsync(() -> {
        try (
                InputStreamReader inputStreamReader = new InputStreamReader(in);
                BufferedReader bufferedReader = new BufferedReader(inputStreamReader)
        ) {
            bufferedReader.lines()
                    .forEach(out::println);
        } catch (IOException e) {
            LOG.error("Failed to redirect process output", e);
        }
    });
}

Advantages of the code above over the other answers thus far:

1. redirectErrorStream(true) redirects the error stream to the output stream, so that we only have to bother with one.
2. CompletableFuture.runAsync runs from the ForkJoinPool . Note that this code doesn't block by calling get or join on the CompletableFuture but sets a timeout instead on its completion (Java 9+). There's no need for CompletableFuture.supplyAsync because there's nothing really to return from the method redirectOut .
3. BufferedReader.lines is simpler than using a while loop.

Answer 11

import java.io.BufferedReader;
import java.io.InputStreamReader;

public class Main {

    public static void main(String[] args) throws Exception {
        ProcessBuilder pb = new ProcessBuilder("script.bat");
        pb.redirectErrorStream(true);
        Process p = pb.start();
        BufferedReader logReader = new BufferedReader(new InputStreamReader(p.getInputStream()));
        String logLine = null;
        while ( (logLine = logReader.readLine()) != null) {
           System.out.println("Script output: " + logLine);
        }
    }
}

By using this line: pb.redirectErrorStream(true); we can combine InputStream and ErrorStream

Answer 12

By default, the created subprocess does not have its own terminal or console. All its standard I/O (ie stdin, stdout, stderr) operations will be redirected to the parent process, where they can be accessed via the streams obtained using the methods getOutputStream(), getInputStream(), and getErrorStream(). The parent process uses these streams to feed input to and get output from the subprocess. Because some native platforms only provide limited buffer size for standard input and output streams, failure to promptly write the input stream or read the output stream of the subprocess may cause the subprocess to block, or even deadlock.

https://www.securecoding.cert.org/confluence/display/java/FIO07-J.+Do+not+let+external+processes+block+on+IO+buffers