How can Gulp be restarted upon each Gulpfile change?

Question

I am developing a Gulpfile. Can it be made to restart as soon as it changes? I am developing it in CoffeeScript. Can Gulp watch Gulpfile.coffee in order to restart when changes are saved?

Cyanite

Answer 1

You can create a task that will gulp.watch for gulpfile.js and simply spawn another gulp child_process.

var gulp = require('gulp'),
    argv = require('yargs').argv, // for args parsing
    spawn = require('child_process').spawn;

gulp.task('log', function() {
  console.log('CSSs has been changed');
});

gulp.task('watching-task', function() {
  gulp.watch('*.css', ['log']);
});

gulp.task('auto-reload', function() {
  var p;

  gulp.watch('gulpfile.js', spawnChildren);
  spawnChildren();

  function spawnChildren(e) {
    // kill previous spawned process
    if(p) { p.kill(); }

    // `spawn` a child `gulp` process linked to the parent `stdio`
    p = spawn('gulp', [argv.task], {stdio: 'inherit'});
  }
});

I used yargs in order to accept the 'main task' to run once we need to restart. So in order to run this, you would call:

gulp auto-reload --task watching-task

And to test, call either touch gulpfile.js or touch a.css to see the logs.

Anderaanderea

That's a nice trick. One thing I'd be tempted to do would be to name the task something like auto-reload, and run it like gulp auto-reload --task foo, otherwise you could get a fork-bomb starting the default task recursively. – Unpaid

Good suggestion @OverZealous. Applied the changes in order to avoid this problem for future readers. – Anderaanderea

That seems complicated, I wish it were as easy as it is with Grunt – Croesus

This method also has a bunch of side effects, like occasional errors (2014-07-29 11:48 gulp[74366] (CarbonCore.framework) FSEventStreamStart: register_with_server: ERROR: f2d_register_rpc() => (null) (-21)), and, e.g. localhost:3000 port number incrementing when using BrowserSync – Deaden

Om Windows instead of require('child_process') better to use require('child-proc') – Mcadoo

@MurraySmith Yes, there are some issues with the approach... I agree. I'm still looking for a good way to do so, but I think it will require a change on Gulp source code. I started working on this change, but got no enough time so far. – Anderaanderea

It's worth noting that this will only work on Windows using the cross_spawn module. npmjs.com/package/cross-spawn – Anchovy

Awesome answer. Worked perfectly fine (Y) – Bookmobile

New behavior in gulp v4: The watcher function (spawnChildren()) is passed a callback argument which must be called after spawnChildren() is done, or else the auto-reload task never goes back to watching for further changes (i.e. the reload only works once). github.com/gulpjs/gulp/blob/v4.0.0/docs/API.md#fn-1 – Vaccine

Answer 2

I created gulper that is gulp.js cli wrapper to restart gulp on gulpfile change.

You can simply replace gulp with gulper.

$ gulper <task-name>

Baliol

Very simple and elegant solution. – Aliber

Wish I'd seen this earlier. Now I'm stuck with my own implementation: github.com/samvv/node-galop – Gunnery

Shame on me, I was using it for a year since I've first saw this answer, and I only now figured out that I did not upvote it. Thanks for the useful tool! – Vitrify

nice for Linux & Mac, but fails on Windows - it tries to launch gulp's CMD as JS – Heirship

You're a star mate :) – Esta

Answer 3

I use a small shell script for this purpose. This works on Windows as well.

Press Ctrl+C to stop the script.

// gulpfile.js
gulp.task('watch', function() {
    gulp.watch('gulpfile.js', process.exit);
});

Bash shell script:

# watch.sh
while true; do
    gulp watch;
done;

Windows version: watch.bat

@echo off
:label
cmd /c gulp watch
goto label

Reconstructionism

Simple, easy to understand (for me at least), certainly no leaks. – Cub

This is the only one that worked for me, since gulper gives errors on Windows. – Aurelio

Answer 4

I was getting a bunch of EADDRINUSE errors with the solution in Caio Cunha's answer. My gulpfile opens a local webserver with connect and LiveReload. It appears the new gulp process briefly coexists with the old one before the older process is killed, so the ports are still in use by the soon-to-die process.

Here's a similar solution which gets around the coexistence problem, (based largely on this):

var gulp = require('gulp');
var spawn = require('child_process').spawn;

gulp.task('gulp-reload', function() {
  spawn('gulp', ['watch'], {stdio: 'inherit'});
  process.exit();
});

gulp.task('watch', function() {
  gulp.watch('gulpfile.js', ['gulp-reload']);
});

That works fairly well, but has one rather serious side-effect: The last gulp process is disconnected from the terminal. So when gulp watch exits, an orphaned gulp process is still running. I haven't been able to work around that problem, the extra gulp process can be killed manually, or just save a syntax error to gulpfile.js.

Hialeah

to restart any task : process.argv.shift(); spawn(process.argv.shift(), process.argv, {stdio: 'inherit'}); process.exit() – Exceeding

Answer 5

I've been dealing with the same problem and the solution in my case was actually very simple. Two things.

1. npm install nodemon -g (or locally if you prefer)

2. run with cmd or create a script in packages like this:

   "dev": "nodemon --watch gulpfile.js --exec gulp"
   

3. The just type npm run dev

--watch specifies the file to keep an eye on. --exec says execute next in line and gulp is your default task. Just pass in argument if you want non default task.

Hope it helps.

EDIT : Making it fancy ;) Now while the first part should achieve what you were after, in my setup I've needed to add a bit more to make it really user friend. What I wanted was

1. First open the page.
2. Look for changes in gulpfile.js and restart gulp if there are any
3. Gulp it up so keep an eye on files, rebuild and hot reload

If you only do what I've said in the first part, it will open the page every time. To fix it, create a gulp task that will open the page. Like this :

gulp.task('open', function(){
return gulp
.src(config.buildDest + '/index.html')
.pipe(plugins.open({
    uri: config.url
}));

Then in my main tasks I have :

gulp.task('default', ['dev-open']);
gulp.task('dev-open', function(done){
    plugins.sequence('build', 'connect', 'open', 'watch', done);
});
gulp.task('dev', function(done){
    plugins.sequence('build', 'connect', 'watch', done);
});

Then modifying your npm scripts to

"dev": "gulp open & nodemon --watch gulpfile.js --watch webpack.config.js --exec gulp dev"

Will give you exactly what you want. First open the page and then just keep live reloading. Btw for livereload I use the one that comes with connect which always uses the same port. Hope it works for you, enjoy!

Gignac

Nodemon is exactly what I needed. I had completely forgotten about it's existence. What a shame. – Sita

I recommend the original solution, not the 'fancy' addition, with two recommendations: 1) add nodemon to optionalDependencies in package.json and 2) Mention in the README – Hold

Answer 6

Another solution for this is to refresh the require.cache.

var gulp = require('gulp');

var __filenameTasks = ['lint', 'css', 'jade'];
var watcher = gulp.watch(__filename).once('change', function(){
  watcher.end(); // we haven't re-required the file yet
                 // so is the old watcher
  delete require.cache[__filename];
  require(__filename);
  process.nextTick(function(){
    gulp.start(__filenameTasks);
  });
});

Precentor

Edit: better using once otherwise you might have memory leaks – Precentor

This still leaks: (node) warning: possible EventEmitter memory leak detected. 11 change listeners added. Use emitter.setMaxListeners() to increase limi – Oleaginous

I've modified it a little bit. The problem is that even if you use once the watcher is still there so you also have to end it – Precentor

Eventually my fix was to stop using gulp heh – Oleaginous

Answer 7

I know this is a very old question, but it's a top comment on Google, so still very relevant.

Here is an easier way, if your source gulpfile.js is in a different directory than the one in use. (That's important!) It uses the gulp modules gulp-newer and gulp-data.

var gulp          = require('gulp'         )
  , data          = require('gulp-data'    )
  , newer         = require('gulp-newer'   )
  , child_process = require('child_process')
;

gulp.task( 'gulpfile.js' , function() {
    return gulp.src( 'sources/gulpfile.js' ) // source
        .pipe( newer(     '.'            ) ) // check
        .pipe( gulp.dest( '.'            ) ) // write
        .pipe( data( function(file)        { // reboot
            console.log('gulpfile.js changed! Restarting gulp...') ;
            var t , args = process.argv ;
            while ( args.shift().substr(-4) !== 'gulp' ) { t=args; }
            child_process.spawn( 'gulp' , args , { stdio: 'inherit' } ) ;
            return process.exit() ;
        } ) )
    ;
} ) ;

It works like this:

• Trick 1: gulp-newer only executes the following pipes, if the source file is newer than the current one. This way we make sure, there's no reboot-loop.
• The while loop removes everything before and including the gulp command from the command string, so we can pass through any arguments.
• child_process.spawn spawns a new gulp process, piping input output and error to the parent.
• Trick 2: process.exit kills the current process. However, the process will wait to die until the child process is finished.

There are many other ways of inserting the restart function into the pipes. I just happen to use gulp-data in every of my gulpfiles anyway. Feel free to comment your own solution. :)

Underproduction

Answer 8

Here's another version of @CaioToOn's reload code that is more in line with normal Gulp task procedure. It also does not depend on yargs.

Require spawn and initilaize the process variable (yargs is not needed):

var spawn = require('child_process').spawn;
var p;

The default gulp task will be the spawner:

gulp.task('default', function() {
  if(p) { p.kill(); }
  // Note: The 'watch' is the new name of your normally-default gulp task. Substitute if needed.
  p = spawn('gulp', ['watch'], {stdio: 'inherit'});
});

Your watch task was probably your default gulp task. Rename it to watch and add a gulp.watch()for watching your gulpfile and run the default task on changes:

gulp.task('watch', ['sass'], function () {
  gulp.watch("scss/*.scss", ['sass']);
  gulp.watch('gulpfile.js', ['default']);
});

Now, just run gulp and it will automatically reload if you change your gulpfile!

Heartbreak

This answer does not address open connections that can lead to errors. See here for the problem % a solution: https://mcmap.net/q/293896/-how-can-gulp-be-restarted-upon-each-gulpfile-change – Exceeding

Answer 9

try this code (only win32 platform)

gulp.task('default', ['less', 'scripts', 'watch'], function(){
    gulp.watch('./gulpfile.js').once('change' ,function(){
        var p;
        var childProcess = require('child_process');
        if(process.platform === 'win32'){
            if(p){
                childProcess.exec('taskkill /PID' + p.id + ' /T /F', function(){});
                p.kill();
            }else{
                p = childProcess.spawn(process.argv[0],[process.argv[1]],{stdio: 'inherit'});
            }
        }
    });
});

Mouthy

this is the only solution that worked for me on Windows. But still it does not kill the previous process, at the end I have to manually kill all the spawned processes. – Nonpros

also, if you use connect.reload(), you need to run connect.serverClose() before process handling. Otherwise port is not released and throws more errors. – Nonpros

Answer 10

Install nodemon globally: npm i -g nodemon

And add in your .bashrc (or .bash_profile or .profile) an alias:

alias gulp='nodemon --watch gulpfile.js --watch gulpfile.babel.js --quiet --exitcrash --exec gulp'

This will watch for file gulpfile.js and gulpfile.babel.js changes. (see Google)

P.S. This can be helpful for endless tasks (like watch) but not for single run tasks. I mean it uses watch so it will continue process even after gulp task is done. ;)

Quach

Better: Add nodemon to "optionalDependencies" within package.json, include "dev":"nodemon --watch gulpfile.js --exec gulp dev" in package.json's "scripts", and finally mention in the project README. – Hold

Answer 11

A good solution for Windows, which also works well with Visual Studio task runner.

/// <binding ProjectOpened='auto-watchdog' />
const spawn = require('child-proc').spawn,
      configPaths = ['Gulpconfig.js', 'bundleconfig.js'];

gulp.task('watchdog', function () {
    // TODO: add other watches here

    gulp.watch(configPaths, function () {
        process.exit(0);
    });
});

gulp.task('auto-watchdog', function () {
    let p = null;

    gulp.watch(configPaths, spawnChildren);
    spawnChildren();

    function spawnChildren() {
        const args = ['watchdog', '--color'];

        // kill previous spawned process
        if (p) {
            // You might want to trigger a build as well
            args.unshift('build');

            setTimeout(function () {
                p.kill();
            }, 1000);
        }

        // `spawn` a child `gulp` process linked to the parent `stdio`
        p = spawn('gulp', args, { stdio: 'inherit' });
    }
});

Main changes compared to other answers:

• Uses child-proc because child_process fails on Windows.
• The watchdog exits itself on changes of files because in Windows the gulp call is wrapped in a batch script. Killing the batch script wouldn't kill gulp itself causing multiple watches to be spawned over time.
• Build on change: Usually a gulpfile change also warrants rebuilding the project.

Calysta

Answer 12

Here's a short version that's easy to understand that you can set as a default task so you just need to type "gulp":

gulp.task('watch', function() {
  const restartingGulpProcessCmd = 'while true; do gulp watch2 --colors; done;';
  const restartingGulpProcess = require('child_process').exec(restartingGulpProcessCmd);
  restartingGulpProcess.stdout.pipe(process.stdout);
  restartingGulpProcess.stderr.pipe(process.stderr);
});

gulp.task('watch2', function() {
  gulp.watch(['config/**.js', 'webpack.config.js', './gulpfile.js'],
    () => {
      console.log('Config file changed. Quitting so gulp can be restarted.');
      process.exit();
    });

  // Add your other watch and build commands here
}

gulp.task('default', ['watch']);

Misread

What would this be for windows? The while true; part returns 'while is not recognized as an internal or external command..' – Celtuce

Answer 13

Install gulp-restart

npm install gulp-restart

This code will work for you.

var gulp = require('gulp'); var restart = require('gulp-restart');

gulp.task('watch', function() { gulp.watch(['gulpfile.js'], restart); })

it will restart gulp where you do changes on the gulpfile.js

Ebonyeboracum

gulp 4.0 on MacOs: TypeError: stdin.setRawMode is not a function at Object.<anonymous> (/.../node_modules/gulp-restart/index.js:18:7) – Garbage

Answer 14

I spent a whole day trying to make this work on Windows / Gulp 4.0.2, and I (finally) made it...

I used some solutions from people on this page and from one other page. It's all there in the comments...

Any change in any function inside "allTasks" will take effect on gulpfile.js (or other watched files) save...

There are some useless comments and console.logs left, feel free to remove them... ;)

const { gulp, watch, src, dest, series, parallel } = require("gulp");
const spawn = require('child_process').spawn;

// This function contains all that is necessary: start server, watch files...
const allTasks = function (callback) {
   console.log('==========');
   console.log('========== STARTING THE GULP DEFAULT TASK...');
   console.log('========== USE CTRL+C TO STOP THE TASK');
   console.log('==========');

   startServer();
   // other functions (watchers) here

   // *** Thanks to Sebazzz ***
   // Stop all on gulpfile.js change
   watch('gulpfile.js', function (callback) {
      callback(); // avoid "task didn't complete" error
      process.exit();
   });

   callback(); // avoid "task didn't complete" error
}


// Restart allTasks
// ********************************************
// CALL GULPDEFAULT WITH THE GULP DEFAULT TASK:
// export.default = gulpDefault
// ********************************************
const gulpDefault = function (callback) {
   let p = null;

   watch('gulpfile.js', spawnChildren);

   // *** Thanks to Sphinxxx: ***
   // New behavior in gulp v4: The watcher function (spawnChildren()) is passed a callback argument
   // which must be called after spawnChildren() is done, or else the auto-reload task
   // never goes back to watching for further changes (i.e.the reload only works once).
   spawnChildren(callback);

   function spawnChildren(callback) {

      /*
      // This didn't do anything for me, with or without the delay,
      // so I left it there, but commented it out, together with the console.logs...

      // kill previous spawned process
      if (p) {
         // You might want to trigger a build as well
         //args.unshift('build');

         setTimeout(function () {
            console.log('========== p.pid before kill: ' + p.pid); // a random number
            console.log('========== p before kill: ' + p); // [object Object]
            p.kill();
            console.log('========== p.pid after kill: ' + p.pid); // the same random number
            console.log('========== p after kill: ' + p); // still [object Object]
         }, 1000);
      }
      */

      // `spawn` a child `gulp` process linked to the parent `stdio`
      // ['watch'] is the task that calls the main function (allTasks):
      // exports.watch = allTasks;
      p = spawn('gulp', ['watch'], { stdio: 'inherit', shell: true });
      // *** Thanks to people from: ***
      // https://mcmap.net/q/73597/-how-do-i-debug-quot-error-spawn-enoent-quot-on-node-js
      // Prevent Error: spawn ENOENT
      // by passing "shell: true" to the spawn options

      callback(); // callback called - thanks to Sphinxxx
   }
}

exports.default = gulpDefault;
exports.watch = allTasks;

Eloquent