dbalabka / dag

Directed acyclic graph of tasks implementation based on Amphp promises

This experimental library helps to organize tasks and its dependencies. Amphp framework support provide ability to execute tasks asynchronously.

Following example show basic usage:

$taskDag = (new TaskGraph())
    
    // root task in graph (there might be multiple root tasks)
    ->addTask(new ClosureTask('1', function () {
        echo $this->name;
    }))
    
    // these two tasks were be executed only after task with name '1'
    // the order of execution is not guaranteed
    ->addTask(new ClosureTask('2', function () {
        echo 2;
    }, ['1'])) 
    ->addTask(new ClosureTask('3', function () {
        echo 3;
    }, ['1']))

    // this is the last task in graph (there might be multiple leaf tasks)
    ->addTask(new ClosureTask('4', function () {
        echo 4;
    }, ['3', '2']));

Amp\Loop::run($taskDag); // will output "1234"

The visualization of the execution graph will be following:

     +------+
   +-+Task 1+-+
   | +------+ |
   |          |
+--v---+  +---v--+
|Task 2|  |Task 3|
+--+---+  +---+--+
   |          |
   | +------+ |
   +->Task 4<-+
     +------+

dag's Issues

Task Names

IMO task names should be assigned on addTask instead of inside the task implementations. That allows having one implementation used multiple times, e.g.

$x->addTask('copy.js', new CopyTask('**/*.js', 'build'));
$x->addTask('copy.css', new CopyTask('**/*.css', 'build'));

Review other similar libraries implementation

1. https://github.com/pinterest/pinball
2. https://github.com/intentmedia/mario
3. https://github.com/Netflix/suro
4. https://github.com/apache/incubator-airflow
5. https://github.com/spotify/luigi

Review deployment tools as future application

This is important to review possible applications for dag library which might highlight important requirements.

1. https://github.com/capistrano/capistrano
2. https://github.com/ansistrano/deploy
3. https://github.com/fabric/fabric
4. https://github.com/gulpjs/gulp