execute a function parallely, while completing execution of rest of the code

Question

I have a code snippet in nodejs like this:

in every 2 sec, foo() will be called.

function foo() 
{
    while (count < 10) 
    {
        doSometing()
        count ++;``
     }
}

doSomething()
{
   ...
}

The limitation is, foo() has no callback. How to make while loop execute and foo() completes without waiting for dosomething() to complete (call dosomething() and proceed), and dosomething() executes parallely?

Answer 1

I think, what you want is:

function foo() 
{
    while (count < 10) 
    {
        process.nextTick(doSometing);
        count ++;
     }
}

process.nextTick will schedule the execution of doSometing on the next tick of the event loop. So, instead of switching immediately to doSometing this code will just schedule the execution and complete foo first.

You may also try setTimeout(doSometing,0) and setImmediate(doSometing). They'll allow I/O calls to occur before doSometing will be executed.

Passing arguments to doSomething

If you want to pass some parameters to doSomething, then it's best to ensure they'll be encapsulated and won't change before doSomething will be executed:

setTimeout(doSometing.bind(null,foo,bar),0);

In this case doSometing will be called with correct arguments even if foo and bar will be changed or deleted. But this won't work in case if foo is an object and you changes one of its properties.

What the alternatives are?

If you want doSomething to be executed in parallel (not just asynchronous, but actually in parallel), then you may be interested in some job-processing solution. I recommend you to look at kickq:

var kickq = require('kickq');

kickq.process('some_job', function (jobItem, data, cb) {
  doSomething(data);
  cb();
});

// ...

function foo() 
{
    while (count < 10) 
    {
        kickq.create('some_job', data);
        count ++;
     }
}

kickq.process will create a separate process for processing your jobs. So, kickq.create will just register the job to be processed.

kickq uses redis to queue jobs and it won't work without it.

Using node.js build-in modules

Another alternative is building your own job-processor using Child Process. The resulting code may look something like this:

var fork = require('child_process').fork,
    child = fork(__dirname + '/do-something.js');

// ...

function foo() 
{
    while (count < 10) 
    {
        child.send(data);
        count ++;
     }
}

do-something.js here is a separate .js file with doSomething logic:

process.on('message', doSomething);

The actual code may be more complicated.

Things you should be aware of

Node.js is single-threaded, so it executes only one function at a time. It also can't utilize more then one CPU.

Node.js is asynchronous, so it's capable of processing multiple functions at once by switching between them. It's really efficient when dealing with functions with lots of I/O calls, because it's newer blocks. So, when one function waits for the response from DB, another function is executed. But node.js is not a good choice for blocking tasks with heavy CPU utilization.

It's possible to do real parallel calculations in node.js using modules like child_process and cluster. child_process allows you to start a new node.js process. It also creates a communication channel between parent and child processes. Cluster allows you to run a cluster of identical processes. It's really handy when you're dealing with http requests, because cluster can distribute them randomly between workers. So, it's possible to create a cluster of workers processing your data in parallel, though generally node.js is single-threaded.