Document why this is better than yield array

[fzaninotto]

Why is co-parallel better than simply yielding an array within co? I can't find any documentation about that, and my tests don't show any particular advantage:

// tasks is an array of 200 async tasks simulated by setTimeout, cf #3
var start = Date.now();
yield parallel(tasks, 10);
console.log('total used %dms', Date.now() - start);

This takes 170ms on average, while:

// tasks is an array of 200 async tasks simulated by setTimeout, cf #3
var start = Date.now();
yield tasks;
console.log('total used %dms', Date.now() - start);

This takes 117ms on average

[juliangruber]

by limiting the maximum concurrency you can control both cpu and memory usage of the task

[fzaninotto]

OK, so parallel is supposed to be slower than yield array by design, is that so?

[juliangruber]

co-parallel will be faster when the array you yield is too big so the node proc would reach its limits, and it will help to ensure not too many sockets are open at once etc.

[kilianc]

what I don't like about this is that we have N * Array.length functions to allocate instead of having a better co-each, am I correct?

[kilianc]

just to be more precise: yield* parallel(list, action, max)