Bug 723555 - Implementation of promise abstraction

[Gozala]

This change change brings low level API for working with promises in the SDK.

https://bugzilla.mozilla.org/show_bug.cgi?id=723555

[ochameau]

You will have to choose between is and has :p

[Gozala]

Addressed all the review comments.

[ochameau]

These two code examples aren't really helpfull. Both are identical whereas we try to explain different behaviors. I'm not sure examples are really important. These two sentences are quite clear.

[Gozala]

First argument to then was supposed to be null.

[ochameau]

Why not using [f, this].concat(Array.slice(arguments)) instead?
or var args = Array.slice(arguments); [f, this].concat(args)
It highlights the issue of arguments, and we are used to Array.slice in our codebase.

Is it possible to process only arguments to reduce. It is really really hard to get why arguments and f/this are processed into the reduce function!? It would be way easier to follow with something like this:

  let self = this;
  return
    // create real array for `arguments`, with `reduce` method
    Array.slice(arguments).
    // reduce it via `promisedConcat` to get promised array of fulfillments
    reduce(promisedConcat, resolve([], prototype)).
    // finally map that to promise of array with `f`
    then(function(args) { return f.apply(self, args); });

Note that execute shortcut doesn't necessary improve comprehension of this code.

[Gozala]

Why not using [f, this].concat(Array.slice(arguments)) instead?
or var args = Array.slice(arguments); [f, this].concat(args)
It highlights the issue of arguments, and we are used to Array.slice in our codebase.

This pull request was importing library that is not written for moz-js in standard js you don't have Array.slice. Also slice is suboptimal for such hotspot codepaths. I have being doing diff profiling to identify fastest option.

[Gozala]

Note that execute shortcut doesn't necessary improve comprehension of this code.

No I agree it's harder to read this code, but it's way more optimal and this function something that is frequently used, so reducing closures introduced on each call is very important. Also see comments on line 162 that explains explain that.

[ochameau]

r+, mainly nits. Feel free to request another round if you feel too, but I think it can land without a new one.
Please consider stashing the history before landing, as we do not really care about all back and forth revisions with various namings.

[ochameau]

I meant this link. This link isn't recognized by the doc parser.