Question of the Day: What about all this asynchrony?

Naveen Chawla naveen.chwl at gmail.com
Tue Nov 7 19:46:55 UTC 2017


async functions create a new promise for you upon every invocation, which
you resolve via `await`, but that's all invisible in the background. It's
basically:

async function doMovesAsync(){
    moves.forEach(
         move=>{
              doMoveAsync(); //another async function
         }
    );
}

...so you can do regular programming, in async world. This is why I believe
it's more powerful than observables, thereby making them redundant.

When I say branching into multiple outputs, I do mean creating new data
that leaves the original data untouched.

On Tue, 7 Nov 2017 at 20:57 Michael Lewis <mike at lew42.com> wrote:

> Also, if you've made it this far, I think it's worth mentioning that these
> async strings are basically all you need for a realtime file system.
>
> File("newFile.ext").append(File("fileA"), File("fileB"),
> ...).transpile().save();
> // --> automatically watches, all inputs (fileA, fileB, etc), caches
> unchanged files, reapplies transforms, writes to file...
>
> Webpack and gulp are basically async plugin systems w/ transforms.
> They're just way too complicated.
>
> Simplify all the things.
>
> And while we're at it, why not make a realtime version control system?
> Not just for files, but for all the things (any data structure inside the
> app).  For example, if we have variable strings, could we enable a history
> on it?  Instead of branching onto a separate entity/value, could we branch *within
> *the string itself, so that we have an entire *verrsion tree *for any
> value?
>
> *What are the fundamental data structures in computer science?*
> The Boolean, obviously.  The Integer.  The String.
>
> Why not a realtime boolean?  I suppose that's just a boolean + change
> events.  What is a "change event"?  Just an array of functions.  But
> JavaScript functions are an abstract concept (compared to processor
> instructions).  What do functions look like at the processor level?
> They're compiled with all the dependent values, right?  How many processor
> ticks does the average line of JavaScript use?
>
> I feel like all languages could boil down to a very small set of
> fundamental data structures, and maybe a slightly larger set of specialized
> data structures.
>
> What are the different types of circuits in a process?  I understand
> (roughly) the basic logic gates, but is there specialized circuitry for
> specialized data structures?  What if those fundamental data structures
> were optimized at the circuitry level?
>
> What if we can optimize our programs to run as nearly instantly as
> possible?  Most scripts are *mostly *instant - at least, there's no
> external input.  For any process that's *nearly* instant, couldn't it
> actually be instant?  In other words, 1 tick of the processor?  Load up all
> the registers with the necessary values, and shine the light down those
> transistors/logic gates, so that we arrive at our result, instantly?
>
> I really feel like this is possible.  Like I mentioned earlier, I've never
> compiled a lick of code in my life, and have very little understanding of
> those things.  But from my sense of JavaScript, it's far from instant.  How
> many processor ticks per line of JavaScript code, on average?
>
>
>
>
> Is anyone still listening?
>
> On Tue, Nov 7, 2017 at 8:47 AM, Michael Lewis <mike at lew42.com> wrote:
>
>> I'm not experienced in async/await enough to know what "using async
>> functions to process [streams]" would look like.
>>
>> You would have to create a new promise for every iteration?  Even if
>> performance isn't an issue, it just doesn't make sense to me.  It's like,
>> you could use `obj.value = "my string"` instead of `var myString = "my
>> string"`, and it will work.  And the performance difference is negligible.
>> But, it just doesn't make as much sense...
>>
>> *Branching vs Mutation*
>> The point you bring up regarding "branching the stream into multiple
>> outputs" is another fundamental concept in programming (that I'm still
>> trying to wrap my head around).  Basically, does an operation (aka a
>> method) operate on the original data, or fork/branch, preserving the
>> original, and creating a clone to apply the transform to.
>>
>> For example, arr.push() manipulates (mutates) the original array, but
>> arr.slice() branches, giving you a brand new array, leaving the underlying
>> array untouched (immutable).
>>
>> This has always been an area of confusion for me.  Which methods are
>> mutators, and which are immutable?
>>
>> *Async Strings*
>> An interesting away to look at all this async stuff, is to consider
>> strings, and their operations (methods), in an asynchronous way.  How can a
>> string be asynchronous?  Just let it change over time, and broadcast change
>> events.
>>
>> What if you compose a string with several pieces:
>>  asyncParentStr.append(asyncStrA, asyncStrB, asyncStrC).
>>
>> Each asyncString can have change events, and will propagate changes to
>> anyone depending on it.   asyncStrB.set("new value") will trigger
>> asyncParentStr.change() event.
>>
>> I feel like this is fundamental functionality that is lacking from
>> JavaScript.  Now that we have `const`, shouldn't `var` automatically set up
>> change events for that "var"?
>>
>> *Async transforms*
>> But lets say we do asyncParentStr.append(asyncStrA,
>> asyncStrB.replace("hello", "goodbye"), asyncStrC).
>>
>> Now we have the question: do we want this .replace() to be a "live"
>> transform?  When we asyncStrB.set("hello world"), does it re-apply the
>> replace?  I think there are many use cases for both: mutate the original
>> asyncStrB, so that all references to this value also exhibit the
>> transform.  And also the alternative, the immutable, branching kind of
>> transform, where you don't mutate the underlying value, and instead are
>> branching.
>>
>> This concept is also the core concept of version control: do we continue
>> down the same path, or branch off?
>>
>> *GUIs will prevail*
>> You can try and create different methods ( ._replace() vs .$replace() )
>> to represent transform vs branching (I don't know which looks more like
>> which).  But, in the end, the GUI will prevail.  Artists can dream about
>> how to envision these version trees, and perfect the GUI/experience.  The
>> code interface just can't compete with GUI, in the long run.
>>
>> I suppose, its necessarily true that the API preceeds the GUI.
>>
>> API before GUI, but GUI all the things.  That's my new motto.
>>
>> *What if variables were automatically async, and referential? *(As
>> opposed to `const` that could be the immutable flavor)
>> var str = "hello world";
>>
>> str.replace("hello", "goodbye"); // transforms `str` var "in place"
>> log(str) // "goodbye world"
>>
>> str = "hello cruel world"; // transform is reapplied
>> log(str) // "goodbye cruel world"
>>
>> This will never happen, but it shows the fundamental difference in
>> logic.  Both are logical/useful...
>>
>> On Tue, Nov 7, 2017 at 8:08 AM, Naveen Chawla <naveen.chwl at gmail.com>
>> wrote:
>>
>>> For me the future is async functions (the present actually). I asked a
>>> question about possible support for async streams / observables here:
>>> https://esdiscuss.org/topic/stream-async-await and I realized that my
>>> use case was much better served by just using async functions to process
>>> each input value in the stream.
>>>
>>> I think using async functions is much more powerful than "observables",
>>> since it allows you to easily branch the stream off into multiple outputs
>>> etc. Using Promise.all etc. is also trivial to use where desired, etc.
>>>
>>> Furthermore, async functions allow while/for loops that include other
>>> async function calls, and this looks like programming with regular
>>> functions, so it's trivial to set up asynchronous iteration, and/or
>>> infinite event processing, etc., even without the new "async iteration"
>>> proposal.
>>>
>>> On Tue, 7 Nov 2017 at 17:25 Michael Lewis <mike at lew42.com> wrote:
>>>
>>>> The email wasn't about my kids, and you don't have to read it (duh).
>>>> If your time is so valuable, maybe you shouldn't be picking fights with
>>>> rambling parents.
>>>>
>>>> Where is the list of approved topics?
>>>>
>>>> On Tue, Nov 7, 2017 at 5:44 AM, Bob Myers <rtm at gol.com> wrote:
>>>>
>>>>> I'm confused. You don't have time to read "The General Theory of
>>>>> Reactivity", yet (1) you have time to write this long, rambling email about
>>>>> your kids, and (2) expect people on this mailing list to spend their
>>>>> valuable time reading it?
>>>>>
>>>>> Please stay on topic for the list.
>>>>>
>>>>> Bob
>>>>>
>>>>> On Tue, Nov 7, 2017 at 4:48 PM, Michael Lewis <mike at lew42.com> wrote:
>>>>>
>>>>>> Good morning JavaScript world,
>>>>>>
>>>>>> Maybe I'll start my mornings with a cup of coffee, and a discussion
>>>>>> prompt.  We'll see how long it lasts.  It's 4:39am.  I live in Aurora,
>>>>>> Illinois, about an hour outside of Chicago.  My kids will wake up in an
>>>>>> hour or two, so I don't have long, and should be working on my framework
>>>>>> anyway.
>>>>>>
>>>>>> *So much asynchrony*
>>>>>> There are callbacks, promises, async/await.  We have streams in
>>>>>> node.js.  There are libraries like RxJS for Observables (that are basically
>>>>>> streams?).
>>>>>>
>>>>>> What's the end game?  What will our children's children be learning
>>>>>> in 100 years?  Let's reduce these pieces, distilling them into their
>>>>>> simplest components.
>>>>>>
>>>>>> This <https://esdiscuss.org/topic/promises-vs-streams> is an
>>>>>> interesting thread (from es-discuss) regarding asynchrony, which references
>>>>>> Kris Kowal's General Theory of Reactivity
>>>>>> <https://github.com/kriskowal/gtor/>, which is too long for me to
>>>>>> dig into at this point in my life.
>>>>>>
>>>>>> The disappointing part, is that this community (who has mastered
>>>>>> asynchrony) doesn't feel like there are any shortcomings, and so we
>>>>>> continue onward without fixing the mess.
>>>>>>
>>>>>> Oh, and generators?  I don't fully understand these things.  Do they
>>>>>> work with promises?  Can you use a generator to process a stream?  How do
>>>>>> generators work with or compare to async/await?  Who knows...
>>>>>>
>>>>>> I think it's safe to say that asynchrony is a confusing mess.  *But
>>>>>> it shouldn't be.   *The concept of "do it now" or "do it later" is
>>>>>> as easy as 123.
>>>>>>
>>>>>> Recently, I read through Jake Archibald's JavaScript Promises: an
>>>>>> Introduction
>>>>>> <https://developers.google.com/web/fundamentals/primers/promises>.
>>>>>> I really enjoy Jake Archibald's writing.  He makes JavaScript less boring.
>>>>>> But wow, understanding promises in their full depth is really complicated.
>>>>>> Sure, a simple promise is more or less a callback, easy peasy.  But once
>>>>>> you start composing parallel and series tasks, add error handling, and try
>>>>>> to understand the control flow - it's a lot.
>>>>>>
>>>>>> I feel like Promises could automatically *render a diagram *when
>>>>>> using them.  In Jake's very practical example (request a list of chapters,
>>>>>> load all chapters in parallel, then append them to the page in order)
>>>>>> there's a lot going on, to say the least.  Wouldn't it be nice to see a
>>>>>> diagram of these tasks?  A timeline maybe?
>>>>>>
>>>>>> Imagine debugging a complex sequence of async actions.  And you have
>>>>>> no idea which piece is failing.  Using the console to log values, and
>>>>>> trying to step through the code with the debugger are two of your basic
>>>>>> approaches.  But honestly, neither of these really *show *you what's
>>>>>> going on.
>>>>>>
>>>>>> Chrome Dev Tools has an awesome timeline GUI.  I've spent an hour
>>>>>> here or there tinkering with it, but can't make sense of a lot of it.
>>>>>> There are 100's if not 1000's of very generic blocks that show up on the
>>>>>> timeline, that don't clearly identify what they are.  And I don't believe
>>>>>> there's any way to visualize promises on this timeline.
>>>>>>
>>>>>> *The problem with Promises*
>>>>>> I want to create a file system framework for node.  I'd like to make
>>>>>> watching the files for changes a default feature.  The problem with
>>>>>> promises, is that you can't re-resolve them.
>>>>>>
>>>>>> So I'm basically left with streams, or plain old callbacks.  Or
>>>>>> trying to recreate the promises every time they resolve...
>>>>>>
>>>>>> What's the end game?  100 years from now?
>>>>>>
>>>>>> Frankly, this is the most important question.  I feel like if we take
>>>>>> a step back, and try to solve these problems for the long term, we'd be
>>>>>> better off.
>>>>>>
>>>>>> And so, it's 5:15.  Well done, Michael.  Well done.
>>>>>>
>>>>>> *The Future*
>>>>>> If anyone has made it this far, I'm going to tell you a quick summary
>>>>>> of my plan:
>>>>>>
>>>>>>    1. make an ultra-simple web framework (almost done?)
>>>>>>    2. use that framework to make a CMS to kill WordPress
>>>>>>    3. turn that CMS into a web OS that does everything a real OS can
>>>>>>    do, only better
>>>>>>    4. turn that web OS into a real, bare metal OS
>>>>>>    5. make lots of amazing (useful) software (like photoshop,
>>>>>>    blender, after effects, CAD, etc)
>>>>>>
>>>>>> Software development is sluggish.  Most software is painful to use.
>>>>>> Windows, Photoshop/Illustrator, many websites...  Open source software
>>>>>> doesn't get the funding/momentum it needs to really kill these proprietary
>>>>>> alternatives.  We need to change that.  I'm going to change that.
>>>>>>
>>>>>> Stay tuned.
>>>>>>
>>>>>> _______________________________________________
>>>>>> es-discuss mailing list
>>>>>> es-discuss at mozilla.org
>>>>>> https://mail.mozilla.org/listinfo/es-discuss
>>>>>>
>>>>>>
>>>>>
>>>> _______________________________________________
>>>> es-discuss mailing list
>>>> es-discuss at mozilla.org
>>>> https://mail.mozilla.org/listinfo/es-discuss
>>>>
>>>
>>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.mozilla.org/pipermail/es-discuss/attachments/20171107/2c072868/attachment-0001.html>


More information about the es-discuss mailing list