Rx glitches aren't actually a problem | Hacker News Reader

Rx glitches aren't actually a problem

staltz.com

37 pointsjessaustin17 comments

The title initially intrigued me because I thought it was talking about errors in prescriptions; it's actually about ReactiveX.

I thought he was talking about reception glitches in radio.

EDIT: The article is about javascript. Everything below does not apply to javascript. You need more than one thread, at least, for any of these problems.

I strongly suspect that combining async streams will prove to be a problem in practice, whether we resolve glitches or not. This article requires quite subtle reasoning about async systems. In my experience this type of reasoning is rarely done effectively on real life projects.

I am not criticising Rx, I am interested in Rx. But I am always concerned when a project/community starts producing articles on very subtle topics and claims 'x isn't a problem if you just understand it properly'. Here properly probably means deeply, and deep knowledge in software is scarce. There are too many things to know, to know many things deeply.

Let me look at one example (and please comment if I'm really not understanding something).

So we consume, asynchronously, two streams. Because they are asynchronous we might hit some point where it's a bit uncertain where two outputs are related.

    errors            ---e1----e2------------------
    userActions       -u1------u2------------u3----
    analyticsMessages ---e1u1--(e1u2 e2u2)---e2u3--

That's a good example, and we fixed that problem with 'errors.withLatestFrom(...'. But, since we are asynchronous don't we also have to worry about this scenario.

    errors            ---e1-------------------e2--
    userActions       -u1--u2--u3-u4-u5-u6-u7-u8--
    analyticsMessages ---e1u1---------------e2u8--

Now, let me clarify what happened here. The error occurred after u2, but the errors thread got descheduled and the userActions thread kept on processing. I am not certain exactly what to expect from 'errors.withLatestFrom(userActions,...' here but I suspect that 'e2u8' may be the expected output.

So this error reporting system is fundamentally unreliable. If this is correct then I would say that combining Rx streams is not appropriate for this use case. This also suggests that although we managed to get rid of our glitches we still produced a terrible piece of software.

Do I have to wrong end of the stick? (The answer was yes, but it was good anyway)

> The error occurred after u2, but the errors thread got descheduled and the userActions thread kept on processing.

Your suggested case is contrived and doesn't apply to JavaScript (the examples were in JavaScript where I assume all these streams are in the same thread). I am not claiming withLatestFrom is the correct mechanism for determining order of events in multi-threaded assumptions. That's an orthogonal concern to the issue at hand.

> I strongly suspect that combining async streams will prove to be a problem in practice

Combining async results is a true problem and people often attempt to solve them with callbacks or heavy usage of Futures/Promises. Rx solves these problems easily, and I've witnessed a lot of developers confirm this when learning Rx. It's not hard to teach the difference between withLatestFrom and combineLatest and their intended use cases. Much easier than managing callback hell, which is what RxJS intends to replace.

(A great thing about HN is getting to talk with the authors directly)

I didn't recognise that you were talking chiefly about javascript. It is true that my example doesn't apply here.

Ok, then I would make one comment (and edit my first comment).

I would avoid the discussion of simultaneity at the start

    'Events in parentheses happen “simultaneously”.
     In practice they happen at slightly different times,     
     but separated by only a couple of nanoseconds, so
     people understand them to be simultaneous. Events (c1c2) 
     are called glitches and sometimes considered a problem
     because one would expect only c2 to happen.'

Because you are working in a single threaded simultaneous events just don't exist. The glitches you describe occur completely independently of timing, nanoseconds or otherwise. Drawing up diagrams where events occur at the same time is adding complexity which doesn't exist.

    errors            ---e1-------e2---------------
    userActions       -u1------u2------------u3----
    analyticsMessages ---e1u1--(e1u2 e2u2)---e2u3--



    errors            ---e1----e2------------------
    userActions       -u1----------u2--------u3----
    analyticsMessages ---e1u1--(e2u1 e2u2)---e2u3--

Both of these scenarios produce glitches (as I understand it). So we don't need any notion of 'simultaneous events' to create them.

Anyway - that is nitpicking. I enjoyed your article. I agree that the scenario described at top does not apply in javascript.

I too strongly dislike callbacks :)

I agree with the essence of what you say. Rx trades off simple semantics for a simple implementation. Given the issues such as the above I think it is generally the wrong tradeoff. Simpler semantics do not require a great deal more work, as demonstrated by numerous other systems that have implemented them.

What are the "numerous other systems" that you had in mind? Links would be greatly appreciated!

Some alternatives are linked in the article. Other systems include FrTime / Flapjax (glitch-free push based systems in Racket and JS respectively), and the work on FRP.

Well, yes, but those don't seem objectively superior. They all have problems to do with it being too easy to accidentally introduce space leaks or time leaks...

(At least as far as I recall. If there's some documentation/papers that suggest otherwise, please link!)

EDIT: I should say: FRPNow! seems to have cracked a bit of this puzzle and Neel Krishnaswami had a recent-ish (2014?) paper on avoiding both space and time leaks with static checking, but I don't think we have anything approaching the mainstream, unfortunately.

Only the Haskell systems have space leaks. The others don't.

Links of PDFs/whatever, plz!

(I'm not being faceitious.)

An issue with this contrived example, as well, is that Rx has an error flow at the stream level, so typically you are more likely to build things similar to:

   userActions(combined)
    \
     \--u1-----------------e1--X
      \--u2---------------------------e2--X
       \--u3--...
        \--u4--...
         \--u5--...
          \...
           e1
            \
             X

While you could do your analytics message on the combined diagonal stream, you'd probably want to do it on the horizontal streams.

Glitches as demonstrated via the diamond pattern can be avoided by traversing the graph in topological order. This was demonstrated in Greg Cooper's PhD thesis: http://cs.brown.edu/~greg/

Transactions, as I understand them, solve a different issue. This is the issue of multiple inputs to the graph occurring at the same time.

> Glitches as demonstrated via the diamond pattern can be avoided by traversing the graph in topological order.

This is correct for DAGs which only propagate value (as in Javelin[1] and my own library DerivableJS[2]), but for graphs which propagate events (as in Rx), topological sorting would only work for those parts of the graph which are effectively propagating value. Events don't have an inherent dedupe operation, so it is very difficult to even imagine ways in which glitch avoidance could be automatically enforced. It would certainly require semantic program analysis.

Personally I think we should be avoiding the proliferation of events (as encouraged by Rx enthusiasts) for exactly this reason. Their imperative nature makes them very difficult to reason about.

[1]: https://github.com/hoplon/javelin

[2]: https://github.com/ds300/derivablejs

The event/value distinction is an interesting viewpoint I hadn't considered. Thanks for writing this.

They aren't a problem if the operations at the leaves of the graph are idempotent, which is usually the case if these operations are UI repaints.

If the operations are for other effects - such as Ajax calls - then glitch elimination is handy. Eliminating glitches at the dataflow level prevents debounce logic from leaking out to the code performing the effects, preserving the clarity and generality of that code.

I find glitch elimination most useful when the dataflow graph is value-propagated, vs. event-propagated, as it is in the dataflow library I helped implement, Javelin - https://github.com/hoplon/javelin/

Javelin also supports transactional input, another feature that's helpful when building dataflow graphs on which effects other than UI-repaint are hung.

It's a feature.

Reply on news.ycombinator.com