debounceTime doesn't work with Subject

[tandu]

We had a discussion here angular/angular#5925. It seems that throttleTime works with Subject and debounceTime doesn't work with Subject.
throttleTime: http://jsfiddle.net/bo6ynqf7/1/
debounceTime: http://jsfiddle.net/8g968xh7/1/

[kwonoj]

Would you elaborate bit more detail of 'doesn't work'? I tried this
http://jsfiddle.net/8g968xh7/2/ which has slightly modified such as below

const { Subject } = Rx;

const button = document.querySelector('button');
const pre = document.querySelector('pre');

var sub = new Subject()

sub.debounceTime(1000)
    .subscribe(_=> pre.innerHTML += _);

button.addEventListener('click', function(){
  sub.next('a');
});

and could see debounceTime emits. Does doesn't work means emitting behavior is not expected, or is it not emitting at all?

[tandu]

Hm, it seems that sub.next() is working for throttleTime but for debounceTime it should pass any value. So for debounceTime sub.next() is not fired but if you send anything like sub.next(0) or sub.next('a') it works as expected.

[tandu]

By doesn't work I meant it was not emitting at all.

[kwonoj]

Ah, that's currently expected behavior from implementation of debounceTime(). Looking at implementation,

...
if (this.lastValue != null) {
      this.destination.next(this.lastValue);
      this.lastValue = null;
    }

it only emits if it explicitly have values to be emitted. So while subject emits next without any value, debounceTime simply skips it since there's no value.

note : just clarify, 'currently expected behavior ' simply means it works in those way for now.

[chicoxyzzy]

looks inconsistent

[kwonoj]

Behavior perspective it does. Few things to be considered are,

• Subject::next() requires value as paramater, it should not be skipped. So in case of above example, calling next(any) ensures operator behavior.
• Edge case in this issue is case of emitting undefined as value, since next() is same as next(undefined). This is case shows different behavior compare to RxJS4.

[staltz]

@kwonoj debounceTime should have a hasValue-based implementation, like e.g. distinctUntilChanged:

class DistinctUntilChangedSubscriber<T> extends Subscriber<T> {
  private value: T;
  private hasValue: boolean = false;

So this bug is legit it seems.

[kwonoj]

@staltz , Yes, @mattpodwysocki suggested same approach in current PR (#1083 (comment)). I'm updating PR to reflect suggestions.

[lock]

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.