Rendering (actually not rendering, just task) could happen multiple times (in multiple micro tasks), getterThrottle could cache wrong value

[3cp]

When clicking "update" button, local i is updated, upper level j is updated (ea calls subscribers in sync, not async mode).

The problem is local j is not updated immediately, it will be updated in a micro task due to the 2 way binding.

local get message() is rendered twice. First time, by local i change in a micro task, then by local j change in another micro task.

Using @getterThrottle() here will cache wrong value "i:2 j:1", the second update "i:2 j:2" is ignored by the cache.

app.js

import {EventAggregator} from 'aurelia-event-aggregator';
import {inject} from 'aurelia-framework';

@inject(EventAggregator)
export class App {
  i = 1;
  j = 1;

  constructor(ea) {
    this.ea = ea;
    this.ea.subscribe('ttt', () => {
      this.j += 1;
    })
  }
}

app.html

<template>
  <require from="./test"></require>
  <test i.bind="i" j.bind="j"></test>
</template>

test.js

import {EventAggregator} from 'aurelia-event-aggregator';
import {inject, bindable, bindingMode, computedFrom} from 'aurelia-framework';
import {getterThrottle} from 'aurelia-getter-throttle';

@inject(EventAggregator)
export class Test {
  @bindable({defaultBindingMode: bindingMode.twoWay}) i;
  @bindable({defaultBindingMode: bindingMode.twoWay}) j;

  constructor(ea) {
    this.ea = ea;
  }

  update() {
    this.i += 1;
    this.ea.publish('ttt');
  }

  // @getterThrottle()
  @computedFrom('i', 'j')
  get message() {
    const m = `i:${this.i} j:${this.j}`;
    console.log('get message: ' + m)
    return m;
  }
}

test.html

<template>
  <button click.trigger="update()">Update</button>
  <p>${message}</p>
</template>

[3cp]

By the time of first getter call, the 2 micro tasks were already created. I could not find a way to inject another micro task (to clear cache) between the 2.

[3cp]

@bigopon any idea to work around?

[bigopon]

Aurelia micro task queue is based on std browser micro task, probably code for micro task queue can be extracted to create a parallel queue that will execute correctly, as they all resolves to underlying browser queue.