Hooking up afterViewerChanged/beforeViewerChanged

[lob743]

Hello,

Previously, in v7, we used the events afterViewerChanged/beforeViewerChanged a great deal to reload stuff upon swapping viewer mode. Now, in v8, it appears these are not hooked up, in the sense that events are never raised. Is there a reason why this was dropped after the transition to v8?

In my attempt to hook these up, I encountered a problem.

In TerriaViewer.ts, I raised the beforeViewerChanged event inside of destroyCurrentViewer(), right where it says "Destroying viewer:". This works fine. However, when attempting to raise afterViewerChanged inside of currentViewer, I get an error for "Cycle detected in computation". I cannot seem to understand why one works and not the other. I would greatly appreciate any insight!
Thanks!

[zoran995]

Already reported in #4752. You probably can use mobx observables to get out around that. I used a combination of reaction and when (not sure why both were needed, long time since it's implemented) to create manual subscribtion to viewer events and it seems to work fine.

  constructor() {
    reaction(
      () => this.cesiumViewer,
      () => this.addUpdateSubscription(),
      {
        equals: (a, b) => {
          return isDefined(b);
        }
      }
    );

    reaction(
      () => this.leafletViewer,
      () => this.addUpdateSubscription(),
      {
        equals: (a, b) => {
          return isDefined(b);
        }
      }
    );

    when(
      () => isDefined(this.cesiumViewer),
      () => this.addUpdateSubscription()
    );
    when(
      () => isDefined(this.leafletViewer),
      () => this.addUpdateSubscription()
    );
  }

  @computed
  get cesiumViewer() {
    return this.terria.cesium;
  }

  @computed
  get leafletViewer() {
    return this.terria.leaflet;
  }

Inside addUpdateSubscription there is a check which viewer is currently active and perform side effects. But it should be possible to separate them completely.

[lob743]

This solution worked wonders as a work-around!
I see that you created a PR on this matter with a more permanent solution (#6190).
Thanks a bunch for solving this @zoran995, very helpful!